Abstract: With the national goals of "dual carbon" and "building a new power system with new energy as the main body", clean and digital have increasingly become the urgent needs of the power system. Based on advanced control, measurement, communication and other technologies, virtual power plants aggregate massive diversified distributed resources on the demand side, which can provide necessary flexibility support for the power system through diversified regulation methods, and help renewable energy consumption at the same time. Digital twin technology uses digital technologies such as big data, cloud computing, and artificial intelligence to virtually model the characteristics, behaviors, processes and performance of physical entities of distributed resources, which is an ideal way to achieve virtual power plant operation optimization. Based on the concept, architecture and characteristics of digital twin, this paper proposes a digital twin virtual power plant system, elaborates the connotation of digital twin virtual power plant, constructs the overall architecture of digital twin virtual power plant based on physical entities, digital twins, twin data, connections and intelligent applications, analyzes relevant key technologies, and summarizes and prospects its future application practices.
0 Introduction
The mainland's goals of "dual carbon" and "building a new power system with new energy as the main body" will promote the rapid evolution of the power system to adapt to large-scale and high-proportion renewable energy. However, large-scale intermittent renewable energy is connected to the grid from the demand side, transmitting uncertainty to the main grid and making it more difficult for the power system to maintain a balance between supply and demand[1].
In the future, large-scale grid-connected power systems with distributed energy resources (DERs) such as demand-side distributed wind power, rooftop photovoltaic, energy storage, electric vehicles, and flexible loads will present the characteristics of complex structure, diversified equipment and complex technology. The traditional "source-load-and-load" mechanism model and optimal control method will be difficult to meet the requirements of power grid operation optimization and system flexibility, and it is urgent to provide the necessary flexible adjustment capabilities with the participation of demand-side distributed resources [2].
In recent years, a new round of electricity market-oriented reform and the promotion of new Internet technology, communication technology and intelligent power consumption technology have made virtual power plant (VPP) technology provide new ideas for the consumption of multiple DERs and the digital transformation of power system on the demand side. Through the aggregation and regulation of massive multiple DERs, virtual power plants not only have the characteristics of stable output and batch power sales of traditional power plants, but also have good flexibility, providing management and auxiliary services for transmission and distribution networks and improving the revenue of DERs.
However, the current research on virtual power plants mainly focuses on the operation and regulation strategy of virtual power plants that aggregate multiple DERs and the coordination of DERs participation in power market transactions in the market environment. In the future, with the rapid increase of demand-side market entities, frequent changes in business rules, rapid growth of external data interaction, and increased system security risks, it is urgent to study how to deeply integrate and apply advanced information and communication technology, big data analysis, artificial intelligence methods and control technology and energy technology, empower virtual power plants through digital means, continuously improve the holographic perception and flexible control ability of demand-side massive multiple DERs in the transmission and distribution network, and build a new business model for virtual power plants to participate in the power market. And from the demand side, it provides flexible control means for the smooth operation of the power grid [3].
In recent years, the digital economy represented by a new generation of information technology has become a hot spot in global economic development, and European and American countries have proposed digital economy strategies; At the same time, under the guidance of the new development concept of innovation, coordination, green, openness and sharing, the mainland also attaches great importance to the development of digital economy[4]. Among them, digital twins combined with technologies such as "big cloud migration intelligent chain" provide new ideas for digital transformation in smart cities, smart healthcare, aviation and other fields. In the past two years, some literature has studied the application of digital twin technology in the energy and power industry, and gradually discussed the enabling role of digital twin based digital technology in the energy and power industry. The use of digital technology, drawing on the concept of the Internet, will accelerate the digital transformation of the power system from energy production to customer service [5-6].
In general, the fidelity, real-time and closed-loop characteristics of digital twin technology determine that it is particularly suitable for complex systems with asset-intensive and high reliability requirements [7], while virtual power plants are comprehensive complex systems that integrate massive multiple DERs on the demand side, which is highly consistent with the application field of digital twin technology. However, at present, the research of digital twin technology in the field of virtual power plants is still in its infancy, such as the team of Professor Ai Qi of Shanghai Jiao Tong University to initially explore the application of digital twin technology in smart microgrids [8].
Based on the research basis of digital twin technology in the field of energy and power system, this paper further proposes a virtual power plant system based on digital twin. As a key digital information technology tool, digital twin can effectively play its characteristics of two-way interaction, scalability, real-time, fidelity and closed loop, play its advantages in virtual power plant model construction, data collection, simulation analysis, simulation prediction, etc., and promote the digitalization, informatization and intelligent development of virtual power plant. With intelligence and digitalization as the link, the digital twin will effectively connect the interaction between demand-side users, aggregators, and transmission and distribution systems, and establish an interactive feedback mechanism between DERs and the power system. Taking the comprehensive utilization of energy as a means, tap the potential of multi-energy complementary synergy and comprehensive demand response, release the flexibility of demand-side adjustment, and improve system efficiency through refined management.
This paper first analyzes the challenges and future development directions of virtual power plants in the process of digital transformation. Secondly, based on the technical characteristics of digital twin, the definition of digital twin virtual power plant is explained for the first time, and the system framework based on DERs physical entities, digital twins, twin data, interactive connections and intelligent applications is further proposed. Thirdly, the key technologies involved in the digital twin virtual power plant are systematically analyzed to provide theoretical support for its subsequent construction and implementation. Finally, its application practices in modeling and simulation, planning, scheduling, health management and value creation are prospected.
1 Virtual power plant
1.1 Research status of virtual power plants
1.1.1 Research Status of Virtual Power Plant Technology
Since Dr. Awerbuch proposed the term "virtual power plant" in 1997, many scholars at home and abroad have interpreted virtual power plants in many ways, but there is no authoritative definition [9-12]. Based on the elaboration of the concept of virtual power plant in the literature, virtual power plant can be defined as the effective aggregation of DERs as a unified overall participation in the power market and power grid optimization regulation and operation of the "source-grid-load-storage" coordinated management system through advanced information and communication technology, intelligent metering and optimization control methods. From a technical point of view, the current research on virtual power plants mostly focuses on two aspects: operation control strategies and participation in electricity market transactions.
