Research On Neural Network Load Forecasting Algorithm Considering Environmental Factors

Power load forecasting is a very important research field,and it is also the focus of this paper.Power load forecasting is to pr edict the load in the next few hours to days,months,or even years.Short-term load forecasting can be used for the power system.Safe operation provides guarantee and also plays an important role in power system scheduling.Medium and long-term load forecasting also plays an extremely important role in the future power grid construction planning and fuel demand.In this paper,the basic theory of load forecasting is introduced in detail,and various load characteristics and related factors affecting power load are deeply analyzed,and the algorithm model from original load data preprocessing to load forecasting is established.First,the local outlier factor detection algorithm is used to check and verify the original load data to find out the abnormal val ues and blank values in the original data,and then use the moving average method to fill in the abnormal data points in the data set.Relevant factors affecting power load factors,discarding external environmental data such as rainfall and wind speed with low correlation,and selecting mu ltiple features such as temperature,humidity,and holiday data with high correlation as input,reducing the input data dimension and improving the prediction accuracy and speed of the model.For the processed load data,the frequency domain decomposition load data is proposed.The fluctuation of the load determines the complexity of the load value.The frequency domain decomposition is enough to effectively extract the data characteristics of different stages.For the loa d data with different data characteristics,it has The huge advantage greatly improves the load forecasting effect and forecasting accuracy.Based on genetic algorithm theory,combined with the performance advantages of long and short-term neural memory network for time series data mining and forecasting,a GA-LSTM load forecasting model is proposed.Taking advantage of the problems existing in the LSTM load forecasting model,such as numerous parameters and the dependence of parameter adjustment on manual experience,the parameter search space of the LSTM load forecasting model is defined in combination with the characteristics of the load data,and a parameter search strategy based on genetic algorithm is proposed,taking the load forecasting accuracy as t he optimization goal,the FDD-GALSTM model is constructed.