Short-term Power Load Probability Density Forecasting Method Based On Support Vector Quantile Regression And Smart Grid

Power load forecasting is the important basis of electric power system planning and scheduling.It is able to ensure economic safety and stable operation of power systems.The sustainable development of environment is the important foundation of the survival and development of human society.The smart grid is the inevitable trend of the future power grid development.And the realization of the smart grid is inseparable from the support of accurate power load forecasting method.Penetration of distributed power and renewable energy sources effectively reduces the traditional energy consumption and protects the environment.However,due to the characteristics such as instability,intermittent,it brings new problems to the stability of the operation of power grids.According to the complexity and uncertainty of operation and dispatch of the power system for smart grid,we need to put forward a new method to improve the accuracy of power load forecasting.In the rapid development environment of smart grid,there are other factors which influence the accuracy of the short-term power load forecasting.Real-time price is also very significant to affect the accuracy of short-term power load forecasting except the historical load and meteorological factors.Moreover,it is one of the uncertain factors that affect the power load forecasting,which has remarkable influence on the power consumption pattern of power consumers.With the rapid development of smart grid,one of the significant changes is that people can adjust the power consumption pattern according to the requirement of electricity and real-time price.So that we can achieve the purpose of peak shaving and load shifting,and improve the efficiency of the power grid equipment and energy efficiency.At the same time,we reduce the electricity expenses of power consumers.In order to improve the accuracy of the short-term power load forecasting considering real-time price,and reflect the uncertainty of power load preferably,this paper proposes a support vector quantile regression(SVQR)method.By means of introducing slack variables to construct the Lagrange function,the results of power load forecasting under different quantiles at any time in a day are evaluated.Meantime,SVQR is combined with kernel density estimation by adopting the Epanechnikov kernel function to perform short-term power load probability density forecasting,which can obtain better prediction results and accurate range at any time future of power load.The paper adopts SVQR method to proceed power load forecasting.It is very vital for this model to consider the kernel function.Thus,a new method is proposed,that is short-term power load probability density forecasting method using kernel-based support vector quantile regression and Copula theory.Three different kernel functions are compared in this work to select the best one for the learning target.Copula theory is proposed in order to analyze the relation between electrical load and real-time price.Besides,the paper evaluates the accuracy of the prediction intervals considering two criteria,namely,prediction interval coverage probability(PICP)and prediction interval normalized average width(PINAW).The historical load and real time price datasets of Singapore's four cases are adopted to proceed the problem of short-term power load probability density forecasting.The results show that the method provides the relevant relationship between real-time price and power load,which has the ability to solve short-term power load probability density prediction problem with considering real-time electricity price.At the same time,the paper gives the relational diagram of real-time price and power load.And diagram of probability density curve and prediction results and prediction intervals,which better illustrate the superiority of the proposed method.