基于改进最小二乘支持向量机的电力系统受扰轨迹在线预测

Power system perturbed trajectory online prediction method based on improved least squares support vector machine

分析了标准最小二乘支持向量机算法用于在线预测时存在的主要问题,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轨迹预测的要求,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。以电科院8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。

Deficiencies of applying the standard least squares support vector machine （LS- SVM） to perturbed trajectory online prediction were specified. Based on the theorem of inverting block matrix, the support vectors were calculated in recursive formula, and learning efficiency was enhanced. In order to satisfy the online trajectory prediction of multi machine system, polymerization technology of trajectories was used to reduce the computing burden. According to the simulation results of 8- machine system of China electric power research institute （CEPRI） and northwest power grid of China, the validity of proposed method was proved in re.spects of prediction precision and computing time.