基于数字孪生的电网变压器故障诊断方法

为减少电网安全隐患,保障用户正常用电,提出基于数字孪生的电网变压器故障诊断方法。 利用传感器设备,结合全要素实体基本信息,模拟变压器运行状态,在虚拟空间内建立数字孪生体,包括几何、物理、规则和行为 4 个模型。 在该孪生体中,利用正交全局与局部保持嵌入算法将初始故障特征集合变换到高维核空间内,计算映射矩阵,获得映射向量最大特征值,提取低维敏感特征;结合提取的故障特征,在规则模型中采用概率神经网络算法,建立概率密度函数,使用差分进化方法,确立初始种群,经过反复的变异、交叉操作,丰富种群多样性,确定最佳平滑因子,当达到最佳迭代次数时,输出最佳诊断结果。 仿真实验表明,所提算法能够有效区分不同故障特征,减少诊断时间,可获得更加精确的诊断结果。

Fault Diagnosis Method of Power Grid Transformer Based on Digital Twin

In order to reduce the potential safety hazards of power grid and ensure the normal power consumption of users,a power grid transformer fault diagnosis method based on digital twin is proposed. Using sensor equipment,combined with the basic information of all element entities,simulating the operation state of transformer,the digital twin is established in the virtual space,including four models:geometry,physics,rules and behavior;in this twin,the initial fault feature set is transformed into a high-dimensional kernel space by using orthogonal globa and local preserving embedding algorithm,then the mapping matrix is calculated,the maximum eigenvalue of the mapping vector is obtained,and the low-dimensional sensitive features are extracted;combined with the extracted fault features,the probabilistic neural network algorithm is used in the rule model to establish the probability density function,and the differential evolution method is used to establish the initial population. After repeated mutation and crossover operations,the population diversity is enriched. The optimal smoothing factor is determined,and the optimal diagnosis result is output when the optimal number of iterations is reached. Simulation results show that the proposed algorithm can effectively distinguish different fault features,reduce diagnosis time and obtain more accurate diagnosis results.