1) The operation regulation strategy of virtual power plant mainly studies its aggregation of multiple DERs, and under the condition of meeting the operating characteristics of various units and the constraints of safe operation of distribution network, the maximum revenue or minimum operating cost/carbon emission is the optimization goal, and the DERs capacity is allocated or the output is optimally dispatched, and there are a large number of relevant research results. For example, on the basis of considering the complementary and cooperative operation characteristics of water, wind and solar, the literature [13] studies the power generation declaration strategies of virtual power plants in isolated island and grid-connected environments. Literature [14-16] discusses the allocation of virtual power plant power capacity under risk constraints and the strategies of participating in power market purchase and sales and energy management. Literature [17] proposes a risk-averse optimal quotation model for virtual power plants to trade in the joint market of electric energy and standby capacity. The literature [18-19] adopts the two-layer optimization method to compare and analyze the overall economic benefits of virtual power plants and the net income of each distributed unit under different capacity ratios.
Considering the uncertainty of distributed renewable energy [20], the literature [21] transforms the uncertainty optimization problem into a stochastic adaptive robust optimization problem by modeling the uncertainty of wind power output and market electricity price prediction. The literature [22] transforms the model into a robust optimization problem by modeling the uncertainty of wind power by confidence intervals and modeling the uncertainty of electricity price prediction by multi-scenario method. In addition, based on the "cloud-group-end" virtual power plant dispatching framework, the literature [23] proposes a standardized modeling method for multivariate DERs and an interactive dispatching strategy for the power grid.
2) Virtual power plants participating in market transactions can quickly respond to changes in market electricity prices based on their flexible and adjustable characteristics. With the gradual improvement of the electricity market mechanism, virtual power plants will promote the liberalization of the electricity market, increase market flexibility, and actively guide users to participate in peak shaving and frequency regulation by participating in the market of day-ahead, real-time, balanced market, and ancillary services. For example, in the literature [24], virtual power plants are defined as aggregators centered on market integration and services, and research is used to participate in market transactions through aggregate DERs to reduce the amount of curtailed wind and light; Literature [25] Based on grid security constraints, a virtual power plant optimization dispatching strategy based on non-cooperative game is proposed.
However, the above research only considers the strategy of virtual power plants to participate in energy market transactions, and with the development of advanced information technology and control methods, virtual power plants will continue to enhance their ability to control DERs, participate in a wider range of electricity market transactions[26], and enhance the income of DERs by providing multiple auxiliary services. For example, by reasonably evaluating the backup capacity and reactive power regulation value of virtual power plants, a transaction decision-making method for virtual power plant reactive power auxiliary services based on multi-level collaborative method is proposed, so that it can support the safe and stable operation of the power grid by participating in system voltage regulation.
1.1.2 Engineering Practices
Europe and the United States have begun to study virtual power plants since the beginning of this century, among which Europe mainly focuses on the grid-connected and power market operation of distributed generation resources on the power side, in order to improve the friendly and intelligent interaction of distributed power grid, so that distributed power plants can participate in the power market more fully, safely, efficiently and more reliably, and create a sustainable and stable development business model. For example, the EU FENIX project (which divides virtual power plants into commercial virtual power plants and technical virtual power plants) and the EDISON project in Denmark [28], the E-Energy project in Germany and the PREMIO demonstration project in France [29].
North America, represented by the United States, focuses on the load side, and realizes real-time supply and demand balance while improving renewable energy consumption and comprehensive energy utilization through automatic demand response and energy efficiency management[12]. Compared with foreign countries, domestic research on virtual power plants started a little later. However, in recent years, the mainland has also carried out the preliminary practice of virtual power plants in Shanghai, Hebei and other places, among which, Shanghai Huangpu District piloted the world's first "commercial building virtual power plant", through control and communication technology, the ability of many energy-using equipment to reduce load as virtual output, participate in market and power grid dispatching and operation[30]; The State Grid "Jibei Virtual Power Plant" was put into operation, becoming the first demonstration project in China to operate in a market-oriented manner [31]. In addition, Continental's practice in virtual power plants has also combined with the construction of ubiquitous power Internet of Things, and initially carried out some exploration of demand-responsive virtual power plants focusing on commercial buildings and flexible loads[32].
1.2 The development direction of virtual power plants in the digital era
1.2.1 Virtual power plant architecture research
Due to the scattered location and wide distribution of DERs aggregated by virtual power plants, it is necessary to carry out the underlying design according to the cell-tissue architecture, realize optimal aggregation through dynamic combination, mutual benefit, and mutual benefit, and realize the global and distributed optimization of internal resources of virtual power plants according to the hierarchical partitioning mode. In addition, combined with the construction requirements of new power systems, the framework research of virtual power plants also needs to integrate digital methods such as big data, cloud computing, and artificial intelligence to promote the construction of virtual power plant architecture system with "electricity" + "computing power", and promote the networking, intelligence and digital transformation of virtual power plants.
1.2.2 Coordinate and optimize the operation of virtual power plants
The diversified and massive characteristics of internal resources of virtual power plants and the flexible aggregation mode make it difficult for the traditional centralized control method of integrating global operation data for unified scheduling by dispatching center to achieve flexible and efficient dispatching control and effective coordination of multiple resources. At the same time, due to the weak predictability and randomness brought by a large number of DERs, the economy of the day-ahead scheduling plan is greatly weakened. Therefore, advanced digital technology and artificial intelligence methods are urgently needed as support, such as grid measurement and transmission, equipment condition awareness and online monitoring, as well as real-time prediction, trend analysis and intelligent decision-making based on big data and artificial intelligence methods.
1.2.3 Virtual power plant support technology
Firstly, due to the complex model parameters, the virtual power generation unit based on DERs aggregation needs to be simplified based on various conditional assumptions during use, resulting in low model accuracy. Based on data-driven, there is a high threshold that requires the accumulation of a large amount of historical data. Therefore, it is necessary to combine physical models and data mining theories to jointly model virtual power generation units to improve the accuracy and usability of the model.
