Research On Ultra Short Term Load Forecasting Based On Neural Network

With the rapid development of China’s power industry,the construction of smart grid is constantly improving,which makes the modern power system develop towards intelligence,marketization and digitization.The structure of power grid is more and more complex and the operation mode is more and more digitalized,which brings great threats and challenges to the planning,operation and security of power system,which requires the power system to improve the speed and accuracy of prediction and shorten the prediction time The predicted time can provide an important theoretical basis for the safe and stable operation of power system and the optimization of dispatching plan.Therefore,it is an important topic to study the Ultra short term load forecasting with shorter forecasting time,higher accuracy and faster speed;and the long-term and short-term memory network is a new neural network method in recent years,so this paper studies the Ultra short term load forecasting based on the long-term and short-term memory network method combined with the actual data.The main contents include:1.Summarize the concept,characteristics and principles of Ultra short term load forecasting;determine the basic steps of Ultra short term load forecasting;select the evaluation index of forecasting error,analyze the causes of forecasting error;summarize the application of Ultra short term load forecasting.2.Research on single neural network method for Ultra short term load forecasting.Long short term memory,LSTM(hidden layer model)neural network model is the basic model,using the hidden layer number and the number of neurons in the hidden layer of the model to take different values,the simulation experiment is carried out on the actual data,and different prediction results are obtained and compared according to different parameters,and compared with the prediction results of BP and RNN neural network model,the prediction effect of long-term and short-term memory network is slightly better.3.Research on neural network method for Ultra short term load forecasting optimization.On the basis of LSTM model,attention is introduced to optimize the weight of LSTM neural network model,and the parameters of long-term and short-term memory network model are optimized by using the powerful global search ability of improved particle swarm optimization algorithm.The attention APSO LSTM model is established.Taking the actual load data as an example,the effectiveness of the above model is verified,which is consistent with the attention LSTM model,PSO-LSTM model and LSTM model Through comparative analysis,the optimized model reduces the prediction error and improves the prediction accuracy.4.Research on combined neural network method for Ultra short term load forecasting.ARIMA-LSTM combined model is proposed to apply to Ultra short term load forecasting of power system.ARIMA model is used to forecast the load,and the error between the original data and ARIMA forecast data is regarded as a nonlinear component;the error series is modified by LSTM neural network,and the final forecast value of Ultra short term load is obtained by adding ARIMA forecast value and LSTM correction value.Based on the actual load data of users,the comparison with ARIMA model,LSTM model and ARIMA-LSTM combination model shows that the proposed method can effectively control the prediction error and improve the accuracy of Ultra short term load forecasting.