Yi Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yi Sun has a total of 53 co-authors. Here are the top co-authors by number of shared publications:
- Shunlin Zheng
School of Electrical & Electronic Engineering North China Electric Power University Beijing People's Republic of China
6 shared publications - Bin Li
School of Electrical and Electronic Engineering, North China Electric Power University, Beijing 102206, China
3 shared publications Bing Qi
Department of State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, Electrical and Electronic Engineering, North China Electric Power University, Beijing, China
2 shared publications- Haobo Guo
Department of State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, Electrical and Electronic Engineering, North China Electric Power University, Beijing, China
2 shared publications Hongyue Yang
School of Electrical & Electronic Engineering North China Electric Power University Beijing People's Republic of China
2 shared publicationsHuiyu Bao
School of Electrical and Electronic Engineering North China Electric Power University Beijing China
2 shared publications
Latest Publications
- Open Access
- 1 January 2026
- Published by Institution of Engineering and Technology (IET) in IET Generation, Transmission & Distribution
- Vol. 20 (1) , e70299
An Adaptive Selection Strategy for an Incentive‐Based Integrated Demand Response Considering a Dynamic Participation Rate and Stochastic Response Deviation
Incentive‐based integrated demand response (IBIDR) has been recognised as a powerful tool to mitigate supply–demand imbalance in energy systems. However, the temporal dynamics of user participation rates and the stochastic fluctuation of users' response power have been central challenges for multi‐energy service providers (MESPs) in designing effective incentive strategies. To address these challenges, this paper proposes an adaptive selection strategy for IBIDR considering the dynamic participation rate and stochastic response deviation. First, users' dynamic participation rate, simultaneously influenced by both imitative behaviour and spontaneous behaviour, is modelled by a Markov process, which improves the user model. Second, an adaptive selection mechanism is proposed based on users' historical response power deviation levels and participation rates, which improves the MESP model. Moreover, the improved user model and the enhanced MESP model are jointly expressed as a bi‐level stochastic programming problem. Utilising the Karush–Kuhn–Tucker (KKT) conditions and probability theory, the original bi‐level stochastic problem is reformulated into a tractable single‐level deterministic optimisation model. Finally, simulation results demonstrate that the proposed model effectively improves the accuracy of user response behaviour prediction and the efficiency of MESP's incentive strategies, thereby reducing MESP's total cost, mitigating response power deviations, and enhancing the utilities for users participating in IBIDR.
- 1 January 2026
- Published by Institution of Engineering and Technology (IET) in IET Generation, Transmission & Distribution
- Vol. 20 (1) , e70299
Open AccessAn Adaptive Selection Strategy for an Incentive‐Based Integrated Demand Response Considering a Dynamic Participation Rate and Stochastic Response Deviation
Incentive‐based integrated demand response (IBIDR) has been recognised as a powerful tool to mitigate supply–demand imbalance in energy systems. However, the temporal dynamics of user participation rates and the stochastic fluctuation of users' response power have been central challenges for multi‐energy service providers (MESPs) in designing effective incentive strategies. To address these challenges, this paper proposes an adaptive selection strategy for IBIDR considering the dynamic participation rate and stochastic response deviation. First, users' dynamic participation rate, simultaneously influenced by both imitative behaviour and spontaneous behaviour, is modelled by a Markov process, which improves the user model. Second, an adaptive selection mechanism is proposed based on users' historical response power deviation levels and participation rates, which improves the MESP model. Moreover, the improved user model and the enhanced MESP model are jointly expressed as a bi‐level stochastic programming problem. Utilising the Karush–Kuhn–Tucker (KKT) conditions and probability theory, the original bi‐level stochastic problem is reformulated into a tractable single‐level deterministic optimisation model. Finally, simulation results demonstrate that the proposed model effectively improves the accuracy of user response behaviour prediction and the efficiency of MESP's incentive strategies, thereby reducing MESP's total cost, mitigating response power deviations, and enhancing the utilities for users participating in IBIDR.
