Research On Spatial Load Forecasting Method Based On CEEMD And AFPSO-LSSVM

Spatial load forecasting(SLF)is the prediction of the spatial and temporal distribution of power load within the range to be predicted.Accurate and reliable load forecasting results can better guide the construction of the power grid.In the past,only the prediction of the load size was far from the requirements of the current power grid planning,so SLF has been paid more and more attention by relevant departments.Firstly,this paper elaborates the advantages and disadvantages of various data pretreatment methods in SLF,summarizes the development process of SLF and domestic and foreign research results,classifies the prediction methods according to different standards,and briefly analyzes the principle of common prediction methods and its advantages and disadvantages.Secondly,the Geographic Information System(GIS)was established,and the layers of class? cells and class ? cells were established.At the same time,the power load is divided from multiple angles to determine the main factors that affect the change of power load and analyze its regularity in the space-time domain.Then,a method is proposed to determine the reasonable maximum value of cellular load by using the complementary ensemble empirical mode decomposition(CEEMD)and runs test technology,which effectively solves the problem of the random fluctuation in the measured cellular load data and reduces the accuracy of spatial load prediction.The method decomposes each class ? cellular load sequence by complementary ensemble empirical mode decomposition technique.Each class ? cell obtains a set of intrinsic mode functions(IMF)and uses runs test technique for each intrinsic mode function.The intrinsic mode function is tested for randomness,and the criterion for identifying the high-frequency function is established.The high-frequency intrinsic mode function that characterizes the random fluctuation of the cellular load is removed,and the remaining intrinsic mode functions that characterize the regularity and trend of the cellular load are reconstructed to obtain the main component,and the maximum value of the main component is taken as the reasonable maximum value of the class ? cellular load.Finally,the reasonable maximum value is combined with the traditional load forecasting method to perform spatial load forecasting based on class ? cells and class ? cells.The engineering example shows that the method is correct and effective.Finally,an adaptive fuzzy particle swarm optimization(AFPSO)method combined with least squares support vector machine(LSSVM)for spatial load prediction is proposed.which effectively solves the problem of low prediction accuracy when using traditional load forecasting methods.This method uses the AFPSO algorithm to optimize the parameters in the LSSVM,which better overcomes the shortcomings of the standard particle swarm optimization algorithm,such as premature particles and low search accuracy,thus establishing a prediction model and applying it to the SLF to improve prediction accuracy.Engineering examples have proved the practicability and accuracy of this method.