Research On Power System Load Forecasting Based On Deep Learning

Power load forecasting plays an increasingly important role in the safe and stable economic operation of modern power systems.Accurate power load forecasting can provide effective guidance for the development and planning of a regional power system to ensure reliable operation of the power grid,cost reduction,and safety and stability.Improving the accuracy of short-term power load forecasts can help alleviate the imbalance between supply and demand,significantly improve the utilization rate of power equipment,and reduce energy consumption.The accuracy of power load forecasting is restricted by many influencing factors.Traditional load forecasting models are difficult to adapt to the characteristics of the current load.It is particularly important to improve the accuracy of short-term load forecasting.Therefore,this paper mainly uses deep learning methods,combined with signal decomposition and feature extraction techniques,to study short-term power load forecasting.in view of the strong nonlinear,non-stationary and temporal characteristics of the load,an empirical modal decomposition(EMD)is proposed.,convolutional neural network(CNN)and long-term and short-term memory network(LSTM)hybrid model short-term power load forecasting methods,combining massive past load data,temperature and The historical electricity price information uses a sliding window method to construct a series feature vector as input.First,use EMD to reconstruct the data into multiple components,superimpose and combine the high,medium and low frequency components,and further use CNN to extract the hidden features of the high and medium components to reduce the weight.The number of values is input into the LSTM network in the form of feature vectors for load prediction,and finally the prediction results of each component are superimposed to obtain the final load prediction value.The experimental results show that this model has higher load forecasting accuracy than BP neural network(Back Propagation Neural Network),support vector machine(SVM),long-term and short-term memory network and EMD-LSTM models.Further considering the insufficiency of EMD for signal decomposition,the variational mode decomposition(VMD)with better performance is used to decompose the load sequence,and the parameter setting experience for variational mode decomposition(VMD)is carried out.Due to the strong subjectivity,a short-term power load prediction method based on the combination model of variational modal decomposition and gated recurrent unit(GRU)optimization parameters based on particle swarm algorithm(PSO)is proposed.The particle swarm algorithm is used to search for the optimal combination of parameters affecting VMD,and the decomposition sub-sequence with the best effect is obtained,which reduces the influence of different trend information on the prediction accuracy.Then,the GRU network is used to establish a GRU-based prediction model for each subsequence component.Finally,the prediction results of each sub-sequence are superimposed to obtain the final prediction value of the short-term power load.The experimental results show that compared with the Back Propagation Neural Network(Back Propagation Neural Network),the support vector machine(SVM),the GRU model,the EMD-GRU model and the unoptimized VMD-GRU model,this model has higher performance load forecasting accuracy.