Research On Short-term Load Forecasting Based On Deep Learning

In the power system,the load prediction data with high accuracy and stability can effectively help the staff to supervise the system operation state and provide important basis for the generation plan.With the advent of the era of big data,the amount of load related data is growing exponentially.How to find the relationship between these data and apply it to meet the increasingly high demand of load prediction has become a research hotspot of scholars.To improve the accuracy of short-term load prediction in power system,this paper improves the training process and network structure of traditional neural network.This paper proposes an improved neural network short-term load prediction model.This method firstly analyzes the load characteristics and adopts the random neural network as the prediction model.For the problems such as the non-optimal model after the training of the random neural network model is stopped and the insufficient data utilization of the hidden layer of the neural network,an improved method is proposed.To verify the superiority of the improved model,a simulation experiment was carried out on the historical load data of a county in Henan Province.Simulation results show that the improved method performs better in prediction accuracy,average absolute error,average absolute error percentage and root mean square error than the traditional neural network,limit learning and short-long term memory models.Therefore,it can be concluded that the improved method of stochastic neural network has academic significance and engineering value in short-term load forecasting.