基于智能算法的空冷火电机组负荷预测研究

Study on load forecasting for air cooling thermal power units based on intelligent algorithm

为及时准确地预测空冷机组的整机性能,引入了基于智能算法的大数据分析方法。针对某600 MW空冷火电机组全年的历史运行数据进行预处理与稳态工况筛选,分别建立了基于反向传播(BP)神经网络和随机森林算法的机组负荷预测模型。预测结果对比分析和模型敏感性分析表明,随机森林预测模型具有精度高、泛化能力强、训练时间短等优点。为优化随机森林模型,通过皮尔森相关系数筛选模型输入特征并根据机组功率进行分负荷工况建模,优化后的模型性能得到了进一步提升。

In order to detect the performance of an air-cooling power unit timely and accurately,a big data analysis method based on intelligent algorithm is introduced.Load forecasting models taking BP neural network and Random Forest algorithm are developed for a 600 MW air-cooling thermal power unit based on pretreatment and steady state screening of its historical data.Through analyzing the prediction results and the sensitivity of the models,the Random Forest prediction model is proven to be of high precision,strong generalization ability and short training period.To optimize the Random Forest model,input characteristic parameters are filtered by Pearson's correlation coefficient and the model is set up according to working conditions under different loads.The optimized model can make more accurate prediction.