An Incremental Learning Framework for Photovoltaic Production and Load Forecasting in Energy Microgrids

Abstract

Energy management is crucial for various activities in the energy sector, such as effective exploitation of energy resources, reliability in supply, energy conservation, and integrated energy systems. In this context, several machine learning and deep learning models have been developed during the last decades focusing on energy demand and renewable energy source (RES) production forecasting. However, most forecasting models are trained using batch learning, ingesting all data to build a model in a static fashion. The main drawback of models trained offline is that they tend to mis-calibrate after launch. In this study, we propose a novel, integrated online (or incremental) learning framework that recognizes the dynamic nature of learning environments in energy-related time-series forecasting problems. The proposed paradigm is applied to the problem of energy forecasting, resulting in the construction of models that dynamically adapt to new patterns of streaming data. The evaluation process is realized using a real use case consisting of an energy demand and a RES production forecasting problem. Experimental results indicate that online learning models outperform offline learning models by 8.6% in the case of energy demand and by 11.9% in the case of RES forecasting in terms of mean absolute error (MAE), highlighting the benefits of incremental learning.

A Next Generation Library of AI-Based Data-Driven Services for the Built Environment

“…This cluster includes a series of forecasting algorithms which are focused on providing predictions for several energyrelated topics, including the overall building's consumption, the consumption of specific sub-systems of a building (i.e. lighting, heating, cooling, ventilation), the consumption of district heating networks in terms of heating and hot water, as well as the production of several renewable energy sources like photovoltaics [19], [20].…”

"Twinp2g: A Digital Twin Architecture for Optimal Power-to-Gas Planning"

“…The forecasting platform in question can handle all type of time series and integrate new datasets with little to no extra development, hence allowing to easily handle the timeseries data ingested in TwinP2G (see Section 3.2.1) and providing forecasts of various (short-, mid-, long-term) horizons. Moreover, supplementing the existing forecasting models, novel practices in the field of machine learning can be used, including incremental analytics for periodically re-training existing models (Sarmas, Strompolas, et al, 2022), as well as transfer learning to handle cases with insufficient data (Sarmas, Dimitropoulos, et al, 2022).…”

A Web-Based Service Supporting Local Governments in SECAP Implementation activities

“…Data democratisation [11] in policymaking and other sectors will allow access to both the raw data [12] of the services, and to the results of them. Significant results, in the cross-sector between AI-data driven services and open data, are demonstrated in multiple occasions and across sectors, such as the building sector [13,14,15], the energy sector [16,17,18] and the local governments policy level. Specifically, actions to address climate change are now widespread as decision-makers prepare for its impacts on areas such as transportation, water supply, public health and infrastructure, utilising data-driven approaches [19,20,21].…”