The role of heating load forecasts in the energy transition is significant,given the considerable increase in the number of heat pumps and the growing prevalence of fluctuating electricity generation.While machine lea...The role of heating load forecasts in the energy transition is significant,given the considerable increase in the number of heat pumps and the growing prevalence of fluctuating electricity generation.While machine learning methods offer promising forecasting capabilities,their black-box nature makes them difficult to interpret and explain.The deployment of explainable artificial intelligence methodologies enables the actions of these machine learning models to be made transparent.In this study,a multi-step forecast was employed using an Encoder–Decoder model to forecast the hourly heating load for an multifamily residential building and a district heating system over a forecast horizon of 24-h.By using 24 instead of 48 lagged hours,the simulation time was reduced from 92.75 s to 45.80 s and the forecast accuracy was increased.The feature selection was conducted for four distinct methods.The Tree and Deep SHAP method yielded superior results in feature selection.The application of feature selection according to the Deep SHAP values resulted in a reduction of 3.98%in the training time and a 8.11%reduction in the NRMSE.The utilisation of local Deep SHAP values enables the visualisation of the influence of past input hours and individual features.By mapping temporal attention,it was possible to demonstrate the importance of the most recent time steps in a intrinsic way.The combination of explainable methods enables plant operators to gain further insights and trustworthiness from the purely data-driven forecast model,and to identify the importance of individual features and time steps.展开更多
Accurately filling in missing heating data is essential for ensuring data quality in applications such as energy management optimization and building efficiency analysis.Traditional machine learning methods use histor...Accurately filling in missing heating data is essential for ensuring data quality in applications such as energy management optimization and building efficiency analysis.Traditional machine learning methods use historical heating data as an input feature to predict the following missing data.However,when the duration of missing data is long,previous estimated values are inevitably used for further imputation,leading to error accumulation and a growing deviation from true values.To overcome this problem,this paper proposes a generative network that can fill missing data solely based on weather and temporal data,without using previous imputed values for further imputation.Our method outperformed the state of the art such as Seq2seq and Transformer,achieving relative normalized root mean square error(NRMSE)reductions of 1.65%to 41.38%,0.30%to 66.43%,and 14.84%to 50.22%across three different data sources.In addition,with our proposed method,the effect of selecting different weather variables on model performance,and the benefits of transfer learning under limited data were also demonstrated.The relative NRMSE reduction is between 3.88%to 15.85%in cold months and from 7.49%to 12.29%in warm months when applying transfer learning.展开更多
基金the German Federal Ministry for Economic Affairs and Climate Action in the framework of the research program EnOB:ML-EBESR 03EN1076B.
文摘The role of heating load forecasts in the energy transition is significant,given the considerable increase in the number of heat pumps and the growing prevalence of fluctuating electricity generation.While machine learning methods offer promising forecasting capabilities,their black-box nature makes them difficult to interpret and explain.The deployment of explainable artificial intelligence methodologies enables the actions of these machine learning models to be made transparent.In this study,a multi-step forecast was employed using an Encoder–Decoder model to forecast the hourly heating load for an multifamily residential building and a district heating system over a forecast horizon of 24-h.By using 24 instead of 48 lagged hours,the simulation time was reduced from 92.75 s to 45.80 s and the forecast accuracy was increased.The feature selection was conducted for four distinct methods.The Tree and Deep SHAP method yielded superior results in feature selection.The application of feature selection according to the Deep SHAP values resulted in a reduction of 3.98%in the training time and a 8.11%reduction in the NRMSE.The utilisation of local Deep SHAP values enables the visualisation of the influence of past input hours and individual features.By mapping temporal attention,it was possible to demonstrate the importance of the most recent time steps in a intrinsic way.The combination of explainable methods enables plant operators to gain further insights and trustworthiness from the purely data-driven forecast model,and to identify the importance of individual features and time steps.
文摘Accurately filling in missing heating data is essential for ensuring data quality in applications such as energy management optimization and building efficiency analysis.Traditional machine learning methods use historical heating data as an input feature to predict the following missing data.However,when the duration of missing data is long,previous estimated values are inevitably used for further imputation,leading to error accumulation and a growing deviation from true values.To overcome this problem,this paper proposes a generative network that can fill missing data solely based on weather and temporal data,without using previous imputed values for further imputation.Our method outperformed the state of the art such as Seq2seq and Transformer,achieving relative normalized root mean square error(NRMSE)reductions of 1.65%to 41.38%,0.30%to 66.43%,and 14.84%to 50.22%across three different data sources.In addition,with our proposed method,the effect of selecting different weather variables on model performance,and the benefits of transfer learning under limited data were also demonstrated.The relative NRMSE reduction is between 3.88%to 15.85%in cold months and from 7.49%to 12.29%in warm months when applying transfer learning.
