Research On Optimal Dispatching Of Distributed Energy Storage In Distribution Network Based On Power Prediction

Large scale penetration of renewable energies like wind power and PV into distribution networks has brought great challenges to the security.At the same time,traditional power sources such as thermal power and hydropower have problems such as long response time and low climbing speed,which cannot meet the development needs of distribution networks under the condition of high energy and high permeability.There are two main ways to improve the stability of new energy access.One is to use energy storage devices to balance the fluctuations of new energy output through charge and discharge operations;the other is to use new technologies to accurately predict wind、light and load power.Short-term forecasting provides a reference for grid scheduling and rotating standby scheduling.In this paper,the distributed energy storage system of distribution network is taken as the research object,and the photovoltaic,wind power and load forecasting technologies are explored and discussed in detail.Based on the power forecast data,the optimized scheduling strategy of distributed energy storage is deeply studied.Mainly carry out the following work:Firstly,based on the wind,light and load power prediction problems,a power prediction model based on long short term memory(LSTM)network is proposed,and the model is used to predict the power demand power interval.The basic principles and structures of traditional neural networks and LSTM are deeply studied.The development process,advantages and disadvantages of neural network technology are analyzed.The advantages of LSTM in dealing with "gradient" problems are discussed.The power prediction process is explored.The LSTM power prediction model is constructed;the energy storage power demand interval is established based on the prediction result,and the energy storage system scheduling plan is provided.By comparing with BP neural network power prediction results,the prediction results prove that the prediction model of long-short neural network has better prediction performance and has certain guiding significance for optimal scheduling of energy storage.Then,aiming at the distributed active power scheduling of energy storage,a distributed optimal scheduling model based on equal cost incremental rate is proposed.Taking the battery as an example,considering the life cost,charge and discharge loss and network loss of the energy storage system,the cost function of energy storage scheduling is constructed.The main application scenarios of the energy storage system are analyzed,and the energy storage scheduling plan is optimized based on the correlation coefficient.In order to solve the problem of distributed energy storage power distribution,a scheduling model based on the principle of energy storage cost incremental rate is proposed to optimize the distribution of energy storage active power.The rationality and economy of the proposed method are verified by an example.Finally,aiming at the optimal scheduling problem of distributed energy storage in distribution network,an optimal scheduling strategy of distributed energy storage in different layers and zones is proposed.In order to minimize the switching power of the tie line,the distribution network is divided into regions.Establishment of distributed energy storage layered partition coordination strategy;Example analysis based on PG&E69 node distribution system verifies the effectiveness and economy of the proposed method.