LSSVM增量式训练的稀疏化算法在短期负荷预测中的应用

Short-term Load Forecasting Approach Based on Imposing Sparseness upon LSSVM Incremental Training Algorithm

最小二乘支持向量机与传统支持向量机相比在训练速度上有所提高,但当训练样本数目较大时,训练速度也相对缓慢。针对这一特点,对最小二乘支持向量机用增量式训练方法,使训练速度得到进一步提高,但是与传统支持向量机相比,丧失了解的稀疏性,影响了二次学习的效率。因此通过改进的剪枝算法对解进行了稀疏化处理,将此方法应用在电力系统短期负荷预测中,并对其预测结果与支持向量机进行分析比较,预测的准确性得到了进一步提高。

Compared with the support vector machine （SVM）, the least squares support vector machine （LSSVM） can improve the computing speed. However, when the sample size is large, the training speed is also relatively slow. Considering this, LSSVM incremental training algorithm is applied to increase the training speed. Compared with SVM, it losses the sparseness of solution, which influences the efficiency of second learning. An improved pruning algorithm is introduced to impose sparseness upon the solution of LSSVM incremental learning algorithm. Through comparison and analysis of load forecasting results between this algorithm and SVM, it is shown that the accuracy of load prediction is improved.