【摘要】 将多能流不稳定能耗系统定性为多元非线性问题，通过短期负荷预测可提高非线性能源系统运行的稳定性和输出能源的质量。神经网络设计过程提出对7种能耗关联因素进行数据挖掘，设计时域滚动数据预测方案，神经网络误差仅为0.00143。优化后的神经网络拓扑结构，网络训练过程输入数据与输出数据作回归分析，R=0.99876，回归效果显著，数据可信。研究成果应用于建筑负荷运行策略，通过生物质燃气分布式能源系统，观察数据与预测数据作回归分析，R=0.999723，回归效果显著。更多还原

【Abstract】 In this paper,the multi-energy flow unstable energy consumption system is characterized as a multivariate nonlinear problem. The stability of the nonlinear energy system operation and the quality of output energy can be improved by short-term load forecasting. Firstly, the data mining of seven energy consumption related factors is carried out during neural network design, and then,the time domain rolling data prediction scheme is designed. The error of neural network is only 0.00143. The regression analysis of the optimized neural network topology,the input data and output data of the network training process are done. It gets R=0.99876,which indicates that the regression effect is significant,the data is credible. The research results are applied to the building load operation strategy,verified by biomass gas distributed energy system,the regression analysis between the observed data and the predicted data shows that R=0.999723,the regression effect is significant,which provides a high-precision prediction alternative way for the development of the operation strategy of the nonlinear energy consumption system.更多还原