基于在线最小二乘支持向量机回归的电力系统暂态稳定预测

提出了一种基于在线最小二乘支持向量机回归的电力系统暂态稳定预测方法。分析了标准最小二乘支持向量机回归算法用于在线预测时存在的主要问题，然后根据分块矩阵求逆定理对标准算法进行改进，实现支持向量的递推式求解，提高了算法的学习效率。为了满足实际多机系统在线稳定预测的要求，引入轨迹聚合技术对多机轨迹进行聚合，进一步减少了计算量。在轨迹降阶的基础上，根据EEAC理论，通过识别聚合轨迹的动态鞍点来判断轨迹的稳定性。最后，以电科院7机系统和我国西北电网为例进行仿真分析，从预测精度和计算时间两方面验证了方法的有效性。

ELECTRIC POWER SYSTEM TRANSIENT STABILITY PREDICTION BASED ON ON-LINE LEAST SQUARES SUPPORT VECTOR MACHINE

In prediction of the power system transient trajectory with least squares support vector machine (LS-SVM), repeatedly inverting matrix is the most important factor in calculation speed. According to the theorem of inverting block matrix, this paper modified the algorithm to save calculation time. In order to satisfy the requirement for on-line transient stability prediction of multi-machine system, the trajectories polymerization technology was used to reduce computing complexity. Based on extended equal area criterion (EEAC), transient stability was estimated by identifying the dynamic saddle point (DSP). According to the simulation results of 36-node system suggested by China Electric Power Research Institute (CEPRI) and China northwest power grid, the validity of proposed method was proved in respects of prediction precision and computing efficiency.