Research On Power Load Forecasting And Optimization Algorithms

Accurate short-term power load forecasting has an important impact on the production scheduling and planning of power systems.Power load forecasting has been a popular research direction since the reform and opening up,and how to construct power load forecasting models and improve the prediction accuracy of power load forecasting models is the most important research in the field of power load forecasting.Nowadays,with the rapid development of artificial intelligence research,intelligent forecasting models represented by Machine Learning(ML)algorithms and Artificial Neural Networks(ANNs)are widely used in power load forecasting modelling problems.The powerful non-linear modelling and learning capabilities of intelligent models are used to better deal with the power load forecasting problem.In addition,one of the research focuses on improving the prediction accuracy of power load forecasting models by using intelligent optimization algorithms to optimize the relevant parameters of intelligent forecasting models.In this respect,although a single power load forecasting model or a single power load forecasting model optimized by standard optimization algorithms can achieve power load forecasts with a certain level of forecasting accuracy,their forecasting results often fail to meet the standards and requirements of modern forecasting accuracy.The main design idea of this thesis is to build an power load forecasting model using the idea of combined forecasting models,using the advantages of multiple techniques to compensate for the disadvantages or shortcomings of another technique.Secondly,a suitable high-performance intelligent forecasting model is selected to build the power load forecasting model.Finally,a multi-strategy improved Sparrow Search Algorithm(SSA)is used to optimize the parameters of the intelligent forecasting model to achieve better forecasting performance and accuracy of the power load forecasting model.The full text is developed in the following sections:(1)A novel FA-CSSA-ELM based model for combined electricity load forecasting;In this thesis,a novel FA-CSSA-ELM-based combined power load forecasting model is proposed: firstly,the sparrow population is initiali zed by a Tent chaos mapping strategy,and the Tent chaos property is used to make the initial sparrow population uniformly distributed in the solution space.Secondly,the Firefly algorithm(FA)is used to update the positions of the optimal sparrow and sparrow flock by using the principle that fireflies with high brightness in the search space can attract fireflies with low brightness to draw closer,so as to improve the search ability and algorithm accuracy of the sparrow search algorithm.The two strategies are combined to construct the FA-CSSA algorithm.Finally,the FA-CSSA algorithm is used to optimize the initial weights and thresholds of the Extreme Learning Machine(ELM)to construct the final combined power load forecasting model.(2)A new combined VMD-CISSA-LSSVM-based model for forecasting electrical loads;In this thesis,a new combined power load forecasting model based on VMD-CISSA-LSSVM is further investigated on the basis of the FA-CSSA-ELM model.The model combines the Variational Modal Decomposition(VMD)data pre-processing technique,the CISSA algorithm based on the improved Tent chaos strategy,the random following strategy and the Levy flight variation strategy optimization,and the Least Square Support Vector Machine(LSSVM)model.Firstly,the raw data is noise-reduced using VMD data pre-processing techniques to minimize the impact of noise on prediction performance.Secondly,in the initialization phase,an improved Tent chaotic mapping is used to generate the initial population to enhance the quality and population diversity of the initial sparrow individuals;in the location update phase,a random-following strategy is used to optimize the location update formula of the joiners in the sparrow search algorithm,balancing the local optimization performance and global search capability of the algorithm;in the late iteration of the algorithm,a Levy flight strategy is used to expand the search range of the population,thus improving the local and global search capability.Finally,the CISSA algorithm is used to optimize the relevant parameters of the LSSVM model to construct the final electricity load forecasting model.(3)A new multi-model based model for forecasting combinations of electrical loads;This thesis presents a novel multi-model based model for combined power load forecasting.The model utilizes a data pre-processing technique,the VMD-Singular Spectral Analysis noise reduction method;two novel improved intelligent optimization algorithms,the Chaotic Adaptive Whale Algorithm(CAWOA)and the Elite Backward Learning-based Chaotic Sparrow Search Algorithm(EOBL-CSSA);and three independent and efficient forecasting models,the Elman Neural Network,the ELM Neural Network and the LSSVM model.Firstly,the VMDsingular spectrum analysis noise reduction method is used to reduce the noise impact of the original data as much as possible by reducing the noise twice.Secondl y,the SSA algorithm,CAWOA algorithm and EOBL-CSSA algorithm are used to optimize the relevant parameters of the Elman neural network,ELM neural network and LSSVM algorithm models respectively to construct the prediction models.Finally,Simulated Annealing(SA)is used to calculate the weighting ratio of the prediction results of the above three independent combination prediction models,and weighting is performed to obtain the final prediction results.