短期负荷预测最大李亚普诺夫指数法的改进

现有采用最大李亚普诺夫指数法进行负荷预测的3种改进技术为:应用待预测点对应的时间窗口上的最大李亚普诺夫指数估计(变化的数值),可以获得更高的准确率;应用"气温-负荷"相关系数等,改进"取舍规则":应用多个"邻近矢量"预测增加抗噪声能力.以此构造的2类改进预测方法是:采用"变最大李亚普诺夫指数、改进的取舍规则以及多邻近矢量"预测法;采用"多邻近矢量对应的1步负荷加权"预测法.后者还可用于非混沌序列预测.数值计算还表明,将原始负荷按照素数间隔抽样,可以进一步提高预测准确率.

IMPROVEMENT OF MAXIMAL LYAPUNOV EXPONENT METHOD TO SHORT TERM LOAD FORECASTING

Three technical improvements for the existing load forecasting methods based on the maximal Lyapunov exponent are presented. The first is use of the maximal Lyapunov exponent estimation in the time window corresponding to load at the time point to be forecasted; the second is to improve the rule of acceptance and rejection by use of the correlation coefficient between atmospheric temperature and load; the third is adopting multi-neighboring vectors in the forecasting to enhance the ability of anti-noise. On the basis of above-mentioned improvements, two kinds of forecasting methods are as following: (A) the forecasting method using varying maximal Lyapunov exponent, improved rule of acceptance and rejection and multi-neighboring vectors; (B) the forecasting method of weighted value of one step load record corresponding to multi-neighboring vectors. The latter can also be used to the load forecasting of non-chaotic series. Numerical calculation results also show that when the original load is sampled by the intervals of prime numbers, the accuracy of load forecasting can be further improved.