- Open Access
- 5 September 2024
- Published by Institution of Engineering and Technology (IET) in IET Generation, Transmission & Distribution
- Vol. 18 (19) , 3071-3084
A coordinated scheduling optimization method for integrated energy systems with data centres based on deep reinforcement learning
As an emerging multi‐energy consumption subject, data centres (DCs) are bound to become crucial energy users for integrated energy systems (IES). Therefore, how to fully tap the potential of the collaborative operation between DCs and IES to improve total energy efficiency and economic performance is becoming a pressing need. In this article, the authors research an optimization coordinated by the energy scheduling and information service provision within the scenario of an integrated energy system with a data centre (IES‐DC). The mathematical model of IES‐DC is first established to reveal the energy conversion process of the electricity‐heat‐gas IES and the DC's energy consumption affected by the scale of active IT equipment. For dynamical providing multi‐energy and computing service by coordinating scheduling energy and information equipment, the formulations of IES‐DC scheduling, which is described as a Markov decision process (MDP), are presented, and it is solved by introducing the twin‐delayed deep deterministic policy gradient (TD3), which is a model‐free deep reinforcement learning (DRL) algorithm. Finally, the numerical studies show that compared with benchmarks, the proposed method based on the TD3 algorithm can effectively control the operation of energy conversion equipment and the number of active servers in IES‐DC.
- 5 September 2024
- Published by Institution of Engineering and Technology (IET) in IET Generation, Transmission & Distribution
- Vol. 18 (19) , 3071-3084
Open AccessA coordinated scheduling optimization method for integrated energy systems with data centres based on deep reinforcement learning
As an emerging multi‐energy consumption subject, data centres (DCs) are bound to become crucial energy users for integrated energy systems (IES). Therefore, how to fully tap the potential of the collaborative operation between DCs and IES to improve total energy efficiency and economic performance is becoming a pressing need. In this article, the authors research an optimization coordinated by the energy scheduling and information service provision within the scenario of an integrated energy system with a data centre (IES‐DC). The mathematical model of IES‐DC is first established to reveal the energy conversion process of the electricity‐heat‐gas IES and the DC's energy consumption affected by the scale of active IT equipment. For dynamical providing multi‐energy and computing service by coordinating scheduling energy and information equipment, the formulations of IES‐DC scheduling, which is described as a Markov decision process (MDP), are presented, and it is solved by introducing the twin‐delayed deep deterministic policy gradient (TD3), which is a model‐free deep reinforcement learning (DRL) algorithm. Finally, the numerical studies show that compared with benchmarks, the proposed method based on the TD3 algorithm can effectively control the operation of energy conversion equipment and the number of active servers in IES‐DC.
- Open Access
- 29 August 2024
- Published by Institution of Engineering and Technology (IET) in IET Renewable Power Generation
- Vol. 18 (14) , 2360-2380
Collaborative forecasting management model for multi‐energy microgrid considering load response characterization
Multi‐energy microgrids (MEMG) have become an effective means of integrated energy management due to their unique advantages, including area independence, diverse supply, flexibility, and efficiency. However, the uncertain deviation of the renewable energy generators (REGs) output and the uncertain deviation of the multiple energy load response cumulatively lead to the deterioration of the MEMG model performance. To address these issues, this article proposes a cooperative forecasting management model for MEMG that considers multiple uncertainties and load response knowledge characterization. The model combines a multi‐energy load prediction model with a management model based on deep reinforcement learning. It proposes multiple iterations of data, fits the dynamic environment of MEMG by continuously improving the long short‐term memory (LSTM) neural network based on knowledge distillation (KD) architecture, and then optimizes the MEMG state space by considering the knowledge of load response characteristics, Furthermore, it combines multi‐agent deep deterministic policy gradient (MADDPG) with horizontal federated (hF) learning to co‐train multi‐MEMG, addressing the issues of training efficiency during co‐training. Finally, the validity of the proposed model is demonstrated by an arithmetic example.
- 29 August 2024
- Published by Institution of Engineering and Technology (IET) in IET Renewable Power Generation
- Vol. 18 (14) , 2360-2380
Open AccessCollaborative forecasting management model for multi‐energy microgrid considering load response characterization
Multi‐energy microgrids (MEMG) have become an effective means of integrated energy management due to their unique advantages, including area independence, diverse supply, flexibility, and efficiency. However, the uncertain deviation of the renewable energy generators (REGs) output and the uncertain deviation of the multiple energy load response cumulatively lead to the deterioration of the MEMG model performance. To address these issues, this article proposes a cooperative forecasting management model for MEMG that considers multiple uncertainties and load response knowledge characterization. The model combines a multi‐energy load prediction model with a management model based on deep reinforcement learning. It proposes multiple iterations of data, fits the dynamic environment of MEMG by continuously improving the long short‐term memory (LSTM) neural network based on knowledge distillation (KD) architecture, and then optimizes the MEMG state space by considering the knowledge of load response characteristics, Furthermore, it combines multi‐agent deep deterministic policy gradient (MADDPG) with horizontal federated (hF) learning to co‐train multi‐MEMG, addressing the issues of training efficiency during co‐training. Finally, the validity of the proposed model is demonstrated by an arithmetic example.
