Research On Unit Commitment Optimization Modeling Based On MIP And LS-SVM Method

In recent years,China’s power generation is in noticeably robust demand and the demand for energy supply is increasing,which has greatly increased the scheduling pressure.The unit commitment(UC)problem is a very important optimization problem.In the research of daily operation scheduling and planning,whether in the short term or in the long term,it has been highly concerned by the industry and academia.It has carried out long-term studies and extensive applications both at home and abroad,and has achieved great success.Complex operating constraints and non-convexity,combined with its large-scale characteristics.It makes solving the UC problem challenging to some extent.As a traditional and important problem in the power system,the UC problem has received extensive attention for decades.Many literatures have reported methods to solve this problem,but most of them are approximate methods.This study presents an improved unit commitment-mixed integer programming(UC-MIP)model based on variable upper bound,which is simultaneously tight and compact.The proposed models were tested on 73 instances over a scheduling period of 24 h.In addition,the numeric experiments show dramatic improvements in computational time for this paper proposed model.We provide evidence that the proposed model has better performance than the previous models.Compared to the 2-bin formulation,the proposed model reduced by at least 6.6%,even 42.1% in the average time of calculation.In addition,the two tighter and relatively compact multi-period formulations are also presented.Both formulations are tighter than the previous two binary(2-bin)variables and the tighter characteristic largely reduces the computational time of the formulations.They are more suitable for UC problems with long scheduling time.The proposed models are tested in the 48-hour scheduling problem,and the experiment shows that the calculation time of the proposed models are significantly reduced in the case of long time scheduling.According to the proposed UC models,this paper deeply analyses the decision problem of UC models,using wavelet packet energy entropy and least square support vector machine(LS-SVM)algorithms.The wavelet packet energy entropy is used to extract the characteristics of the unit data,and then according to the given unit data used the LS-SVM to classify and selected the optimal model.It can improve the efficiency of the solution while obtaining high-quality solutions.For large-scale unit data,the most suitable unit commitment model can be selected for calculation in just a few seconds,greatly reducing the calculation time.This research is composite the mixed integer programming unit models,wavelet packet energy entropy and machine learning algorithm to find the efficient and optimal solution of the UC problem in power system.