Secondly, DERs have many types, large numbers, small monomer capacity and multiple uncertainties, which require effective aggregation and parameter identification. Thirdly, based on the analysis and mining of massive data on the demand side, accurate portraits of power users can assist electricity sales companies in decision-making and provide differentiated value-added services including load forecasting, demand-side response implementation, and electricity sales package formulation. The above supporting technologies cover the whole process of data collection, storage, calculation, analysis and decision-making, and need to ensure the accuracy of models and real-time decision-making through real-time data collection and dynamic update, so a data business platform with strong data management capabilities and super computing capabilities is required as support.
1.2.4 Business model of virtual power plants
In terms of business model, it is necessary to integrate the interests of all parties within the virtual power plant and design an appropriate revenue distribution scheme. Based on the benefit evaluation data of virtual power plants, the reasonable commercial development financing mode of virtual power plants is studied. At the same time, considering the possibility of the development of the mainland electricity market and the demand for various energy value-added services, a variety of commercial operation modes of "source-grid-load-storage" virtual power plants are finally formed to maximize the comprehensive benefits of virtual power plant operation.
In addition, studying the blockchain-based distributed tradable energy nearby consumption mechanism can fully mobilize the enthusiasm of DERs to participate in market-oriented transactions. Since the structure of virtual power plant participation in the market includes multiple DERs at the equipment level, aggregators at the user layer and electricity market information at the application layer, the data structure is complex and diverse, and it is in different software ecosystems, so it is urgent to establish a data management platform to facilitate virtual power plant operators to mine data at a deeper level, enhance the cooperation and integration of multi-dimensional models, and make differentiated intelligent decisions in response to different scenarios.
2 Digital twin
2.1 Overview
2.1.1 The concept of digital twins
As early as around 2003, the concept of digital twin appeared in the product life cycle management course taught by the University of Michigan Grieves, but the term "digital twin" did not officially appear in NASA's technical report until 2010, and was widely used in the aerospace field in the following years, including airframe design and maintenance, aircraft capability evaluation, aircraft failure prediction, etc. [33-35]. In recent years, thanks to the development of a new generation of information technology such as the "big cloud migration intelligent chain", digital twins have been widely used in the fields of electric power [36], ships[37], smart cities[38], and medical and health care[39].
Currently, the concept of digital twins is constantly developing and evolving. To summarize the research of all parties, digital twin is based on physical model, makes full use of sensor technology to perceive and store multi-source data such as physical entities, simulation, knowledge, applications, etc., integrates multi-disciplinary, multi-physical quantity, multi-spatiotemporal scale simulation analysis, and then forms a digital twin in the virtual space, so as to reflect the transformation and connection process of physical entities and twin data corresponding to the whole life cycle. From a functional point of view, digital twin applications require the support of physical infrastructure, where products, services, and process data are synchronized into the virtual space, and the models and data in the virtual space are fed back and interacted with the application process. By entering physical world information and related stimuli, real-time analysis results including prediction, simulation, optimization, and health monitoring are output.
2.1.2 Features of digital twins
Conceptually, digital twins have the following typical characteristics [40-41]:
1) Interoperability: Physical entities and digital space two-way mapping, dynamic interaction and real-time connection, so the physical-digital twin system can measure and obtain real-time data to update the digital model, and at the same time transmit the corrected and calculated control parameters in the digital model back to the physical controller through the control interface to achieve precise control of the physical equipment.
2) Scalability: Digital twin technology can easily integrate, add and replace digital models, which is convenient for scale scaling for multi-timescale, multi-physical, and multi-level models.
3) Real-time: Through the digital representation of physical entities that change with time axis, a digital spatial mapping of the real-time state of physical entities is formed, in which the objects to be represented include the appearance, state, attributes and internal mechanism of physical entities.
4) Fidelity: Digital twins require digital twins and physical entities to maintain geometric similarities, that is, digital twins simultaneously simulate physical entities in state, phase and tempe.
5) Closed-loop: The digital twin describes the internal mechanism of the physical entity through visual means, monitors the state of the physical entity, optimizes the state and operating parameters of the physical entity through intelligent algorithms, and realizes the closed loop through the feedback decision-making function.
2.1.3 Digital twin technology
Digital twin is a system engineering built with models and data as the main elements, suitable for the use of big data, artificial intelligence and other complex tasks processing, is an important means to promote the digital transformation of enterprises and promote the development of the digital economy. According to the different levels of the digital twin system, the key technologies involved include digital model construction, data interaction, simulation analysis, and decision support [33,42]. For example, the multi-source, heterogeneous all-element massive data generated by the process of physical perception, model generation, simulation analysis and other processes in twin space can fully mine effective information and effectively support system decision-making by using big data analysis methods.
In addition, digital twins are elastic at scale, and computing and storage needs are rapidly increasing from the unit level to the complex system level, and cloud computing can take advantage of its on-demand and distributed sharing model to dynamically meet digital twin computing and storage needs. Third, based on artificial intelligence algorithms, in-depth knowledge mining of twin data is carried out without the participation of data experts, providing customized and accurate services, and improving the added value of data. In addition, blockchain technology ensures that twin data is immutable, traceable, traceable, etc.
2.2 Research status of digital twin in energy and power system
In recent years, some studies have begun to gradually explore the application of digital twin technology to the power system. For example, in terms of framework and review research, literature [43] summarizes the development experience of digital twin technology for smart energy systems at home and abroad, analyzes its deployment and application in the smart energy industry, and gives development suggestions from three aspects: technology, ecology and policy. The literature [44] puts forward the concept, construction framework and key technologies of digital twin power grid, and puts forward typical application scenarios according to the characteristics of the power industry. Literature [45] proposes a digital twin framework for energy Internet, and gives application routes for scenarios such as equipment status evaluation, "source-grid-load-storage" autonomous regulation, online analysis of power grid, comprehensive energy autonomous coordination, and user data mining. Literature [46] analyzes the value of integrated smart energy digital twin system in smart city development, and explains its key technologies and its integration and application with the planning, operation, control and optimization of integrated energy system. Literature [47] Design the framework of digital twin power system, explore the key problems and core technologies faced by its construction, and discuss its application and development prospects in multiple fields of power system; The literature [48] introduces the key technologies required for the realization of digital twins of power equipment, and analyzes its challenges in data collection, model construction and solving, and platform use.