- Open Access
- 15 August 2024
- Published by Institution of Engineering and Technology (IET) in IET Generation, Transmission & Distribution
- Vol. 18 (18) , 2956-2971
Integrated demand response optimization strategy considering risk appetite under multi‐dimensional uncertain information
Integrated demand response (IDR) is deemed as an effective tool to balance energy supply and demand. User’s uncertain information containing prior uncertain information and posterior uncertain information is a key factor affecting the implementation effectiveness of IDR, but existing studies fail to consider the two types of uncertain information, response risk caused by the uncertain information, and risk appetite comprehensively. Based on the principal‐agent theory of optimal incentive contract under uncertain information and Markowitz's mean‐variance portfolio theory, a new IDR model is established in this paper, and an IDR optimization strategy considering risk appetite under uncertain information is proposed. By proposing the user model considering multi‐dimensional uncertain information and the risk appetite‐based integrated energy service providers (IESP) model based on the principal‐agent theory and Markowitz's mean‐variance portfolio theory, we have achieved effective modelling of the user’s uncertain information and the risk borne by IESP. The arithmetic examples have verified advantages of the model in enhancing the accuracy of user’s actual response prediction and the superiority of incentive strategies, which is beneficial to reduce the cost of IESPs and enhance the benefit of users participating in IDR.
- 15 August 2024
- Published by Institution of Engineering and Technology (IET) in IET Generation, Transmission & Distribution
- Vol. 18 (18) , 2956-2971
Open AccessIntegrated demand response optimization strategy considering risk appetite under multi‐dimensional uncertain information
Integrated demand response (IDR) is deemed as an effective tool to balance energy supply and demand. User’s uncertain information containing prior uncertain information and posterior uncertain information is a key factor affecting the implementation effectiveness of IDR, but existing studies fail to consider the two types of uncertain information, response risk caused by the uncertain information, and risk appetite comprehensively. Based on the principal‐agent theory of optimal incentive contract under uncertain information and Markowitz's mean‐variance portfolio theory, a new IDR model is established in this paper, and an IDR optimization strategy considering risk appetite under uncertain information is proposed. By proposing the user model considering multi‐dimensional uncertain information and the risk appetite‐based integrated energy service providers (IESP) model based on the principal‐agent theory and Markowitz's mean‐variance portfolio theory, we have achieved effective modelling of the user’s uncertain information and the risk borne by IESP. The arithmetic examples have verified advantages of the model in enhancing the accuracy of user’s actual response prediction and the superiority of incentive strategies, which is beneficial to reduce the cost of IESPs and enhance the benefit of users participating in IDR.
- 4 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 20 (6) , 9037-9047
Learning-Based Terminal-Edge Collaborative Energy-Efficient Routing Algorithm for Green RWSN
In recent years, wireless smart sensors powered by solar energy have been widely deployed to ensure the green and sustainable operation of remote industrial systems monitoring. Such devices can offload computing tasks locally or using the edge server by transmitting the raw data wirelessly. Since random renewable energy harvesting has a detrimental effect on the energy balance of nodes in these green rechargeable wireless sensor networks (RWSN), the rational synergy between terminal-edge collaborative tasks offloading (TECTO) and network topology optimization (NTO) is of great significance for improving the sustainability. Therefore, this article presents a learning-based terminal-edge collaborative energy-efficient routing algorithm. First, a system model is developed to integrate TECTO and NTO, and the original problem is decoupled into two layers. Then, the NTO layer is aimed at quickly generating an energy-efficient network topology by variable cycle block coordinate descent method based on the greedy strategy. Finally, the TECTO layer adopts deep reinforcement learning based on the dynamic baseline to understand the energy efficiency feedback law of the NTO layer and rationally adjusts the TECTO scheme. The simulation results show that the presented algorithm can reasonably generate the network topology and TECTO scheme according to the node's energy state change and efficiently consume the renewable energy distributed in the green RWSN, which significantly enhances its sustainability.
