摘要
通过对广义回归神经网络(GRNN)在预测方面的研究,结合变压器特征气体检测的实际情况,建立了一种基于GRNN的变压器油中特征气体发展趋势的预测模型,用于等时间间隔和非等时间间隔采样,预测未来任意时刻变压器油中特征气体值、产气速率以及产气速率超出限定值的时间点,在吉林省多台变压器上应用,证明该方法预测误差均在允许范围之内,可避免设备故障的发生,提升了电网的运行水平。
With the study on general regression neural network in prediction and combing with the practical result of transformer feature gas, a new feature gas development trend forecasting method for transformer oil based on general regression neural network has been proposed. This method is used for interval and non-interval sampling and forecasting the feature gas value, production rate at any time in the future, the time that gas production rate exceeds the limit value. This method has been verified by certain amount of practical examples of transformer oil feature gas development trend.
出处
《吉林电力》
2014年第6期11-14,共4页
Jilin Electric Power
