Short Term Load Forecasting Of Smart Grid Based On Improved LSSVM

The development of smart grid promotes the optimal allocation of energy structure and the diversified development of power system.Wind energy,photovoltaic and other clean energy rely on the smart grid to realize the grid and consumption,and emerging power demand such as electric vehicles can be realized through the smart grid.Compared with the traditional thermal power grid,the generation side of smart grid has more intermittent energy,and the energy consumption mode of demand side is becoming more and more diversified.In order to balance the supply and demand of smart grid,Pumped storage,hydrogen storage and chemical energy storage are widely used in load regulation of smart grid.Therefore,the short-term load of smart grid shows more complex variation rules and characteristics.Accurate prediction of short-term load of smart grid is the basis of power grid dispatching,which is of great significance to ensure the safe and stable operation of smart grid and provide reliable power load.In view of the load characteristics,influencing factors and model construction will affect the load forecasting effect of smart grid,this paper first uses the complete ensemble empirical mode decomposition(CEEMDAN)and singular value decomposition(SVD)to analyze the load series of smart grid.At the same time,from the generation side,the grid itself and the demand side of the smart grid,the improved grey relational analysis(IGRA)is used to calculate the correlation between the proportion of clean power,energy storage capacity and the number of electric vehicles and the smart grid load.At the same time,combined with the traditional power load factors,the smart grid load influencing factor system is constructed.The adaptive strategy is introduced to improve the colony foraging algorithm,and the adaptive bacterial foraging algorithm(ABFO)is proposed.The penalty parameters and kernel parameters of LSSVM are optimized by using ABFO.CEEMDAN-SVD-ABFO-LSSVM smart grid load forecasting model is constructed.Three data sets of Beijing and Hebei South power grid are selected to verify the proposed load forecasting model of smart grid,and the forecasting effect is tested from the aspects of forecasting accuracy and stability.The results show that the proposed load forecasting model of smart grid achieves high forecasting accuracy,and has strong stability and good generalization.With the development of clean energy and in-depth study of UHV technology,load forecasting has important application value and practical significance in smart grid dispatching optimization and power supply guarantee.Therefore,this paper establishes a more comprehensive system of smart grid load influencing factors from the generation side,the grid itself and the power side.CEEMDAN-SVD complex decomposition technology is used to realize the short-term load stabilization and characteristic analysis of smart grid.The ABFO-LSSVM optimization model is used to realize the accurate load forecasting of smart grid,which provides references for the development and stable operation of smart grid.