In terms of the specific application of digital twin technology, literature [49] proposes an ultra-short-term prediction mechanism of photovoltaic power generation power based on digital twin. The literature [50] proposes the construction method of energy Internet digital twin, and introduces the energy Internet planning platform CloudIEPS based on digital twin. Literature [8] proposes a multi-agent control architecture for smart microgrid driven by digital twins, constructs twin agent model components and establishes information interaction and transmission mechanism between components. Literature [51] Aiming at the fluctuation of wind power output, a control strategy based on digital twin hybrid energy storage is proposed to optimize the operation efficiency of energy storage equipment. Literature [52] proposes the principle of longitudinal protection of flexible HVDC transmission system based on digital twin, and establishes an accurate digital twin model of DC line. Literature [53] proposes an online analysis architecture of power system based on digital twin, and verifies the effectiveness of the proposed method through simulation analysis.
3 Digital twin virtual power plant
3.1 Digital twin virtual power plant definition
Based on the above research status of virtual power plant and digital twin technology in energy and electricity, this paper further proposes the concept of virtual power plant based on digital twin. Digital twin virtual power plant is a development form of virtual and real integration and coexistence of physical space, physical and information dimensions. By accepting various types of information from physical DERs entities, environments, markets, etc., the digital space creates a virtual space corresponding to the physical virtual power plant, and evolves synchronously with the physical virtual power plant, reflecting the state of the virtual power plant in the real environment in the form of holographic simulation, dynamic monitoring, real-time diagnosis, accurate prediction, etc., and then promoting the digitalization of all elements of the virtual power plant, real-time visualization of the whole state, multi-level operation management collaboration and intelligence, and realizing comprehensive and accurate monitoring of the physical virtual power plant. Interact synergistically with the virtual space and feed back the results of analysis, such as diagnostics, forecasting, participation in the electricity market, and dispatch control, to the virtual plant physical entity.
By combining the virtual power plant decision-making system with the digital twin, an intelligent decision support system that can continuously learn and evolve is formed, and the allocation efficiency of material, intellectual and information resources on the demand side is improved, so as to promote the overall optimization of the virtual power plant. As a closed-loop enabling system connecting massive multiple DERs and transmission and distribution network data resources, the digital twin virtual power plant will open up a new mode of digital smart grid construction and operation management through global resource identification, accurate status perception, real-time data analysis, scientific model decision-making, and intelligent and accurate execution.
3.2 Digital twin virtual power plant system construction
The construction of digital twin virtual power plant should synchronize the planning of the virtual power plant physical entity and the digital twin virtual space, and build a data middle platform from the modeling stage to form a static property database. At the same time, during the operation process, simulation, knowledge, application and other related models and management data are continuously imported into the virtual space, and the data middle platform database is constantly improved; In the operation stage, relying on the intelligent analysis platform to achieve decision-making support and optimal management of virtual power plants. For the DERs that have been built and put into use, they are incorporated into the digital twin virtual power plant system through digital modeling and deployment of IoT facilities, and the information hub data center is supplemented and improved through intelligent perception and data collection.
In terms of optimized operation, the virtual twin space and physical entities enable parallel interaction between the virtual and the real through efficient connection and real-time transmission. Through the intelligent perception of the Internet of Things and real-time information collection technology, "from the real to the virtual" is realized; The virtual power plant physical entity and virtual space realize virtual and real iteration through the feedback mechanism, and realize "virtual control from reality" through the support of intelligent decision-making platform and real-time optimized operation control.
In view of the above requirements for the construction of digital twin virtual power plants, this paper refers to the five-dimensional model of digital twins in the industrial field [54] and proposes a digital twin virtual power plant system (DTVPPS) that includes physical entities, digital twins, twin data, connections and intelligent applications, as shown in Equation (1).
Formula: PVPP is a virtual power plant physical entity; VVPP is a digital twin; DTDate is twin data; CN is the connection between the parts of the system; DT&S stands for the services provided by the twin system; The overall architecture of DTVPPS is shown in Figure 1, PVPP digitally describes the characteristics, behaviors, formation processes and performance of physical objects such as DERs and markets through digital technologies such as smart sensors and the Internet of Things, and builds historical and real-time operation databases. VVPP uses data integration and simulation operations to form a complete digital twin mapping, and realizes the evolution and prediction of the future state of the virtual power plant through real-time simulation, which is displayed to users in a visual way, providing DT&S such as safety warning and fault diagnosis while supporting the safe and stable operation of the virtual power plant. Finally, through efficient connection and real-time transmission, all parts can be realized to interact collaboratively, feedback control of physical entities and iterative growth of simulation models in VVPP.
3.2.1 Physical Entities (PVPP)
PVPP construction can be done from the bottom up according to the hierarchy, including:
1) Equipment level: including distributed photovoltaic, distributed wind power, gas turbine and other power generation resources on the distribution network side, flexible loads such as household/public electric vehicle charging piles, air conditioners, water heaters and other electric/gas/cold/hot loads, as well as distributed energy storage, heat storage tanks, ice cold storage, PtoX and other energy storage equipment;
2) User level: including schools, industrial parks, commercial buildings, energy stations, microgrids, data centers, residential areas, energy storage power stations, and intelligent power distribution systems;
3) Virtual power plant level: including source-side power sources such as photovoltaic power plants and gas-fired power stations, as well as virtual power plants dynamically aggregated according to the characteristics of DERs.
A single DERs digital twin is the smallest virtual unit of DTVPPS, and real-time monitoring and fault prediction of DERs can be realized through digital twin construction. Multiple DERs aggregate to form a user-level virtual power generation unit to achieve optimal operation between its internal resources. Multiple user-level virtual power generation units can dynamically aggregate according to the complementary characteristics of multi-energy and geographical location, and then build a virtual power plant-level digital twin system to optimize the dispatch, market participation, and energy efficiency improvement of massive and diversified DERs in the region. Multiple DTVPPS build a larger range of DTVPPS through collaboration and game, predict the evolution of distribution stations, increase the proportion of renewable energy access, optimize the capacity allocation of DERs, and improve system flexibility.
