Research On Reliability Of Power System Based On Bi-level Cross Entropy And Multivariable Kernel Density Estimation

With the gradual development of society,people’s awareness of environmental protection and energy conservation has gradually increased,and the proportion of clean energy such as wind energy has gradually increased in power system.However,the penetration of wind energy has also brought additional uncertainty to the power system.In addition,there are also a lot of uncertainties on the load side of the power system,which makes the system modeling more complicated and more time-consuming to evaluate.Therefore,how to accurately and efficiently simulate the changing laws of these uncertain factors and establish an effective probability model is of great significance.In this paper,the power system considering the randomness and correlation of the wind speed and node load is the object of our research,and how to improve the efficiency of power system reliability evaluation is the focus of this paper.As one of the uncertain factors in the power system,the random fluctuations of load will definitely bring trouble to the reliability of the system.In power system reliability evaluation,if the discrete multidimensional load samples are sampled directly,the sampling efficiency will be very low because the probability of each sample being drawn is the same.Therefore,importance sampling based on cross entropy method is used to optimize the probability of samples being drawn,so that the samples that have a greater impact on reliability assessment are highlighted.Specifically,this paper proposes a single-level cross entropy method based on multistate discrete variables(SCE-MDV),this method sorts the original multi-dimensional discrete samples that have approximate effect on the reliability evaluation results into one class.On this basis,according to the different contributions of the samples in each class to the reliability evaluation results,the probability of the class is optimized,which finally achieves the purpose of optimizing the probability of every original discrete sample.In addition,considering that the multivariate kernel density estimation method is accurate in modeling but has low sampling efficiency,so according to its sampling characteristics,this paper will combine the single-level cross entropy method with multivariate kernel density estimation(MKDE),optimizing its two parameters,weight and bandwidth,to solve the problem of low sampling efficiency.Reliability analysis was conducted through RBTS,RTS79 and RTS79 system with derated load to verify the accuracy and effectiveness of the single-level cross entropy method.For the problem when there are too many classifications during single-level cross-entropy algorithm,some samples that have a larger impact on reliability indicators cannot be drawn,resulting in inaccurate reliability evaluation results,this paper proposes a bi-level cross entropy algorithm.Based on the single-level cross entropy method in Chapter 3,this method divides the class in the first level into several subclasses,and regards the probability of the class in the first level and the subclass in the second level as the parameters to be optimized.According to the different contributions of the samples in different classes or subclasses to the reliability evaluation results,while optimizing the probability of each class of the first level,the probability of each subclass of the second level is also optimized,so as to finally achieve the purpose of optimizing the probability of each original discrete sample,All in all,the bi-level cross entropy method can effectively avoid the shortcomings of the bi-level cross entropy algorithm.Similarly,the bi-level cross entropy method is combined with MKDE to optimize the two important parameters,weight of each kernel and bandwidth,to achieve the purpose of improving the speed of reliability assessment,while ensuring the accuracy of related modeling.Finally,reliability analysis was conducted through RBTS,RTS79 and RTS79 with multiple wind farms to verify the accuracy and effectiveness of the bi-level cross entropy method.The results show that this bi-level method improves the evaluation efficiency more and has a wider application range.