摘要
暖通空调系统是建筑耗能的主要设备,研究空调系统的设备能耗模型对制定高效的优化策略起到重要作用.通过获取实际地源热泵系统的运行参数对系统主要能耗设备进行预测对比,使用MATLAB进行建模,建立反向传播神经网络(BPNN)、长短时记忆神经网络(LSTM)、支持向量机(SVM)、随机森林(RF)、极限学习机(ELM)、卷积神经网络(CNN)、基于遗传算法的反向神经网络(GA-BP)模型和基于思维进化算法的反向神经网络(MEA-BP)模型,并将预测模型的误差进行对比.在机组能耗预测中,预测效果最好的是MEA-BP模型,均方根误差(RMSE)、平均绝对误差(MAE)和平均绝对百分比误差(MAPE)分别为1.487、0.939和2.429％.在用户泵能耗预测中,预测效果最好的是GA-BP模型,RMSE、MAE和MAPE分别为0.098、0.085和2.314％.一般情况下,优化复合模型的准确性通常高于单一模型,但在一些实际工程上,个别单一模型在特定工程中的表现接近或超过组合模型.因此,在评价新型算法的性能时,必须全方位考量,并将其与传统单一模型进行详细对比.
Abstract
The HVAC system is the main energy consuming equipment in buildings,and studying the equipment energy consumption model of the air conditioning system plays an important role in formulating efficient optimization strategies.This article compares and predicts the main energy consuming equipment of a ground source heat pump system by obtaining the operating parameters of the system.MATLAB is used for modeling,and backpropagation neural network(BPNN),long short-term memory neural network(LSTM),support vector machine(SVM),random forest(RF),extreme learning machine(ELM),convolutional neural network(CNN),genetic algorithm based inverse neural network(GA-BP)model,and thinking evolution algorithm based inverse neural network(MEA-BP)model are established.The errors of the prediction models are compared.In the prediction of unit energy consumption,the MEA-BP model has the best prediction performance,with root mean square error(RMSE),mean absolute error(MAE),and mean absolute percentage error(MAPE)of 1.487,0.939,and 2.429％,respectively.In the prediction of user pump energy consumption,the GA-BP model has the best prediction effect,with RMSE,MAE,and MAPE of 0.098,0.085,and 2.314％,respectively.In general,the accuracy of the optimized composite model is generally higher than that of the single model,but in some practical projects,individual single models are close to or better than the combined model in specific projects.Therefore,when evaluating the performance of new algorithms,it is necessary to consider them comprehensively and compare them in detail with traditional standalone models.
维普
万方数据