3.2.2 Digital twins (VVPP)
VVPP is a mirror image of the virtual power plant entity in the virtual space, including the physical model (MM), the data-driven model (MD), and the co-driven model (MM&D) of physics and data, as shown in Equation (2).
The multivariate model is used to describe and characterize PVPP from the multidimensional space-time scale, and the physical mechanism modeling is selected when the physical process is fully observable and there is a fixed mathematical model. When the physical entity parameters cannot be directly measured, the physical mechanism model is difficult to solve, or the physical mechanism of DERs is not clear, the data-driven model is obtained by integrating and refining the collected data. When the two models alone cannot meet the requirements of accuracy and accuracy, the joint modeling method is adopted, and the data-driven model is used as the error compensator to compensate and correct the mechanism model to improve the accuracy of the overall model.
VVPP can get rid of the limitations of space, time and environment in the physical world, predict potential faults through the deduction and prediction of the operation status of virtual power plants, or give reasonable solutions to existing faults, so as to achieve the prevention or immediate solution of virtual power plant failures.
3.2.3 Twin data (DTDate)
The mapping of PVPP to VVPP is based on perception technology and acquisition terminal to realize the intelligent perception and acquisition process of multiple massive data such as electrical quantity, state quantity, physical quantity, environmental quantity, spatial quantity, and behavior quantity in each link of "source-network-load-storage". Therefore, DTDate is the core of DTVPPS, including data collection, storage and management system, including physical entity data (DTpvpp), digital twin data (DTVVPP), intelligent entity application data (DTS), knowledge data (DTK) and data generated by multi-data fusion (DTF), such as Equation (3):
DTpvpp includes the physical element attribute data of the physical entity specifications, functions, performance, relationships, etc. of the virtual power plant, as well as the dynamic process data reflecting the operating status, real-time performance, environmental parameters, and sudden disturbances of the physical entity of each link of the virtual power plant. DTVVPP includes data generated by simulation including steady-state power flow, voltage, and real-time operation constraints, rules, analysis and evaluation of virtual power plants. DTS includes unstructured data such as service objects, statistical analysis, rules, and results; DTK includes expert knowledge, industry standards, rule constraints, reasoning inference, common algorithm libraries and model libraries, etc. DTF is the derivative data obtained after conversion, classification, association, integration and fusion of DTpvpp, DTvvpp, DTS, DTK, DTF, etc., and obtains cyber-physical fusion data by fusing virtual power plant physical reality data with historical statistical data, expert knowledge and other information data, so as to realize information sharing and value-added.
3.2.4 Connection (CN)
CN is a network connection that realizes efficient connection, real-time transmission, collaborative interaction and iterative optimization of all aspects of DTVPPS, including the connection between PVPP, VVPP, DT&S, and the interconnection between them and DTDate. In terms of interaction with DTDate, PVPP data is collected by various sensors, embedded systems, data acquisition cards, etc., and transmitted to DTDate through the fieldbus; VVPP stores simulation and related data to DTDate in real time, obtains relevant data in real time for dynamic simulation, and feeds back the data or instructions processed through simulation to physical entities to realize the optimal scheduling of virtual power plants. Similarly, while reading history, rules, algorithms, and model data for business support and operation management, App Service also stores the generated data in real time to DTDate.
DT&S and VVPP realize two-way communication through software interface, complete direct instruction transmission and message synchronization, etc. VVPP can use the real-time operation data of PVPP to update the correction model, and PVPP can use the VVPP analysis results to achieve feedback control of the virtual power plant. In addition, in view of the transmission requirements of massive and diversified data in virtual power plants, plug-and-play and wide-area flexible access of various sensors can be combined with "cloud-edge" collaboration technology, forming a highly reliable data transmission network covering the entire DTVPPS.
3.2.5 Application Services (DT&S)
DT&S includes "functional services (SF)" that encapsulate various data, models, algorithms, and simulation results required in the application process of DTVPPS to support the operation and implementation of internal functions, and "business services (SB)" that meet the needs of multiple users in various fields of virtual power plants in the form of application software, as shown in Equation (4).
DT&S's SF includes model management services for VVPP, data storage, cleaning, mining, fusion and other services for DTDate, and data collection, sensory access, transmission, protocol, interface and other connection services for CN. SB includes instructive services for terminal operators; Professional technical services for technicians, such as multi-time scale prediction of DERs, participation in electricity market strategies, dynamic optimization of scheduling, etc.; Intelligent decision-making services such as demand analysis and risk assessment for managers, functional experience for end users, and product services such as remote maintenance.
3.3 Key technologies of virtual power plant digital twin
Digital twin virtual power plant is a complex technology and application system for new digital smart distribution network with multi-source data integration, multi-category technology integration and multi-type platform functions, which will effectively support the digitalization of the power grid by combining with the existing virtual power plant management and control platform. First, the typical structure of a virtual power plant consists of an equipment layer, a user layer, a virtual power plant layer, and an application layer. Among them, the equipment layer includes multiple DERs such as distributed power supply, energy storage equipment, electric vehicles, flexible load, and PtoX equipment; The user layer includes energy storage service providers, data centers, commercial buildings, public buildings, residents, and small and medium-sized users with multiple energy storage devices. DERs are actively aggregated between the virtual power plant layer and flexible resources through centralized control, so that virtual machines can optimally dispatch demand-side resources through price or other signals. The application layer relies on the virtual power plant management platform to review the access conditions, access schemes, external characteristics, and response information of the virtual power plant.
The power trading platform conducts market transactions according to the access review results, and sends the clearing results to the dispatching platform for safety verification; The dispatching system issues dispatching requirements to the virtual power plant management platform, which further decomposes the dispatching instructions to the user layer and the equipment layer to achieve closed-loop control. Correspondingly, the DTVPPS technology ecosystem consists of a physical layer, a perception layer, an information hub layer, and an intelligent application layer, as shown in Figure 2.
3.3.1 Physical Layer
The physical layer is the basis of the digital twin virtual power plant, which mainly includes the "source-grid-load-storage" related links involved in the virtual power plant, physical entities, power markets, distribution network operations and personnel behavior. The physical layer is the carrier of twin data, providing the perception layer with various physical entity parameters, market operation indicators, power grid parameters and personnel behavior related data. In addition, the physical layer also includes communication lines, network systems, host server systems, storage devices, etc. and other basic hardware facilities, which is the premise for ensuring the operation of the entire information system.
