The Application Of Non-parametric Kernel Density Estimation Load Model In Power System Reliability Evaluation

This dissertation is supported by National Natural Science Foundation of China (50977094, 50607021).With the scale of power grid expanding increasingly and the degree of power network interconnected raising continuously, the application of quantitative reliability evaluation in electric power systems has now evolved to the point at which most utilities use these techniques in one or more areas of their planning, design, and operation. In recent years, power system reliability in model algorithm or engineering applications has made great progress. As a typical random phenomenon, the accuracy of load model has a great impact on assessment of reliability. How to combine organically short-term load forecasting with power system operation risk assessment, how to consider the nodes load uncertainty and correlation effect on the reliability of assessment results, become in-depth study of important issues.This paper considers the temporal and spatial correlation between the loads and adopts "non-parametric kernel density estimation" method, which is emerging technologies of mathematical statistics, on the joint probability density distribution of timing and nodes loads.The method based entirely on data-driven, which can effectively avoid the subjective experiences.The temporal correlation between the system timing loads was taken into account. And a short-term load forecasting model using non-parametric kernel density estimate was established, as the historical load data as nonlinear time series: focusing on the model order, smoothing parameters and prediction confidence intervals. The model avoids selecting the factors artificially, without a lot of the factors information as a data support.By improving the algorithm, the expected value of load probability density estimation and conditional probability density distribution can be obtained. The effectiveness and availability of the model are verified on two case studies: a real power system and the IEEE-RTS79 test system. Predicted results not only be a more objective reflect the system load trend, but also provide the load information for the bulk power system reliability evaluation to get the system ahead risk level.In addition, using non-parametric kernel estimation in the bulk power system reliability evaluation achieved load correlation, the joint probability density of bus loads have obtained. The presented method is a kind of data-driven approach without any assumption for underlying form, and is capable of uncovering the structure information hidden in the historical load data. Improved through the IEEE-RBTS reliability test system, calculated the node load correlation and uncertainties related to the reliability of the grid, and verified the correctness of the model.