基于数据驱动的高炉煤气复合预测模型

对钢铁企业高炉煤气系统科学准确的预测，可以为煤气的合理调度提供依据，对企业提高能源利用效率、减少煤气放散和环境污染有着非常重要的意义。针对钢铁企业高炉煤气系统设备工况复杂、煤气量波动频繁、难以准确预测的问题，依据小波分析方法、BP神经网络、最小二乘支持向量机的性质建立了基于数据驱动的高炉煤气的复合预测模型。该模型综合考虑高炉煤气系统生产计划和检修计划，对高炉煤气系统的产耗用户在不同工况下分别建立训练数据集，利用多组模型参数预测高炉煤气产生量、消耗量和缓冲量。利用某大型钢铁企业实际数据进行测试，该模型能够结合设备的实际生产工况变化，实现煤气的准确预测。结果表明，该模型平均绝对百分比误差小于4.95%，对变工况煤气系统有较好的预测效果。

Composite prediction model of blast furnace #br# gas based on data driven

The scientific and accurate prediction of the blast furnace gas system in iron and steel company can provide a basis for the reasonable dispatching of the coal gas, which was very important for the enterprise to improve the energy efficiency, reduced the gas discharge and the environmental pollution. The equipment conditions of blast furnace gas system in iron and steel works were complex, and the fluctuation of gas quantity was frequent, so it was difficult to predict accurately. In terms of these issues, a data driven compound prediction model of blast furnace gas was built on the basis of wavelet analysis, BP neural network and least squares support vector machines. The model taked the production plan and maintenance schedule of the blast furnace gas system into account, sets up training data sets for the production and consumption users of blast furnace gas system under different working conditions and uses multi group model parameters to predict the production, consumption and buffering capacity of blast furnace gas. Tested by the actual data of a large iron and steel enterprise, the model combined with the actual production equipment condition and the accurate prediction of gas can accurately predict the gas. The results showed that the average absolute percentage error of the model was less than 4.95%, which had a good prediction effect on the variable working gas system.