Research On Short-term Load Forecasting Model Based On Uncertain Load

The load forecasting of power system can help the related departments to arrange the production and planning of electric energy more reasonably,and it is also an important research topic of Power System Operation Management.Nowadays,electric energy permeates every aspect of our life and affects the whole society and national economy.In order to realize the goal of power system security and stability,it is of great significance to ensure accurate load forecasting.However,when more uncertain loads,such as electric vehicle charging posts and energy storage,are connected to the grid,it increases the complexity of the power system and makes load forecasting more difficult.With the development of electric power industry,it is difficult for traditional load forecasting methods to maintain high forecasting accuracy.Nowadays,a large number of scholars have put their research on power system load forecasting or uncertain load such as electric vehicle charging post,thus a large number of new load forecasting algorithms,such as neural network and deep learning,have emerged,time series method and so on.These algorithms can extract the hidden features in the learning history data,and then perform better in the field of load forecasting.In this paper,the research background and significance of short-term load forecasting are described,and the research status of short-term load forecasting is also described.After analyzing the uncertainty and basic change characteristics of power system load,this paper chooses the Prophet prediction model,which performs very well in analyzing the learning time series data,holiday items and other parts,but this model is easy to over-fitting or under-fitting in the part of each addition item,and then affect the final prediction results,so this paper adopts the combination model,by combining the weights of the two models,the latter not only inherits the advantages of the latter,but also embodies the ability of the latter in analyzing and learning the hidden features of the long-term historical data,finally,a numerical example is given to demonstrate the effectiveness of the combined model.Electric Vehicles(evs)and other uncertain loads are connected to the power grid,which has an important influence on the power grid.Different from the overall load of electric power system,the load fluctuation of electric vehicle charging pile is more uncertain and irregular,a load forecasting method based on wavelet transform and the composite model of Prophet-LSTM for electric vehicle(EV)charging post is proposed.Wavelet transform can decompose the volatile data into a number of components containing a large number of eigenvalues and a number of stationary components which have regular changing trends,so that the original signal becomes more deterministic and less volatile,then the decomposed components are predicted by using the LSTM prediction model,and then the predicted results are reconstructed to get the predicted results from the LSTM prediction model,and the original data are input to the Prophet prediction model to get the predicted results,at this point,the predictions of the LSTM and Prophet models are combined by least squares weight.The experimental results show that the improved model based on wavelet decomposition has higher prediction accuracy and better practicability than the traditional model.