Analysis of the Integration of Drift Detection Methods in Learning Algorithms for Electrical Consumption Forecasting in Smart Buildings

Abstract

Buildings are currently among the largest consumers of electrical energy with considerable increases in CO2 emissions in recent years. Although there have been notable advances in energy efficiency, buildings still have great untapped savings potential. Within demand-side management, some tools have helped improve electricity consumption, such as energy forecast models. However, because most forecasting models are not focused on updating based on the changing nature of buildings, they do not help exploit the savings potential of buildings. Considering the aforementioned, the objective of this article is to analyze the integration of methods that can help forecasting models to better adapt to the changes that occur in the behavior of buildings, ensuring that these can be used as tools to enhance savings in buildings. For this study, active and passive change detection methods were considered to be integrators in the decision tree and deep learning models. The results show that constant retraining for the decision tree models, integrating change detection methods, helped them to better adapt to changes in the whole building’s electrical consumption. However, for deep learning models, this was not the case, as constant retraining with small volumes of data only worsened their performance. These results may lead to the option of using tree decision models in buildings where electricity consumption is constantly changing.

Dynamic Modeling With Integrated Concept Drift Detection for Predicting Real-Time Energy Consumption of Industrial Machines

“…According to the literature, concept drift can be detected using active or passive methods, as described in Section 2 [14], [15], [35], [39]. Active methods are primed to detect changes and are re-trained when a trigger is recognized, while passive methods are re-trained at regular intervals regardless of whether a change has occurred [20]. Equations 1 and 2 identify the changes in the distributions.…”

“…Compared with the nonadaptive model, their proposed adaptive strategy outperformed the conventional approaches in terms of energy prediction performance. Hernandez et al [20] utilized active and passive concept drift detection approaches in the DT and deep learning models. The results indicate that constant retraining of the decision tree models, together with change detection methods, can improve their ability to adapt to changes in the total electrical consumption of a building.…”

“…The limitations of MAPE are paired with the MAE, which displays how much inaccuracy is expected from the forecast on average, assisting in determining which models are superior. Because the MAE has trouble differentiating between significant and minor mistakes, it was paired with the RMSE to be safe [9], [20].…”

“…In the best-known active drift detection systems, the user determines the window size, which uses the most recent batch of data with the most recent training instances [20]. However, standard drift detection methods are not designed for complex data streams from industrial machines.…”

“…As a result, conventional machine-learning algorithms struggle to handle changes in real-world streaming data [15]. Traditional electricity energy forecasting models are often trained once and then not re-trained with new data; thus, the prediction performance of the original model deteriorates since industrial data generally changes over time [20]. Most researchers traditionally split all available data into certain parts for training and testing [12], [21], [22].…”