Research On Multi-Time Scale Optimization And Control Strategy Of Active Distribution Network Considering Different Source And Load Characteristics

With the large-scale integration of distributed power sources,air conditioning/heat pumps,and electric vehicles into distribution systems,modern distribution systems become more flexible and effective.Yet the uncertainties and fluctuations have introduced many challenges to the economic and safe operation of distribution systems.This thesis proposes integrated optimization and control strategies for different time scales based on the characteristics of different power sources and loads and the operational requirements of distribution networks.The proposed strategies try to fill the research gap of economic and safe operation of the active distribution systems by making full use of the flexibility of the renewables and loads.Specifically,this thesis proposes a coordinated scheduling strategy based on a distributed optimization model for flexible power sources and loads,and a real-time scheduling strategy based on deep reinforcement learning for highly random power sources and loads.Furthermore,an integrated optimization and control strategy for dual time scales is proposed with different time scale characteristics.In addition to improving economic efficiency,the safety operation issues need to be guaranteed in the optimization and control of active distribution systems at different time scales,i.e.,network congestion and voltage limits.This thesis conducts research on optimized scheduling and safety control issues of active distribution systems considering different power source and load characteristics.The main research topics of the thesis are summarized as follows:(1)To address the research issue that conventional centralized scheduling methods are hard to coordinate a large number of distributed power sources and loads,a distributed scheduling strategy for active distribution systems is proposed,considering user privacy.First,the system model of the AC/DC hybrid microgrid-active distribution network is established,where the optimization objective is to reduce the operating costs of the distribution network and multiple microgrids while minimizing system network losses under voltage safety constraints.Then,considering the information barriers and privacy protection requirements,the centralized optimization problem is decomposed into multiple sub-problems using the auxiliary problem principle.Finally,the alternating direction multiplier method is used to solve the primal problem in a distributed manner.The results of the case study show that the proposed distributed solution can achieve fast convergence while the formulated scheduling plan by the interconnected systems achieves power mutual assistance and effectively reduces system operating costs.(2)To address the safety operation issues(e.g.,network congestion)caused by the mismatch between power sources and loads during the day-ahead scheduling,a distributed coordination strategy for active distribution systems is proposed considering large-scale pre-cooling control of air conditioning.First,a demand response model with an efficient solution at the scheduling time scale is established based on an equivalent thermal model of residential,which assumes the air conditioning load as an automatic pre-cooling/pre-heating price response load.Secondly,considering the reactive power support capability of photovoltaic inverters,voltage safety constraints,and network congestion constraints,a locational marginal electricity pricing model of the distribution network is established to regulate the air conditioning loads.Then,considering the equilibrium of the interests of the distribution network operator and users,the mechanism of the failure of large-scale demand response regulation is analyzed thoroughly.Finally,a distributed coordination strategy for large-scale demand response is proposed.The case study validates that the proposed method can effectively reduce the operating costs of distribution systems while avoiding network congestion and voltage limits caused by the mismatch between power sources and loads.And further,facilitate renewable energy consumption.The proposed distributed coordination framework can greatly reduce calculation time to enable daily scheduling.(3)To address the research issue of real-time scheduling of random electric vehicle load in active distribution systems,a real-time scheduling strategy based on deep reinforcement learning for active distribution system is proposed.The real-time scheduling problem of the system,which is affected by the high randomness of electric vehicles,is modeled as a Markov decision process with unknown state transition probabilities.The deep deterministic policy gradient algorithm is adopted to solve the problem.To enable the intelligent agent to fully perceive the environment,a state perception model of electric vehicles is established.By interacting with the fuzzy controller of electric vehicles,the model provides a decision-making basis for the intelligent agent.To improve the learning capabilities of the algorithm,serval corresponding improvement strategies are proposed in this thesis.The results of case study show that the proposed strategy can achieve real-time scheduling of the system with randomness while significantly reducing the deviation between load and photovoltaic output.The proposed method can effectively guarantee the source-load coordination and improve the economic operation of the distribution system.(4)To address the safety issue caused by the rapid fluctuation of source/load during real-time scheduling,a real-time voltage control strategy for active distribution network with distributed photovoltaic generation is proposed.The real-time voltage control problem is modeled as a constraint Markov problem with the minimum network loss.The multi-agent deep reinforcement learning algorithm is adopted to solve it.To address the safety issue of constraint Markov problem,a safety projection strategy is proposed to ensure safety and promote effective exploration of the multi-agent policy.To address the communication delay of the multi-agent deployment,a real-time decision-making algorithm based on state synchronization is proposed.The results show that the proposed strategy can make real-time decisions with local observation information and achieve no voltage violation with minimum network loss.(5)To address the research issue of coordination of source-load at different time scales,a hierarchical reinforcement learning based dual time scale optimized control strategy for active distribution network is proposed.Firstly,the dynamic pricing problem of charging stations at a slow time scale is modeled as a Markov decision process with maximizing the long-term revenue of the distribution network.Then,the real-time control of photovoltaic inverters at the fast time scale is modeled as a constraint Markov decision process with minimizing network loss,and a safety projection network is used to transform the problem into a general Markov decision process.Finally,an improved hierarchical reinforcement learning strategy is proposed to solve the above problems.The experimental results show that the proposed strategy can effectively coordinate source load resources at different time scales and improve the operation efficiency of the distribution network in a highly uncertain environment with no voltage violations.In summary,the optimized control strategies are proposed for active distribution systems at different time scales considering the multi-source characteristics in the active distribution systems,which can ensure the economic and safe operation of the distribution systems.The optimization and control strategies proposed in this thesis research can be applied to different types of distribution systems.