Study On Short-Term Load Forecasting Of Power System Under Complex Environment

With the large-scale integration of energy storage,electric vehicles,renewable energy and the increasing diversification of society's demand for power supply services,the quantity,variety,randomness and distribution characteristics of load influencing factors have become increasingly complex and complicated.Research on short-term load forecasting applicable to the complex environment is urgent.In this paper,the composition and influencing factors of load were analyzed in detail,and short-term load forecasting models that can effectively deal with the complex and nonlinear relationship between load factors and load factors were established.The main work done is as follows:(1)The composition of power system load,short-term load influencing factors and input variables selection of load forecasting model were analyzed in detail,and the abnormal data were detected and corrected based on multiple linear regression model,which lays the foundation for the subsequent load forecasting modeling.(2)A short-term load forecasting method based on deep belief network was proposed.In the process of model parameter pre-training,the influence factors such as weather,date type and time-of-use price were taken as the input data of the prediction model,and the Gaussian-Bernoulli Restricted Boltzmann Machine(GB-RBM)was used as the first module in the stack to compose the deep belief network,so as to deal with the multiple types of real-valued input data that have an impact on the load more effectively.The partially supervised training algorithm,which is a combination of unsupervised training algorithm and supervised training algorithm,was used to pretrain the complex non-linear relationship between load factors and load to be predicted.Complex reality was better approximated by this way.The Levenberg-Marquardt(LM)optimization algorithm was used to fine-tune the network parameters obtained in the pre-training stage and converge to the optimal solution more quickly,which avoids the disadvantage that the gradient descent method converges slowly and easily falls into local optimum.The results show that the proposed method has higher prediction accuracy when the training samples are larger and the load influencing factors are complex.(3)A short-term load forecasting method based on EMD-mRMR-PSO-LSSVM was proposed.Firstly,EMD was used to decompose the original load sequence into several sub-sequences,and then the optimal input variables set of each load sequence were selected according to the mRMR criterion.Then,each load sequence was predicted by the load forecasting model based on least squares support vector machine(LSSVM)whose parameters were optimized by the particle swarm optimization algorithm.The results show that this method can predict the short-term load sensitive to external factors more accurately than other methods.