Intermittent Energy Output Power Forecasting And Storage System Planning

The forecasting of intermittent energy output power and the installation ofstorage equipment are the effective ways to mitigate the negative influence on powersystem operation and management which caused by the integration of intermittentenergy in power grid. This thesis focuses on issues related to the forecasting of theintermittent energy output power and the planning of storage systems.First of all, a wind speed forecasting model based on improved empirical modedecomposition(EMD) and GA-BP neural network is proposed. In comparison with thetest results of the traditional GA-BPNN model and the combined EMD and GA-BPNN model, it shows that the forecasting accuracy of the proposed method is moreaccurate than them, and the proposed method is suitable for ultra short term(10min)and short term(1hour) wind speed forecasting.Secondly, for photovoltaic output forecasting, two methods are proposed. One isa combination model based on EEMD and GA-BP for forecasting the hourlyirradiance under similar day. Another is a combination model based on grey modeland BP neural network for predicting the photovoltaic output power directly. Theresult of test example shows that the two forecasting methods have higher accuracythan the other methods. Those methods have a potential value in practical applications.Thirdly, this thesis proposes a storage planning model and a method under thedefinite transmission network with wind farms, and also a coordinated planningmodel with energy storage station and transmission network. According to the resultsof a test system, the rationality of the models and efficiency of the method are verified.Meanwhile, the comparison and analysis of different planning models are discussed.The proposed model and algorithm can provide a theoretical reference for thedevelopment of future grid planning with intermittent energy.Fourthly, this thesis proposes a planning model and a method for the optimalsizing and siting of ESS in distribution network with high penetration photovoltaics.Using the net present value (NPV) of the total system cost after ESS allocation asobjective function and meeting the constraints of the ESS operating control strategyand the distribution network operation, an improved adaptive particle swarmoptimization (APSO) algorithm is proposed to solve the problem of optimal allocationof multiple types of energy storage systems in distribution networks with highpenetration photovoltaics. The comparison and analysis of the influence on economic and node voltage by different ESSs in the distribution network are conducted. Finallythe feasibility of the models and the efficiency of the method are verified bysimulation results of a test system.Finally, by considering whether electric vehicle feed energy into the grid andwhether electric vehicle controlled by electricity price, this thesis builds three kinds ofelectric vehicle load models based on Monte Carol simulation. The impact of threedifferent electric vehicle load models and different scale of electric vehicles on thegrid is simulated and discussed. Meanwhile, a large-scale electric vehicle chargingmanagement strategy with demand response is proposed. The proposed strategy notonly meets the requirement of system dispatching but is also customer satisfactionconstrained. In order to provide ancillary services for intermittent energy, the abilityof adjustment of electric vehicle, as a form of energy storage to implement auxiliaryservice, is discussed.