An Architecture to Improve Energy-Related Time-Series Model Validity Based on the Novel rMAPE Performance Metric

Abstract

In this paper, an architecture based on computational intelligence for time series modeling is proposed to guarantee the automatic adjustability of trained models no matter the dynamic behavior of the modeled phenomena. The proposed method can assess the performance; and then proposes a maintenance routine for the time-series model. Thus, an auditor is devised to identify when a model must be updated before losing forecast performance. It has been determined that the MAPE (Mean Absolute Percentage Error) metric could not reveal changes in the model predicted curve contributing to invalidating the model, in particular if a non-stationary behavior is expected in the studied phenomena. Therefore, the novel rMAPE performance metric is proposed, so that the auditor does detect that the updating process does not achieve better performance; the system opts for replacing the time-series modeling techniques included in an available knowledge base. The intelligent system allows building time-series models automatically considering exogenous variables such as weather, calendar, and statistical transformations that can lead to the number of models required for a particular application. The proposed approach has been experimentally tested for power consumption and energy price via simulation. The forecasting results showed an improvement in the MAPE of up to 23% in the tests performed.INDEX TERMS. Forecasting, intelligent systems, time series modeling.