Learning Optimization Of The Dispatch For Inter-regional Power Grid Considering The Stochasticity Of Source And Load

Inter-regional power grid has the advantages of allotting the power resource,promoting reliability of power grids and utilization ratio of renewable energy.However,the uncertainties of new energy sources such as wind power,photovoltaic and load demand,as well as the expansion of system scale caused by regional interconnection,all require for high accuracy and rapidity of inter-regional power grid dispatching.Therefore,in order to maintain safe,economical and clean operation of inter-regional power grid system,it is necessary to find an intelligent automatic dispatching method for inter-regional power grid.The inter-regional power grid system consisting of thermal power generator units,wind units,photovoltaic units,rigid loads and flexible loads is researched in this dissertation.On purpose of reducing the overall operation cost of inter-regional power grid system,the dispatch problem of direct current tie-line and two-layer in inter-regional power grid with uncertain renewable source and load are researched.Firstly,because of the time-series correlation of the renewable energy output and load demand in inter-regional power grids,the stochastic dynamic characteristics of source and load are formulated as continuous time Markov processes approximately.And on the basis,the regional source and load model and the inter-regional direct current tie-line model are established according to the characteristics of each unit of inter-regional power grid system.Secondly,based on practical operation requirements including the power balance constraint,the optimal dispatch problem for tie-line in inter-regional power grid is described as a discrete time Markov decision process model.According to the power of renewable energy output and load demand,the optimization strategy for the power of tie-line in each period is adjusted to promote the running benefit of the system in this model.Finally,a learning optimization method is adopted to obtain the optimization strategy,simulation results show that the operational efficiency of inter-regional power grid is significantly enhanced by the proposed learning optimization method which can cope with the stochasticity of source and load.In addition,according to the actual power grid dispatch,the dispatching process of inter-regional power grid is divided into tie-line dispatch of upper layer and unit power adjustment,flexible load scheduling from different areas of lower layer.Then on the base of these,the corresponding hierarchical discrete Markov decision process model is established.Finally,a hierarchical reinforcement learning method is adopted to obtain the optimization strategy of the model,and IEEE 300 bus system is adopted in the simulation results and analysis.The strategy can realize automatic adjustment of tie-lines and units in inter-regional power grid systems and automatic removal of flexible loads from different areas,which verifies the effectiveness of the proposed method.Meanwhile,this dissertation combines the transfer learning method into the hierarchical optimization algorithm.By transferring the regional knowledge from pre-learning to the hierarchical optimization process,the learning speed is accelerated extremely and the cost of re-learning is reduced.