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
Variable selection is a universal problem in building multivariate calibration models, such as quantitative structure-activity relationship(QSAR) and quantitative relationships between quantity or property and spectra...Variable selection is a universal problem in building multivariate calibration models, such as quantitative structure-activity relationship(QSAR) and quantitative relationships between quantity or property and spectral data. Significant improvement in the prediction ability of the models can be achieved by reducing the bias induced by the uninformative variables. A new criterion,named as C, is proposed in this study to evaluate the importance of the variables in a model. The value of C is defined as the average contribution of a variable to the model, which is calculated by the statistics of the models built with different combinations of the variables. In the calculation, a large number of partial least squares(PLS) models are built using a subset of variables selected by randomly re-sampling. Then, a vector of the prediction errors, in terms of root mean squared error of cross validation(RMSECV), and a matrix composed of 1 and 0 indicating the selected and unselected variables can be obtained. If multiple linear regression(MLR) is employed to model the relationship between the RMSECVs and the matrix, the coefficients of the MLR model can be used as a criterion to evaluate the contribution of a variable to the RMSECV. To enhance the efficiency of the method, a multi-step shrinkage strategy was used. Comparison with Monte Carlo-uninformative variables elimination(MC-UVE), randomization test(RT) and competitive adaptive reweighted sampling(CARS) was conducted using three NIR benchmark datasets. The results show that the proposed criterion is effective for selecting the informative variables from the spectra to improve the prediction ability of models.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.
基金supported by the National Natural Science Foundation of China (21475068, 21775076)
文摘Variable selection is a universal problem in building multivariate calibration models, such as quantitative structure-activity relationship(QSAR) and quantitative relationships between quantity or property and spectral data. Significant improvement in the prediction ability of the models can be achieved by reducing the bias induced by the uninformative variables. A new criterion,named as C, is proposed in this study to evaluate the importance of the variables in a model. The value of C is defined as the average contribution of a variable to the model, which is calculated by the statistics of the models built with different combinations of the variables. In the calculation, a large number of partial least squares(PLS) models are built using a subset of variables selected by randomly re-sampling. Then, a vector of the prediction errors, in terms of root mean squared error of cross validation(RMSECV), and a matrix composed of 1 and 0 indicating the selected and unselected variables can be obtained. If multiple linear regression(MLR) is employed to model the relationship between the RMSECVs and the matrix, the coefficients of the MLR model can be used as a criterion to evaluate the contribution of a variable to the RMSECV. To enhance the efficiency of the method, a multi-step shrinkage strategy was used. Comparison with Monte Carlo-uninformative variables elimination(MC-UVE), randomization test(RT) and competitive adaptive reweighted sampling(CARS) was conducted using three NIR benchmark datasets. The results show that the proposed criterion is effective for selecting the informative variables from the spectra to improve the prediction ability of models.