- 4 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 20 (6) , 9037-9047
Learning-Based Terminal-Edge Collaborative Energy-Efficient Routing Algorithm for Green RWSN
In recent years, wireless smart sensors powered by solar energy have been widely deployed to ensure the green and sustainable operation of remote industrial systems monitoring. Such devices can offload computing tasks locally or using the edge server by transmitting the raw data wirelessly. Since random renewable energy harvesting has a detrimental effect on the energy balance of nodes in these green rechargeable wireless sensor networks (RWSN), the rational synergy between terminal-edge collaborative tasks offloading (TECTO) and network topology optimization (NTO) is of great significance for improving the sustainability. Therefore, this article presents a learning-based terminal-edge collaborative energy-efficient routing algorithm. First, a system model is developed to integrate TECTO and NTO, and the original problem is decoupled into two layers. Then, the NTO layer is aimed at quickly generating an energy-efficient network topology by variable cycle block coordinate descent method based on the greedy strategy. Finally, the TECTO layer adopts deep reinforcement learning based on the dynamic baseline to understand the energy efficiency feedback law of the NTO layer and rationally adjusts the TECTO scheme. The simulation results show that the presented algorithm can reasonably generate the network topology and TECTO scheme according to the node's energy state change and efficiently consume the renewable energy distributed in the green RWSN, which significantly enhances its sustainability.
- 23 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Smart Grid
- Vol. 15 (4) , 3624-3638
A Risk Diversification Strategy for Integrated Demand Response Under Imperfect Rationality
Integrated demand response (IDR) is acknowledged as a cost-effective and low-carbon tool that helps alleviate imbalances between energy supply and demand in energy systems. However, the imperfect rationality of consumers can lead to a series of new issues, such as cognitive characteristics of bounded rationality as well as irrationality, subjective uncertainty, and correlated risks. This paper proposes an improved consumer model by using a dynamic subjective weight function with a rationality degree indicator to cope with cognitive characteristics, coupled with a deduced skew distribution modelling the subjective uncertainty. Besides, MESP model is also improved to deal with the correlated risks by managing the diversification level of risks that is defined based on Shannon entropy. The mathematical formulation of the proposed model is expressed as a bi-level stochastic optimization problem, which is equivalently converted into a single-objective optimization problem to be solved efficiently. Simulation results validate advantages of our model in enhancing the accuracy of consumer behavior prediction, effectiveness of incentive strategies, diversification level of response risks, which contributes to achieve a win-win situation between consumers and MESP by decreasing not only total response power deviation as well as cumulative response risks of MESP but also consumer’s discomfort level.
- 23 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Smart Grid
- Vol. 15 (4) , 3624-3638
A Risk Diversification Strategy for Integrated Demand Response Under Imperfect Rationality
Integrated demand response (IDR) is acknowledged as a cost-effective and low-carbon tool that helps alleviate imbalances between energy supply and demand in energy systems. However, the imperfect rationality of consumers can lead to a series of new issues, such as cognitive characteristics of bounded rationality as well as irrationality, subjective uncertainty, and correlated risks. This paper proposes an improved consumer model by using a dynamic subjective weight function with a rationality degree indicator to cope with cognitive characteristics, coupled with a deduced skew distribution modelling the subjective uncertainty. Besides, MESP model is also improved to deal with the correlated risks by managing the diversification level of risks that is defined based on Shannon entropy. The mathematical formulation of the proposed model is expressed as a bi-level stochastic optimization problem, which is equivalently converted into a single-objective optimization problem to be solved efficiently. Simulation results validate advantages of our model in enhancing the accuracy of consumer behavior prediction, effectiveness of incentive strategies, diversification level of response risks, which contributes to achieve a win-win situation between consumers and MESP by decreasing not only total response power deviation as well as cumulative response risks of MESP but also consumer’s discomfort level.
- Open Access
- 19 January 2024
- Published by Association for Computing Machinery (ACM)
- p. 148-152
Application of data-driven user-transformer relationship identification technology in load control of station area
With the rapid development of the new power system, the proportion of distributed new energy and new power load in the intelligent distribution area increases, which brings great challenges to the lean operation of the low-voltage power distribution system. This paper systematically sorts out the typical classification and regulation requirements of load regulation at the present stage, discusses its advantages and disadvantages, applicability and implementation difficulties, and proposes a data-driven load regulation framework for low-voltage transformer area. Through sensing and processing the data of intelligent sensing equipment and measuring equipment provided by intelligent acquisition devices, and integrating topological recognition technology, accurate user-transformer relationship relationship is obtained. It provides a research idea for the fine regulation and implementation of low-voltage transformer area load.
