Short Term Load Forecasting Model Of Power System Considering Real-time Meteorological Factors

Short term load forecasting mainly refers to forecasting the power load in the short-term period in the future.It plays a key role in making safe,reliable and economic operation strategies for the power system.With the economic growth,the base number of temperature regulating load is also rising,and the difference between peak and valley of daily load is widening,which makes the load forecasting work of power grid face new challenges.At the same time,it puts forward higher requirements for the accuracy of forecasting than before.However,the daily meteorological factors are different at every moment,which has a "real-time" impact on the load.In the short-term load forecasting work,we should make full use of the advantages of big data to deeply mine the potential laws of load and meteorological information,study the real-time influence of meteorological factors on load curve,and establish a forecasting model that can reflect the law of short-term load change,which is of great significance to improve the level of forecasting accuracy and promote the refined management of load forecasting work.By collecting data,analyzing the load characteristics of power grid and building short-term load forecasting model,this thesisproposes a short-term load forecasting method based on temporal convolutional networks and a short-term load forecasting method based on temporal convolutional networks and least squares support vector machine combination model.The main contents of this thesis are as follows:(1)From the time scales of year,week and day,the periodic rule of load characteristics contained in grid load is mined.Then,based on this,the data of load and meteorological data are processed,and the correlation between meteorological factors and load is analyzed in detail from the perspective of daily characteristics and real-time characteristics.Based on the theory of weighted grey correlation degree,a method of similar day selection is proposed,which is based on the mixed daily features and real-time meteorological features.(2)Considering the characteristics of "real-time" and "time difference" of the influence of meteorology on load change,the real-time meteorological factors including real-time comprehensive meteorological index are introduced as part of the model input,and the temporal convolutional networks which can fully consider and contain the "time difference" between real-time meteorological factors and load is used for short-term load forecasting modeling of power grid,that is,based on A short-term load forecasting modeling method based on temporal convolutional networks is proposed.The simulation results show that the model can effectively improve the accuracy of short-term load forecasting.(3)In order to further improve the prediction accuracy,this thesis proposes a combined prediction model of TCN and IPSO-LSSVM based on the advantage matrix method,and verifies that the combined prediction model has a better prediction effect than the original single prediction model through simulation.