Voltage Analysis And Optimal Planning Of Pv-integrated Distribution Network

Due to the development of renewable generation technology,distributed Photovoltaic(PV)generation’s penetration is steadily increasing in distribution network,which causes the problem of voltage limit violations and power reverse transmission,affected by the PV’s decentralized,instable,and intermittent power generation.While reducing traditional energy consumption and improving the environment,distributed PVs also bring risks to the distribution network.According to the power flow calculation algorithm,the operation status of the distribution network is determined by the real-time load and distributed output power of each node.Therefore,accurate load and PV generation estimation is necessary for voltage risk assessment and the PV installation planning.Under the scenario of PV-integrated network,the paper finishes the short-term load forecasting on the load side,and scenario generation of PV’s output on the PV side.Based on the result of both sides,the paper do research on the location and capacity optimization of PV planning.First,aiming at the problem of short-term load forecasting on the load side,a short-term load forecasting model based on the Hilbert-Huang decomposition of long-and short-term memory neural network is proposed.Considering the high volatility and uncertainty of load data,a sole load forecasting method could not meet the accuracy requirements,so the idea of load decomposition is proposed.The load sequence is decomposed into several components by Hilbert-Huang transformation,to which the deep long and short-term memory neural network(LSTM)is utilized to predict respectively.The final forecasting value is reconstructed by the predicted value of each component.The accuracy of load forecasting model is evaluated on the Tensorflow-based deep leaning platform.Secondly,aim at the scenario generation on the PV side,data-driven distribution system scenario generation method is proposed in this paper.The model is based on local density based clustering(LDC)algorithm and copula functions,with considering the change of correlation of PV station’s output.LDC is used to classify the historical data into several clusters.In each cluster,copula function is utilized to build the joint probabilistic density functions(PDF)to reflect the inherent relations between PV stations.The scenario data are generated by applying Latin Hypercube Sampling(LHS)to the PDF of PVs’ output power.The proposed method aims to obtain the joint probabilistic distribution of multiple PV outputs more accurately.The proposed method is conducted in the probabilistic power flow calculation of an 30-node,real world based 10 k V distribution network in Jiangsu.Finally,the optimization of PV location and capacity optimization was researched.The distributed PV location optimization model is constructed with the objective function of probabilistic voltage sensitivity and network loss,while the capacity optimization model is constructed with the objective function of voltage over-limit risk,annual network loss,and installed PV capacity.To solve the problem that the existing optimization algorithm is not effective in optimizing multi-objective problem,the NSGA-III algorithm is introduced as the solution to generate the Pareto front.In addition,the influence of medium and long-term load increase on PV optimization results is considered.The effectiveness of the optimization method is also verified by conducting PV planning on the same distribution network in Scenario generation’s case study.