| With the continuous development of social economy,the power system is gradually becoming a super large-scale complex system.Power system reliability assessment analyzes the safety level and weak links of the power system from the perspective of probabilistic risk,which can provide important decision-making reference information for power grid planning and operation.The random changes of the load is a typical random factor in the operation of the power system.Accurately identifying the internal change law of the load and establishing an effective load probability model is of great significance to improve the accuracy and practicability of the reliability evaluation of the power system.This paper conducts an in-depth study on the accurate construction of the load probability model and applies it to the reliability assessment of large power grids.The main research contents are as follows:(1)The typical construction methods of load probability model,namely semiparametric estimation model(Gaussian mixture model)and the basic principles of non-parametric kernel density estimation model are introduced in detail,and their advantages and disadvantages are deeply analyzed and summarized: although the Gaussian mixture model is a sparse probability model,the number of Gaussian components is often selected based on subjective experience,and it is difficult to optimize and determine;although the number of kernel functions of a non-parametric kernel density estimation model depends on the number of objective samples,it is a typical nonsparse probability model.Aiming at the above-mentioned advantages and disadvantages,the paper explains in detail through related calculation examples,and further points out the significance of the sparsity probability model based on the data-driven method to determine the number of kernel functions.(2)In order to have both the advantages of Gaussian mixture model and nonparametric kernel density estimation,the paper proposes a probability density estimation method based on The Generalized Cross Entropy(GCE),and first applies it to univariate(such as system load)Probability density estimation.The GCE method is based on the cross-entropy assumption,takes the minimization of the generalized cross-entropy measure as the objective function,establishes an optimization problem,and solves it through the Lagrangian duality theory,and obtains the sparsity probability model.Subsequently,the model was applied to power system reliability evaluation.Taking RBTS,IEEE-RTS79 and IEEE-RTS96 test systems as examples,the results show the effectiveness and practicability of the model established by this method in power system reliability evaluation.(3)Considering that in the actual operation of the power grid,there are complex correlation changes between node loads,and it is of great significance to establish an accurate correlation probability model for node loads.In view of this,this article extends the aforementioned GCE method to the joint of multi-dimensional random variables Probability density estimation.Compared with the univariate probability density estimation,the multi-dimensional joint probability density estimation based on the GCE method not only needs to consider the solution of the bandwidth coefficient,but also the choice of the bandwidth matrix form.The optimal bandwidth coefficient is determined by the generalized moment constraint,and the appropriate bandwidth matrix form is selected.Taking the multi-dimensional node load as an example,a correlation probability model is established and its fitting accuracy is verified.At the same time,the probability density model is applied to the reliability evaluation of the power system.Taking the RBTS and IEEE-RTS79 test systems as examples,the results verify that the method proposed in this paper can establish an accurate correlation probability model for the multi-dimensional node load,and It can be effectively used in power system reliability assessment. |
PDF Full Text Request