Studies On Online Sequential Decision-Making For Power Load Restoration By Source And Load Cooperation

In recent years,the global climate has been deteriorating,leading to an increase in the frequency and intensity of extreme weather events and power outage incidents.Substations,as the basic units of the power system,serve as the link between the transmission and distribution networks.Improving the load restoration capability of substations after outages can effectively reduce the impact of outages and minimize economic losses,which is an effective method to enhance power system resilience.However,under extreme weather conditions,the power supply capacity of the transmission network will decrease,and the demand of air conditioners will increase,leading to a sharp increase in the load pickup amount after outages,and the challenge of source-load matching for substation load restoration.In addition,the changes of the power grid operation conditions and external environmental are difficult to predict after outages,and the restoration process is uncertaint.Therefore,in-depth research on online sequential decision-making methods for power load restoration by source and load cooperation is of great theoretical significance and practical value for effectively improving the restoration ability of power system and enhancing grid resilience.This dissertation aims at the challenges of insufficient power supply capacity of upper-level power grid and sharp increase of air conditioners load pickup amount caused by extreme weather.Taking substation as the core and feeder as the basic unit,this dissertation uses modern communication technology to realize adaptive source-load matching and develop the online sequential decision-making for substation load restoration in the restoration process.Based on an equivalent thermal parameter model for air conditioner,an aggregate power evaluation model is established,which includes both fixed-frequency and inverter air conditioner groups,to rapidly and accurately evaluate the load pickup amount after outages.By fully utilizing the demand response capability of air conditioners,a two-stage optimization decision-making method is proposed to formulate safe and efficient feeder restoration schemes for substation load restoration in scenarios where the upper-level power grid supply capacity is insufficient.After completing the feeder restoration operation,the power supply capacity of the transmission and the microgrid are evaluated in real-time,taking into account the dynamic changes in air conditioners during extreme weather events,and a sequential decision-making method based on deep reinforcement learning is proposed for feeder critical load restoration.Building on previous research,an online sequential decision-making method for coordinated source-load power load restoration are studied in this dissertation,utilizing techniques such as evolutionary computation,model predictive control,distributed robust optimization,and deep reinforcement learning.The main contributions and innovations of the dissertation are described as follows:(1)A fast load pickup amount evaluation method considering the pickup characteristics of air conditioners is proposed by deriving the fixed-frequency air conditioners and inverter air conditioners power evaluation model based on thermodynamic processes.The probability of the fixed-frequency air conditioner being turned on and the relationship between the working frequency of the inverter air conditioner and the outage duration and environmental temperature are derived based on the equivalent thermal parameter model of an air conditioner.The heterogeneous air conditioner group power evaluation model containing fixed-frequency and inverter air conditioners is established.The empirical mode decomposition method is used to decompose the pre-outage load data into thermostatically controlled loads and non-thermostatically controlled loads.The air conditioners are selected as the representative of thermostatically controlled loads.Based on the outage data,meteorological data,and statistical information of air conditioner parameters,the ratio of the post-outage aggregate power of the fixed-frequency and inverter air conditioner groups to the pre-outage power is calculated,and the range of the thermostatically controlled loads pickup amount is estimated to quickly obtain the range of the load pickup amount after the outage.The calculation results show that the proposed air conditioner group power evaluation model can quickly and accurately evaluate the aggregate power of the air conditioner groups under different outage duration and environmental temperature,considering the proportion of fixed-frequency and inverter air conditioners.The analysis using the proposed model shows that the higher the proportion of fixed-frequency air conditioners in the air conditioner group and the higher the environmental temperature,the greater the ratio of the post-outage aggregate power to the pre-outage.In actual outage scenarios,the proposed fast evaluation method can quickly obtain the value range of the substation load pickup amount,which is consistent with the load pickup amount collected by the system,which has high practical value.(2)In scenarios where extreme weather results in insufficient power supply from the upper-level power grid,a substation load restoration optimization method for source-load adaptive matching is proposed to explore the demand response potential of air conditioners during the restoration process.In order to adapt to the constantly changing power grid operating conditions and outage scenarios,an adaptive load restoration framework is established to guide the substation load restoration process by considering the flexible control of air conditioners and the supply capability of multiple recovery resources.To ensure load restoration safety constraints,a two-stage substation feeder restoration optimization method is proposed,which uses multi-step substation feeder restoration optimization and single-step air conditioning control online correction to ensure source-load adaptive matching during the restoration process.With the optimization objective of maximizing load restoration benefits,a multi-step substation feeder restoration model considering the flexible control of air conditioners is established,and the particle swarm optimization algorithm is used to optimize the feeder restoration scheme and air conditioners control schemes in several steps.To cope with the uncertainty of the restoration process and reduce the impact of air conditioner control on customers,a mixed-integer linear programming model for air conditioners control is established to ensure the reliability of single-step feeder restoration operations.The simulation results show that the proposed method can improve load restoration efficiency by avoiding failure caused by excessive load pickup amount beyond the maximum restorable load amount by the flexible control of air conditioners.This two-stage load restoration optimization method can fit to changes in load composition,environmental temperature,and available power of the upper-level power grid.It also formulates the optimal load restoration plan,ensures real-time source-load matching,and ensures reliable feeder restoration while reducing the impact of air conditioners control on customers.(3)To cope with the uncertainty of power supply capability and load restoration in transmission and distribution microgrids,a sequential decision-making method for substation feeder load restoration with interaction of transmission,distribution,and micro-grids is proposed to coordinate top-down and bottom-up recovery power to ensure key feeder load restoration.Based on a multi-agent framework for substation feeder restoration decision-making,the substation feeder load restoration problem is decomposed into different subtasks and allocated to transmission/distribution/microgrid agents to complete the load restoration decision-making through multi-agent interaction.Considering the source-load uncertainty in microgrids,a power supply capability evaluation model based on distributed robust optimization is established to quickly evaluate the microgrid power supply capability as the basis for feeder load restoration decision-making.Based on the theory of Markov decision processes,a reinforcement learning model for feeder load restoration is established,and the deep Q-network is used to process the model to achieve sequential decision-making for feeder load restoration.The simulation results show that the proposed power supply capability evaluation method based on distributed robust optimization can quickly evaluate the microgrid power supply capability according to the decision maker’s requirements for different levels of risk.Compared with traditional optimization methods,the proposed sequential decision-making method based on deep Q-network has little difference in load restoration optimization performance,but has significant advantages in computational efficiency.The proposed method can fully utilize the power supply capability of transmission and microgrids to ensure the reliable restoration of feeder critical load in the substation and obtain higher load restoration benefits.