3.3.2 Perception Layer
The perception layer is the medium of DTVPPS data perception access, which collects the electrical quantity of the physical entity of the virtual power plant through high-performance sensors, and the environmental quantity, distribution network status quantity, physical quantity, etc. related to the access range of the virtual power plant, so as to realize the status perception, data transmission, environmental monitoring and behavior tracking of the physical objects aggregated by the virtual power plant. The perception layer stores and incorporates collected data, or knowledge obtained from the environment, into the digital twin system, enabling the connection between the physical entity of the virtual power plant and the virtual space. Based on artificial intelligence algorithms such as deep learning, the mirror system of virtual power plants in the digital twin space is constructed, and the system computing efficiency is improved by combining offline learning and online decision-making based on cloud computing and edge computing, and supports the intelligent optimization operation of "source-grid-load-storage" of virtual power plants.
3.3.3 Information Hub Layer
The information hub is the intelligent brain of DTVPPS, which consists of a data middle platform, a twin model, and an intelligent analysis platform. Among them, the data middle platform SCADA obtains and stores online data of power grid, application and knowledge data such as weather/season/society, power grid fault data and equipment status data, etc., to form a historical data, real-time data and model simulation database including virtual power plant internal DERs, market, environment, etc.
The data middle platform can realize data management and sharing, which is the core driving force of the digital twin virtual power plant, and supports the functions of batch data access, statistics, browsing, query, deletion and management. The twin model construction uses simulation data such as prediction, status, behavior, and early warning, and builds twins based on three methods: physical model, data-driven model, and physical and data fusion model. The intelligent analysis platform includes functions such as data analysis, simulation calculation, scenario simulation, intelligent computing, and artificial intelligence algorithms. In addition, it will also involve related technologies such as software-defined terminal management and access technology, high-performance transmission and storage technology, and data sharing services.
3.3.4 Decision application layer
The decision-making application layer is divided into device management, distribution network access, application services, and operation management modules according to scenarios. Among them, the equipment management module includes DERs equipment autonomous state management, real-time fault diagnosis, maintenance plan formulation, etc.; Distribution network access management includes security warning, online analysis, DERs access optimization, electric vehicle charging pile planning, energy storage configuration, etc. The business module includes the prediction of distributed power generation, behavior, fault, etc., as well as user data mining and value-added interaction. Operation management includes end-user asset classification aggregation and virtual power plant external feature management, virtual power plant participation in multiple power market transactions, settlement, and tracking, multi-dimensional spatiotemporal scale optimization and scheduling, and full value chain collaborative optimization and visualization.
4 Digital twin virtual power plant practice prospect
4.1 Typical applications
4.1.1 Virtual power generation unit modeling and simulation
Modeling and simulation is based on digital form to accurately reproduce the main factors such as virtual power plant aggregation resources, distribution network and market, and is used to carry out virtual power plant market bidding, DERs monitoring analysis and operation optimization. Virtual power plant aggregates multiple DERs with complex internal structure, makes full use of DTVPPS information hub data middle platform, combines the advantages of physical mechanism and data-driven modeling, realizes the complete mapping of physical resources of virtual power plant to virtual space through organic integration, builds digital twins, and forms accurate digital simulation models.
At present, the modeling and simulation of virtual power plants is still mainly based on models based on mathematical formulas and physical mechanisms, and some studies based on data-driven modeling, such as literature [55] for the problem of different characteristics of participating demand response users, large differences in response ability, and high-dimensional, nonlinear and non-convex overall response characteristics after multi-type resource combination, a data-driven modeling method is proposed to achieve effective characterization of the complex response characteristics of aggregates. Literature [56] Based on data analysis, this paper studies the data-driven modeling methods of power generation, load and energy storage equipment using deep learning and cluster analysis techniques. In the future, algorithms such as category balancing algorithm, policy network and value network data learning can also be used to overcome the problem of inconsistency and lack of original data. Based on cost-sensitive learning, machine learning inversion and parameter recognition, the shortcomings of difficult modeling and low accuracy of mechanism model modeling are overcome. And through the ensemble learning algorithm, improve the versatility of the system operation state evaluation method.
4.1.2 Digital twin virtual power plant planning
With the gradual liberalization of the power market, many aspects such as power system planning, operation and management will change, and the traditional yearly on-time planning can no longer adapt to the current rapidly changing power system needs. The introduction of digital twin concept into virtual power plant planning can first reflect the changes of market price signals in a timely manner, so that the multi-DERs planning on the demand side can adapt to market demand. Secondly, through the multi-time scale simulation and rehearsal of the digital twin system, trial and error in the digital space are tried and made at low cost, avoiding the excessive construction of hardware facilities such as power supply, power grid, substation, and energy storage, replacing on-time planning with on-demand planning, and accurately quantifying the investment scale of virtual power plants. Third, in extreme cases, the digital twin virtual power plant can carry out abnormal identification and safety warning based on data analysis, simulation calculation, scenario simulation and other methods, and timely feedback the results to the physical power grid to guide the construction of the virtual power plant and avoid the lagging impact of timely planning, as shown in Figure 3.
It should be emphasized that operators predict and deduce the development of the virtual power plant physical entity according to simulation technology in the digital space, and then feedback the results to the physical entity, thereby affecting the development trajectory of the virtual power plant physical entity and forming a closed-loop feedback. Virtual power plant planning based on digital twin can combine big data analysis and artificial intelligence algorithms to deduce its optimal evolution direction, give customized countermeasures, and can affect the development trajectory of physical entities at any time scale through rehearsal and feedback mechanisms [13].
4.1.3 Digital twin virtual power plant optimization scheduling
At present, the modeling method based on physical model of virtual power plant is difficult to effectively cope with the problem of "source-grid-load-storage" collaborative optimization operation of distribution network. At the same time, the traditional virtual power plant simulation analysis and decision-making model based on simplification and assumption conditions is difficult to accurately describe the dynamic behavior process of virtual power plant, and it is difficult to achieve real-time interaction with the environment and physical entities. The digital twin virtual power plant can sense the status of multiple DERs and related data such as distribution network and market in real time, and embed artificial intelligence algorithms such as deep reinforcement learning in the intelligent analysis platform to support intelligent evaluation and fault warning of virtual power plant operating status, as shown in Figure 4.