- 19 January 2024
- Published by Association for Computing Machinery (ACM)
- p. 148-152
Open AccessApplication of data-driven user-transformer relationship identification technology in load control of station area
With the rapid development of the new power system, the proportion of distributed new energy and new power load in the intelligent distribution area increases, which brings great challenges to the lean operation of the low-voltage power distribution system. This paper systematically sorts out the typical classification and regulation requirements of load regulation at the present stage, discusses its advantages and disadvantages, applicability and implementation difficulties, and proposes a data-driven load regulation framework for low-voltage transformer area. Through sensing and processing the data of intelligent sensing equipment and measuring equipment provided by intelligent acquisition devices, and integrating topological recognition technology, accurate user-transformer relationship relationship is obtained. It provides a research idea for the fine regulation and implementation of low-voltage transformer area load.
- 5 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Green Communications and Networking
- Vol. 8 (2) , 601-618
Online Collaborative Energy-Network Resource Scheduling for WPT-Enabled Green Edge Computing
The operations of IoT devices (IoTD) and 5G base station (BS) contribute to most of carbon emission and on-power energy consumption in wireless edge network. To reduce operational costs and achieve low-carbon computing, this paper investigates a long-term average operational expenditure (OPEX) minimization problem, and proposes an online joint energy-network resource scheduling algorithm, including computation offloading, wireless power transmission, energy sharing, and task migration in wireless edge network powered by renewable energy, energy storage, and power grid. Differing from existing works, ours consider the energy loss of battery, dynamic carbon emission related with on-power energy, and constraint of spatial electric network into our model. Then, we apply the Lyapunov technique to decompose the proposed problem into three real-time sub-problems. Furthermore, a federal gradient descent based full-distributed online algorithm is proposed to obtain solution while protecting the privacy of network operators. We also prove the convergence of proposed algorithm and provide the tradeoff between network stability and optimal OPEX. Simulation results reveal that the proposed algorithm outperforms existing benchmarks in reducing on-grid power dependence and OPEX.
- 5 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Green Communications and Networking
- Vol. 8 (2) , 601-618
Online Collaborative Energy-Network Resource Scheduling for WPT-Enabled Green Edge Computing
The operations of IoT devices (IoTD) and 5G base station (BS) contribute to most of carbon emission and on-power energy consumption in wireless edge network. To reduce operational costs and achieve low-carbon computing, this paper investigates a long-term average operational expenditure (OPEX) minimization problem, and proposes an online joint energy-network resource scheduling algorithm, including computation offloading, wireless power transmission, energy sharing, and task migration in wireless edge network powered by renewable energy, energy storage, and power grid. Differing from existing works, ours consider the energy loss of battery, dynamic carbon emission related with on-power energy, and constraint of spatial electric network into our model. Then, we apply the Lyapunov technique to decompose the proposed problem into three real-time sub-problems. Furthermore, a federal gradient descent based full-distributed online algorithm is proposed to obtain solution while protecting the privacy of network operators. We also prove the convergence of proposed algorithm and provide the tradeoff between network stability and optimal OPEX. Simulation results reveal that the proposed algorithm outperforms existing benchmarks in reducing on-grid power dependence and OPEX.
- 12 September 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 11 (4) , 7013-7023
Adaptive Payoff Balance Among Mobile Wireless Chargers for Rechargeable Wireless Sensor Networks
Wireless power transfer (WPT) based on mobile platforms such as unmanned aerial vehicle (UAV) has been recognized as a promised technique to prolong battery lifetime of resource-constrained wireless sensors in the Internet of Things (IoT) era. However, it is challenging to collaborate multiple mobile wireless chargers (MWCs) for coordinating heterogeneous energy requirements of massive rechargeable wireless sensors, where the efficient optimization of quantitative collaboration utility among MWCs is difficult. Hence, this paper investigates the adaptive payoff balance among MWCs for rechargeable wireless sensor networks (RWSN). Firstly, the collaborative wireless powered system based on charging sectors control among MWCs is proposed. Then, the charging payoff function and the corresponding optimization problem for maximizing the minimum payoff are designed, and it is decomposed into two layers by the hierarchical decompose method to be solved quickly. In the bottom layer, the payoff of each MWC with the given charging sector is maximized by the convex optimization theory. According to the payoff feedback of the bottom layer, intelligent charging sectors allocation is realized by the deep reinforcement learning. The simulation results show that our algorithm can ensure efficient energy allocation of any single MWC, and the overall utility of collaborative wireless powered system on this basis can be optimized by rationally allocating charging sectors, which significantly improves the sustainability of RWSN.
