基于密度聚类与ARIMA模型短期电力负荷预测

Short-term electric load forecasting based on DBSCAN-ARIMA method

为了提高短期电力负荷预测的准确度,提出了将密度聚类算法(DBSCAN)与自回归移动平均模型(ARIMA)相结合的方法,进行短期电力负荷预测。首先,对数据进行归一化、天气状况类别数据编码、缺失值填补等预处理;然后,利用DBSCAN对负荷均值进行聚类与剔除噪音点。ARIMA模型的参数根据差分后的时间序列及热力图确定;最后,重构分解后的曲线,并根据历史数据对未来短期负荷进行预测。实验结果表明,预测结果的误差在合理范围内。

To improve the accuracy of electric load forecasting,a method which combines DBSCAN(Density-Based Spatial Clustering of Applications with Noise)with ARIMA(Autoregressive Integrated Moving Average Model)are proposed to predict short-term electrical load.Firstly,data preprocessing is carried out such as normalization,weather condition category data encoding,missing data imputation etc.Secondly,outliers are eliminated and filled by DBSCAN and adjacent data points.Parameters of the Arima model are calculated on the basis of thermodynamic chart.Finally,curve of reconstruction and load forecasting are carried out based on historical data.Experimental results indicate that errors are within reasonable ranges.