At the same time, DTVPPS provides real-time connection between virtual space and physical entities, supports real-time interaction between physical and data-driven fusion models and environments, realizes dynamic iterative optimization of models, and improves control accuracy. In addition, by reflecting equipment load, user load, new energy output, distribution network operation status, etc. in real time, virtual power plant status monitoring is realized, and the optimal scheduling strategy is adjusted and optimized in a timely and effective manner. Third, through self-learning, the intelligent application platform formulates the optimal aggregation strategy based on the complementary benefits of multi-energy, so as to enhance the pertinence of virtual power plants, improve their participation in the power market, improve the flexibility of the power system, and promote the absorption of renewable energy.
4.1.4 Intelligent health management of digital twin virtual power plants
At present, the management of demand-side multiple DERs faces problems such as difficult condition monitoring, complex fault diagnosis and maintenance. Through the intelligent perception and collection of DERs equipment status and information, the digital twin virtual power plant constructs the twin image of the virtual power plant physical equipment in the virtual space, and then carries out the health status management and evaluation of virtual power plant equipment through multi-data access specification and fusion, status assessment and fault identification diagnosis. The construction of the virtual power plant digital twin provides an accurate description of the evolution of the whole life cycle of the virtual power plant, the operation and maintenance behavior of personnel and the interaction process with the environment, etc., which is helpful to realize real-time status monitoring, fault prediction/diagnosis, on-site and remote interaction, maintenance guidance, etc. of the whole life cycle of the equipment inside the virtual power plant, realize DERs, distribution station voltage, power quality abnormal tracing analysis, research and judgment of abnormal causes, intelligent recommendation of improvement measures, and automatic fault analysis report generation. Effectively reduce the operation and maintenance cost of virtual power plants and improve the service life of equipment, as shown in Figure 5.
4.1.5 Value creation and comprehensive evaluation
For a long time, the idea of using virtual models for simulation optimization and providing services based on the optimization results was not new. However, the widespread use of virtual power plant digital simulation platforms with more complex modeling and simulation capabilities, better interoperability, and more ubiquitous IoT sensors has made it possible to create more elaborate digital simulation models and provide value-added services based on them. In recent years, more and more enterprises have applied digital twin technology to transform from product sales to "product + service". However, at the same time, it also faces problems such as uneven levels of data collection capabilities in enterprises/industries, inability to obtain underlying key data, and high idle degree of existing data, lack of integrated applications of data correlation and in-depth mining.
In the future, based on DTVPPS, virtual power plants will be able to effectively use advanced technologies such as artificial intelligence, big data, cloud computing, incremental clustering, and blockchain consensus to integrate massive data in the virtual power plant ecosystem, fully release the potential of demand-side DERs regulation, provide diversified, refined and customized services to users, aggregators, distribution networks and other parties, and enhance the added value of virtual power plants. In the future, DTVPPS will effectively manage the environment, model, analysis results and other data generated in the process of digital twin model construction, simulation analysis, reliability evaluation and intelligent application, and use big data analysis methods to dig deep and make full use of the effective information in the database, which will establish an effective reliability evaluation index system for virtual power plants.
4.2 Digital twin virtual power plant advantages
1) Coordinate the internal DERs of virtual power plants to achieve optimal allocation of energy resources. Based on digital twin technology, virtual power plants can dynamically aggregate multiple adjustable resources through the Internet of Things and blockchain technology to promote the optimal allocation of DERs. Based on the data middle platform and intelligent analysis platform, through big data and advanced artificial intelligence technology to assist users to tap the potential of energy conservation and emission reduction, promote the transformation of energy consumption from a single and passive mode to an efficient utilization mode that integrates multiple needs, active participation and customization, and promote the popularization of electric vehicles, electric energy substitution, energy conservation and emission reduction, and the development of comprehensive energy services. In addition, with the continuous advancement of the power market, DTVPPS incorporates real-time market information into the decision-making system, which can efficiently coordinate the interests of different market entities and further realize the optimal allocation of all-factor resources of virtual power plants.
2) Realize the digitalization of all elements of virtual power plants and promote the construction of digital power grids. Based on digital twin technology, the digital twin virtual power plant effectively aggregates multiple DERs to support the real-time dynamic response of the power system, takes data as the core production factor, digitizes the physical world DERs, and related people, things, events and other elements, builds a virtual power plant digital twin, and opens up the information of each link of "source-network-load-storage" through real-time and efficient connection, so as to realize the comprehensive, accurate, measurable and highly controllable virtual power plant. The dynamic feedback of the physical world to the digital world accurately and in real time through sensors, and the realization of virtual-real interaction through networking and intelligence, will promote the construction of a new digital power system and form a new form of distribution network development that combines virtual and real and twin interaction.
3) Improve the operation level of virtual power plants and realize multi-dimensional spatiotemporal scale optimization management. The optimal management of demand-side DERs is an effective means to provide the flexibility and adjustment ability of the power system. Building a digital twin virtual power plant with the help of digital twin technology will greatly change the extensive management mode of traditional DERs through data empowerment, which is the key to achieving the optimal configuration of demand-side DERs, and is also the core of DERs to achieve scheduling, operable and tradable. In the target dimension, coordinate electrical, physical, environmental, market and other multi-source data to assist virtual power plants to participate in electricity energy and ancillary service market transactions; In the time dimension, it supports the construction of an optimal scheduling mechanism for multi-time scale coordination; In terms of spatial dimension, on the one hand, the virtual power plant can optimize the quotation strategy to participate in the power market with the help of the powerful analysis and computing capabilities of the data middle platform in the information hub and the intelligent analysis platform, and on the other hand, the virtual power plant can achieve precise control of multiple DERs in the autonomous area.
In summary, through the construction of DTVPPS, intelligent and accurate insights into the core business of virtual power plants are realized, and accurate, convenient and intelligent services are realized by providing new services and new models, improving flexible and interactive personalized service capabilities, improving customer acquisition and satisfaction, and serving virtual power plant collaborative planning and precise investment to improve power grid security risk management and control capabilities. The advantages of DTVPPS over traditional virtual power plants are summarized in Table 1.