- 12 September 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 11 (4) , 7013-7023
Adaptive Payoff Balance Among Mobile Wireless Chargers for Rechargeable Wireless Sensor Networks
Wireless power transfer (WPT) based on mobile platforms such as unmanned aerial vehicle (UAV) has been recognized as a promised technique to prolong battery lifetime of resource-constrained wireless sensors in the Internet of Things (IoT) era. However, it is challenging to collaborate multiple mobile wireless chargers (MWCs) for coordinating heterogeneous energy requirements of massive rechargeable wireless sensors, where the efficient optimization of quantitative collaboration utility among MWCs is difficult. Hence, this paper investigates the adaptive payoff balance among MWCs for rechargeable wireless sensor networks (RWSN). Firstly, the collaborative wireless powered system based on charging sectors control among MWCs is proposed. Then, the charging payoff function and the corresponding optimization problem for maximizing the minimum payoff are designed, and it is decomposed into two layers by the hierarchical decompose method to be solved quickly. In the bottom layer, the payoff of each MWC with the given charging sector is maximized by the convex optimization theory. According to the payoff feedback of the bottom layer, intelligent charging sectors allocation is realized by the deep reinforcement learning. The simulation results show that our algorithm can ensure efficient energy allocation of any single MWC, and the overall utility of collaborative wireless powered system on this basis can be optimized by rationally allocating charging sectors, which significantly improves the sustainability of RWSN.
- Open Access
- 1 September 2023
- Published by Institution of Engineering and Technology (IET) in IET Renewable Power Generation
- Vol. 18 (3) , 456-475
A collaborative training approach for multi energy systems in low‐carbon parks accounting for response characteristics
Existing researches lack systematic design of the operation mechanism and overall implementation process of park oriented multi‐state energy system; The existing researches on multi‐state energy systems ignored the dynamic response characteristics of load and the coupling effect of units considering the impact of carbon emissions, and did not go deep into the collaborative management of multiple multi state energy system parks. To this end, this paper first proposes a two‐stage multi‐energy system operation architecture for low carbon parks, and establishes a mathematical model considering carbon emissions, operating costs and other objectives. The response space of load is optimized by considering the dynamic interaction characteristics of load response and the coupling characteristics of unit operation. Then, considering the problems of data security and training efficiency, F‐DDPG technology is used to conduct collaborative training for multiple campus systems. Finally, an example is given to prove the effectiveness of the method, which is not only conducive to reducing the operating cost of the system and the carbon emissions on the supply side, but also improves the convergence of model training. In the case of collaborative training in multiple parks, the method in this paper converges faster and has better economic benefits.
- 1 September 2023
- Published by Institution of Engineering and Technology (IET) in IET Renewable Power Generation
- Vol. 18 (3) , 456-475
Open AccessA collaborative training approach for multi energy systems in low‐carbon parks accounting for response characteristics
Existing researches lack systematic design of the operation mechanism and overall implementation process of park oriented multi‐state energy system; The existing researches on multi‐state energy systems ignored the dynamic response characteristics of load and the coupling effect of units considering the impact of carbon emissions, and did not go deep into the collaborative management of multiple multi state energy system parks. To this end, this paper first proposes a two‐stage multi‐energy system operation architecture for low carbon parks, and establishes a mathematical model considering carbon emissions, operating costs and other objectives. The response space of load is optimized by considering the dynamic interaction characteristics of load response and the coupling characteristics of unit operation. Then, considering the problems of data security and training efficiency, F‐DDPG technology is used to conduct collaborative training for multiple campus systems. Finally, an example is given to prove the effectiveness of the method, which is not only conducive to reducing the operating cost of the system and the carbon emissions on the supply side, but also improves the convergence of model training. In the case of collaborative training in multiple parks, the method in this paper converges faster and has better economic benefits.