基于PCA-ANN的碱性电解水系统气体纯度预测

Gas Purity Prediction of Alkaline Water Electrolysis System Based on PCA-ANN

绿氢是以可再生能源为能源供给并通过电解水技术制备的氢。为了实现碳中和的目标,化工领域原料氢的一部分可以用绿氢替代。目前各种电解水技术中,碱性电解水技术制氢适用于化工过程。若要大规模地将碱性电解水技术用于化工生产,则关键指标氧中氢(HTO)决定了装置运行的安全性和经济性。为了预测HTO值,采用HYSYS对碱性电解水系统进行建模,并对生成的数据采用皮尔逊相关性分析:选取7个相关变量为输入参数,建立了基于主成分分析-人工神经网络(PCA-ANN)的气体纯度预测模型,得出最佳模型的平均绝对误差(MAE)为0.2030,均方差(MSE)为0.0956,决定系数(R2)为0.7154,优化边界条件异常数据后模型R2达到0.9507。

Green hydrogen is produced by water electrolysis technology with renewable energy as energy supply.In order to achieve the goal of carbon neutrality,the raw material hydrogen in chemical industry can be partly replaced by green hydrogen.At present,among all kinds of water electrolysis technologies,hydrogen production by alkaline water electrolysis technology is suitable for chemical processes.For large-scale application of alkaline hydrolysate in chemical production,the safety and economy are determined by the hydrogen to oxygen(HTO),a key index of plant operation.In order to predict the HTO value,the basic electrolytic water system was modeled by HYSYS.Pearson correlation analysis was used to select seven relevant variables as inputs for the data generated by HYSYS.The gas purity prediction model based on principal component analysis and artificial neural network(PCA-ANN)was developed.The MAE(Mean Absolute Error)value is 0.2030,MSE(Mean Square Error)value is 0.0956,R2 is 0.7154,and R2 of the model can reach 0.9507 after optimizing the abnormal data of boundary conditions.