Table 1 Advantages of DTVPPS compared to traditional virtual power plants
4.3 Development trend of digital twin virtual power plants
DTVPPS provides virtual power plant operators with simulation simulation, diagnosis and prediction, visual monitoring, and optimal scheduling with its accurate and reliable twin models, multi-source, massive twin data, and real-time virtual-real interaction. With the rapid development and deep integration of technologies such as the "Big Cloud Mobility Intelligent Chain", the future digital twin virtual power plant will develop in the following directions:
1) Refinement: Through physical mechanism and data-driven fusion modeling, a high-precision DERs digital twin model is established, which reduces the aggregation cost while quantifying its adjustment flexibility to participate in the multi-power market and realize refined management throughout the whole life cycle.
2) Systematization: In the future, it will break the construction and management mode of massive multiple DERs resources on the demand side, integrate the digital twin of the equipment layer, user layer and virtual power plant layer, and provide support for accurately grasping the real-time operation status and development evolution trend of virtual power plant and distribution station area.
3) Intelligence: With the improvement of the intelligent level of digital twin virtual power plants, it will effectively support multi-time scale distributed power generation and load forecasting, research and judgment and guidance of distribution station area development planning, and intelligently respond to system uncertainty and emergencies through iterative learning. In addition, with the addition of emerging technologies such as blockchain, such as the application of DERs distributed transaction process, it will effectively protect the security and privacy of user data [57-58].
4) Generalization: With the maturity of digital twin related technologies and the rapid development of DERs, the cost of building digital twin virtual power plants will be significantly reduced, and in the future, it will be widely used in the optimization and control of DERs and the planning, construction and operation of distribution station areas.
5) Openness: The integration of digital twins and virtual power plants will increase the inclusiveness of digital twin virtual power plants, and will jointly provide support for the digital transformation of power systems through cross-integration with concepts such as digital twin transportation, digital twin architecture, and digital twin cities. In addition, connecting multiple DTVPPS through a distributed framework will effectively support the modular and customized services of virtual power plants.
4.4 Digital twin virtual power plant development platform outlook
At present, domestic and foreign manufacturers such as Siemens, ANSYS, GE, Microsoft, BitVision and Tongyuan Soft Control have provided digital twin related solutions [40]. However, depending on the industry or the target target, the platform developed and used by each manufacturer is also different. For example, ANSYS developed Twin Builder for digital twin simulation modeling, and based on MATLAB, Simulink's multidomain modeling tools can be used to create physical models; Model visualization platform, including GIS geographic information system, unity3D visualization, WebGL/Canvas web page visualization, etc.; BigChainDB database technology based on blockchain digital twin; Microsoft's Azure IoT digital twin platform, which can create a comprehensive digital model, and HanClouds' industrial Internet platform, launched by Hanyun Technology.
However, the application of digital twins in the energy and power industry is still in its infancy, and only the literature [50] mentions the cloud platform CloudIEPS for integrated energy system planning. In the future, based on the existing simulation analysis software in the energy and power industry, combined with the key technologies of digital twin virtual power plants introduced in the third part of this paper, it is necessary to further cooperate to build an open digital twin virtual power plant development platform, explore application solutions that can be implemented, and gradually accumulate and form some demonstration products.
5 Conclusion
This paper summarizes the development direction of virtual power plants under the goal of "building a new power system", puts forward the overall framework of digital twin virtual power plants based on the characteristics of digital twin technology, analyzes its key technologies, and looks forward to its application practice and development trend. The construction of the digital twin virtual power plant will promote the construction of a new power system with new energy as the main body, promote the digital development of the power system, and is the key to the implementation of the "dual carbon" strategy of the power system. Secondly, based on comprehensive, accurate and transparent data collection and analysis, the digital twin virtual power plant uses advanced computer technology to provide data analysis and regulation potential mining, which will effectively enhance the decision-making level of the virtual power plant. By building a digital twin virtual power plant, creating a flexible and efficient demand-side DERs management and configuration platform, and improving the holographic perception, flexible control, optimal scheduling, health management and intelligent service capabilities of the virtual power plant, it will support the development of renewable energy, enhance the flexible adjustment ability of the distribution station area, and promote "clean substitution" and "electric energy substitution". Finally, there are two points to note about the future development of digital twin virtual power plants:
1) Challenges for the future development of digital twin virtual power plants: First of all, the core of digital twin technology is data, and the accuracy, consistency and transmission stability of twin data are key to realizing the full potential of digital twin technology. At the same time, digital twins require the original closed system of relevant enterprises to gradually transform into an open system, so when applying virtual power plant data to value-added service improvement, it will also face data sharing and data security challenges from different platforms, users, and power grid enterprises.
Secondly, DTVPPS integrates physical entities and virtual space, which is a multi-disciplinary, multi-physical quantity, multi-time scale, and multi-uncertainty simulation process, which requires the integration of multi-dimensional systems: in terms of platform construction, it is necessary to build a data middle platform containing electrical, environmental, physical, market, user behavior and other multi-source data from the systematic level of virtual power plants; In terms of openness, it is necessary to open up the platform software ecology such as prediction, simulation, and market trading involved in the operation of virtual power plants; At the model level, it is necessary to enhance the customized aggregation and parameter identification of DERs. Thirdly, in the future, it is necessary to further explore the basic theory of computer algorithms and integrated fusion technologies that are more in line with the application of digital twin virtual power plants, and build an intelligent application platform that can be iteratively updated and independently evolved.
Finally, in terms of business model, the business model of digital twin virtual power plant is not perfect, limited by the process of the power market, it is urgent to break through the integration of multiple technologies and form a market mechanism for virtual power plants to participate in multiple markets as soon as possible.
2) Driving force for future development: The implementation of policies such as digital grids, the development of computing equipment/hardware, the growth of data scale, the evolution of models and algorithms, and the participation of professionals will pave the way for the implementation of digital twin virtual power plants.
Source: Proceedings of the CSEE, authors: Yan Xingyu, Gao Ciwei, etc