Font Size: a A A

Short-Term Power Load Forecasting Based On VMD And Improved Extreme Learning Machine

Posted on:2020-02-01Degree:MasterType:Thesis
Country:ChinaCandidate:J H WuFull Text:PDF
GTID:2392330596979297Subject:Engineering
Abstract/Summary:PDF Full Text Request
As an efficient green energy,electric power plays an important role in today's society.Short-term load forecasting of power system has important reference significance for rational dispatch and stable operation of power sector,and is also the fundamental pocess tof optimal operation of power grid.The research o'f short-term load forecasting has lasted for many years,but with the increasing influence factors on power load,the change of power load becomes more complex.Traditional forecasting models can no longer meet the needs of forecasting accuracy.Therefore,selecting appropriate forecasting models to improve forecasting accuracy has certain research significance.Based on the power load of an area in Anhui Province,the load characteristics and short-term load forecasting model are studied in this paper.First,because the limitations of traditional forecasting models such as BP neural network and SVM support vector machine on short-term load forecasting,in this paper,a short-term load forecasting model of extreme learning machine is established.l1owever,the randomness of initial input weights and hidden neuron thresholds in extreme learning machines leads to unstable forecasting performance and even inadequate forecasting accuracy.The paper combined improved-particc swarm optimization and genetic algorithm to optimize the input weights and hidden layers thresholds of extreme learning machine to improveSecond,because the power load is affected by human activities and many external factors,the noise in load sequence signal is random and non-linear.In this paper,the variational mode decomposition technique is used to decompose the original power load sequence into sub-scqucnccs with limited bandwidth to extract detailed features and avoid mode aliasingharing.Then the sample entropy of the subsequence is calculated.The sample entropy reflects the complcxity of the sequence.The subsequences with similar sample entropy values are merged into a new sequence.which is then combined with the optimized forecasting model of the extreme learning machine.The forecasting results of each subsequence are superimposed to get the final forecasting results.Finally,the real power load data of a certain area are taken as sample data for simulation test.The simulation results show that the forecasting performance of the two improved models in different day load forecasting has been improved significantly,and the results are satisfactory in engineering practice.
Keywords/Search Tags:Short-term load forecasting, Extreme learning machine, Improved-particle swarm optimization, Variational mode decomposition, Sample entropy
PDF Full Text Request
Related items