随机系统模型参数在线辨识的一种算法及负荷预报

本文提出了一种简化的多变量随机系统状态模型参数在线辨识方法。与最小二乘自适应递推算法比较,不仅需要辨识的参数减少,而且针对一类模型参数缓慢变化的系统,可以通过选择不同的遗忘因子序列来控制参数变化的幅度,解决了电力系统负荷预报中季节模型的老化问题。本方法基于带有随机噪声状态模型的典范型,大大节省了计算机的运算量和存贮容量,适于微处理机的在线应用。

A simplified on-line multivariable stochastic system state model parameters estimation algorithm is developed. We choose a particular canonical form for estimation purpose. Proposed algorithm is compared with extended least-squares algorithm. It reduces equation parameters, memory requirements and execution per iteration, and also suits for slowly changing Systems. In the application to load forecasting in electric system, the seasonal model is updated by properly selecting the forgetting factors, thus a long-term continuous load forecasting is achieved.