基于神经网络的变压器损耗计算方法

Calculation Method of Transformer Power Loss Based on Neural Networks

常规的变压器损耗计算只考虑了负载率的影响,却忽略了谐波以及三相不平衡带来的附加损耗,因此需要一种新的方法建立变压器损耗和其影响因素之间的映射关系。人工神经网络能够通过不断的学习来拟合负载率、谐波畸变率、三相不平衡度等特征参数和变压器损耗之间复杂的非线性映射,通过仿真,训练过的神经网络输出结果误差小、响应速度快,只需提供特征参数就能得出变压器损耗数据,与传统方法相比,该方法不仅考虑的因素更为全面,还减少了运算过程,在变压器损耗预测中也能起到积极的作用。

Conventional transformer loss calculation only considers the impact of the load rate of transformer, but ignored the additional losses caused by harmonics and unbalance. So it is necessary to find a new approach to es- tablish the mapping relationship between transformer losses and its influencing factors. Artificial neural networks can fit the complex nonlinear mapping between the load factor, harmonic distortion, imbalance, other characteristic parameters and transformer losses through continuous learning. With Matlab simulation, the error of trained neural network output is small and the speed of response is fast. Transformer loss can be obtained only by providing the characteristic parameters. Compared with the traditional method, this method not only considers more comprehen- sive, but also reduces the computation process. It also can play a role in the tansformer loss prediction.