A Study On Short-term Load Forecasting Of Feeders Considering Photovoltaic Users

With the reform of electricity market system and the rapid development of smart distribution network,our country encourages the development of new energyin order to deal with the problem of environmental pollution and energy shortage.Solar energy,as one of the clean energy,has attracted more and more attention and gradually promoted the use of photovoltaic power generation in the electrical power system.When photovoltaic power generation technology is widely used,more and more photovoltaic users will be introduced into the power distribution system,which will change the original power flow direction and the load distribution of the power distribution system.If the traditional feeder short-term load prediction method is still used,the fluctuation and intermittent of photovoltaic load will cause large errors in the prediction results.Therefore,an in-depth study on short-term load prediction of feeders considering photovoltaic users is an improvement and optimization of traditional feeder short-term load prediction.It will assist the power distribution system to complete planning,power supply and other work,which has a very important practical significance.Among many load forecasting methods,artificial neural network algorithm can autonomously adapt to a large number of non-structural and inaccurate laws and has strong nonlinear function fitting ability and learning ability.So it can get high accuracy in short-term load forecasting of power distribution system.Therefore,based on the neural network algorithm,this paper studies the short-term load prediction of feeders considering photovoltaic users.The paper proposes to divide the feeder load considering photovoltaic users into traditional feeder load and photovoltaic load,and select the optimal artificial neural network algorithm respectively for short-term load prediction.Then the two vectors are superposed and combined to obtain the feeder load considering photovoltaic users with higher prediction accuracy.The main work is as below.Firstly,the feeder load considering photovoltaic users is divided into traditional feeder load part and photovoltaic load part by using the different directions of powerflow in the power distribution system.The load characteristics of the two parts are analyzedrespectively and the influencing factors of short-term load change are judged according to their periodicity and seasonal characteristics.Secondly,the correlation coefficient method is used to screen out the main influencing factors of traditional feeder load and photovoltaic load,and then they are used as input variables in the artificial neural network algorithm prediction model.According to the load characteristics of both,a feeder short-term load prediction model considering photovoltaic users is proposed,which is composed of the traditional feeder short-term load prediction module based on GRNN neural network and the photovoltaicshort-term load prediction module based on ELMAN neural network.Finally,taking a 10kV feeder considering photovoltaic user in a certain region of Guangdong province as an example,the load data and corresponding meteorological data were simulated on the MATLAB simulation platform.The results show that the short-term load prediction results obtained by dividing the traditional feeder load and photovoltaic load into two parts and then superimposing them are better than that obtained by direct short-term load prediction.The improvement of prediction accuracy proves the validity of decomposition and the validity of the prediction algorithm.