绝缘子覆雪闪络特性及其改进量子神经网络的预测模型

Study on Snow Covered Insulator Flashover Characteristics and Its Improved QNN Prediction Model

准确评估绝缘子覆雪融雪过程中闪络电压特性,对恶劣气象条件下电网安全运行水平评估和线路检修策略决策具有重要意义。针对FXBW-35/70型复合绝缘子,采用人工覆雪闪络电压试验方法,对均匀覆雪和不均匀覆雪绝缘子串的覆雪期、融雪初期和融雪后期3个阶段的闪络电压特性进行了试验研究,分析了各阶段闪络电压和泄漏电流变化特征及其产生原因。通过分析覆雪绝缘子的电场分布,探究了融雪期闪络电压变化的原因。研究基于深度自编码网络与量子神经网络结合的覆雪闪络电压特性变化建模方法,建立基于改进量子神经网络的覆雪闪络电压预测模型。人工覆雪闪络电压试验测试和验证结果表明,提出的改进量子神经网络闪络电压预测模型与单一的反向传播神经网络、支持向量机模型和遗传算法模型相比,能够更加准确地反映北方地区线路绝缘子覆雪融雪过程中的闪络电压变化特性,其预测结果可为电网安全评估与检修决策提供有效指导。

To evaluate safe operation level of power grids and maintenance strategies of transmission line under severe meteorological conditions, it is significant to accurately assess flashover voltage characteristics during snow-melting process on insulators. In this paper, artificial snow-covered flashover voltage test method with FXBW-35/70 composite insulators is used to test the flashover voltage characteristics of snowcovering, initial snow-melting and later snow-melting stages of uniform snow-covered and non-uniform snow-covered insulator strings. Flashover voltage characteristics and leakage current variations at each stage and the causes are analyzed. By analyzing the electric field of snow-covered insulators, the cause of flashover voltage change during snow-melting period was preliminarily analyzed. Based on combination of deep self-encoding network and quantum neural network（QNN）, a snow-covered flashover voltage characteristic change modeling method is established. The snow-covered flashover voltage prediction model based on improved QNN is set up. Artificial flashover voltage test result shows that the proposed improved flashover voltage forecasting model of quantum neural networks can more accurately reflect the change of flashover voltage during snow-melting process on line insulators in Northern regions than a single back-propagation neural network, support vector machine model and genetic algorithm model. Forecast results from the proposed method can provide effective guidance for grid security assessment and maintenance decisions.