A neural network model-based approach for power data collection and load forecasting accuracy improvement

Abstract Accurate power load forecasting is a prerequisite for opening up the field of power system generation and development, and its reliability is sufficient to eliminate the dilemmas caused by its inherent irregularity, randomness and non-stationarity, so as to realize the effective scheduling of the balance between power supply and demand, the attainment of energy saving and emission reduction, and the improvement of economic efficiency. In this paper, the categorization of power load forecasting is first elaborated to determine the length of the forecasting period in this paper. An example analysis of the power load sequence is carried out to summarize the characteristics of the power load sequence. The influencing factors, including date factors, meteorological factors and other factors, are investigated. Then the electricity data preprocessing method is described, LSTM neural network is used for modeling, and Particle Swarm Algorithm (PSO) is introduced to optimize the parameters in the LSTM model. Finally, the PSO-LSTM model is used to predict the power load data after optimizing the parameters of the particle swarm algorithm, and the final loss value of the PSO-LSTM model is at least 0.1962, which proves that the model optimized by the particle swarm algorithm has a higher prediction accuracy than that of the model before the optimization.