基于标准化高斯pLSA协同过滤的用电量预测模型

Load-forecasting model based on normalized Gaussian pLSA collaborative filtering

现有的电力负荷预测算法在中长期预测时存在不同程度的局限性.究其原因,是因为影响复杂非线性系统输出的变元过多,难以用解析的方法对其进行描述.本文提出利用概率潜在语义分析使历史随机数据呈现出各种有规律的示象(aspect),结合对内容的协同过滤技术去建立用电量预测模型,从而利用统计学习的方法避开了对影响系统输出的隐含变元的寻找与刻画.采用MATLAB进行数值仿真实验的结果表明该算法相比于神经网络和灰色预测在准确度方面具有优势.

To some extent the existing long-term load-forecasting algorithms have their limitations because the variables influencing the output of the complex non-finear system are too many to be described. By combining the probabilistic Latent Semantic Analysis （pLSA） that can cluster random data into respective aspects and content-based collaborative filtering, a novel load forecasting model based on normalized Gaussian probabilistic latent semantic analysis collaborative filtering is proposed in order to avoid seeking and describing of the hidden variables mentioned above. Simulating experiments via MATLAB show that this method gains the advantage in accuracy over neural network and grey prediction.