Load Forecasting Under Concept Drift: Online Ensemble Learning With Recurrent Neural Network and ARIMA

Abstract

Rapid expansion of smart metering technologies has enabled large-scale collection of electricity consumption data and created the foundation for sensor-based load forecasting on individual buildings or even the household level. With continuously growing energy consumption, the importance of energy management including load forecasting is increasing in order to remedy the energy effect on the environment. Numerous machine learning techniques have been proposed for sensor-based load forecasting but most are offline approaches: the model is trained once and then used to infer future consumption. However, these approaches are not able to adapt to concept drift: for example, their accuracy will degrade when the building use changes or new equipment is installed. Thus, an approach capable of learning from new data as they arrive is needed. This paper proposes adaptive online ensemble learning with Recurrent Neural Network (RNN) and ARIMA for load forecasting under concept drift. The RNN part of the ensembles consists of Online Adaptive RNN as its underlying RNN learner has the ability to model temporal dependencies present in load data while its online nature enables continuous learning from arriving data. The adaptation to the concept drift is improved by adding Rolling ARIMA to the ensemble. The performance of the proposed approach has been examined on the four individual homes with different degrees of concept drift. The results show that the proposed ensemble achieves better accuracy than its constituent algorithms alone and, moreover, the analysis demonstrates the need to examine load forecasting approaches in respect to how they handle concept drift.

Design of Ensemble Forecasting Models for Home Energy Management Systems

“…Nowadays, linear regression is often used as a comparative method to evaluate the performance of more elaborated machine learning methods. Studies that used ARIMA as an energy prediction method for buildings may be found in [38][39][40].…”

Deterioration of Electrical Load Forecasting Models in a Smart Grid Environment

“…A similar phenomenon occurs in dynamic electrical environments, which involves S.G. and D.G.M. Some other studies considering the concept drifts are: Jagait, R. K., et al [40] presented an ensemble approach for online learning using R.N.N.s and ARIMA. The study proposed an adaptive model based on concept drift using R.N.N.s and considered ARIMA for creating a rolling window operation to create an online scenario for the ensemble model.…”

“…These models are capable of self-learning and adaptation to changes. Though the researchers are covering some dynamics in other fields for the transformation of models in classification applications [49][50][51], some propositions in regression works also exist [25,40]. Still, in regression, many features need to be incorporated when the task is specifically of electrical load forecasting to enable the model to be eligible for multi-modality or Smart Grid.…”

Stacking Ensemble Methodology Using Deep Learning and ARIMA Models for Short-Term Load Forecasting

“…First, the proposed methodology is guided by the ensemble methods (learning) concept, where an improved predictive performance is expected when employing/combining multiple methods versus any of the individual methods. More specifically, this work is centered on a stacking ensemble type approach, which in a broad sense involves training a machine learning method to merge the predictions of some other methods [41]. To reinforce this point, the literature in the forecasting field documents report that when dealing with an unstable and varying data pattern, it is convenient to use dissimilar models to improve time-series forecasting accuracy [42].…”