Robust ensemble learning framework for day-ahead forecasting of household based energy consumption

Abstract

Smart energy management mandates a more decentralized energy infrastructure, entailing energy consumption information on a local level. Household-based energy consumption trends are becoming important to achieve reliable energy management for such local power systems. However, predicting energy consumption on a household level poses several challenges on technical and practical levels. The literature lacks studies addressing prediction of energy consumption on an individual household level. In order to provide a feasible solution, this paper presents a framework for predicting the average daily energy consumption of individual households. An ensemble method, utilizing information diversity, is proposed to predict the day-ahead average energy consumption. In order to further improve the generalization ability, a robust regression component is proposed in the ensemble integration. The use of such robust combiner has become possible due to the diversity parameters provided in the ensemble architecture. The proposed approach is applied to a case study in France. The results show significant improvement in the generalization ability as well as alleviation of several unstable-prediction problems, existing in other models. The results also provide insights on the ability of the suggested ensemble model to produce improved prediction performance with limited data, showing the validity of the ensemble learning identity in the proposed model. We demonstrate the conceptual benefit of ensemble learning, emphasizing on the requirement of diversity within datasets, given to sub-ensembles, rather than the common misconception of data availability requirement for improved prediction.

Game Theoretical Demand Response Management and Short-Term Load Forecasting by Knowledge Based Systems on the basis of Priority Index

“…The authors in [32] calculated the weighted mean load of every hour for three preceding and similar days for short-term load forecasting. Moreover, the impact of temperature on prediction of short-term load is also considered by means of exponential association between power demand and temperature.…”

“…Moreover, the impact of temperature on prediction of short-term load is also considered by means of exponential association between power demand and temperature. Likewise, the mean prediction error for a daily peak load of France was attained 2.74% in [32]. Besides, the consequences of temperature, wind pressure and humidity, was scrutinized in [33].…”

“…Since |M| is invertible; thus, |M| is positive if its eigenvalues are non zeros and show a symmetry property. Also, the solution presented in Equation (32) depicts that κ uc 0 > 0. (27) is a best optimum solution for raising profits.…”

Developing and Comparing Different Strategies for Combining Probabilistic Photovoltaic Power Forecasts in an Ensemble Method

“…Relevant literature on load and generation forecasting is quite heterogeneous; this is highlighted by the comparative dissertations in reviews and surveys [8,9], clearly showing that no method outperforms the others in every aspect. Major efforts have been devoted to point prediction, for which researchers and practitioners often individuate Artificial Neural Networks (ANN) [10,11], K-Nearest Neighbors (KNN) [12], support vector regression [13], Random Forests (RF) [14], and multiple linear regression models [15] as the best solutions.…”

“…Pinball Score values averaged over the tasks[11][12][13][14][15]. Bold values highlight the best results for each zone.…”