Blog 5 - Use of machine learning to forecast generation and demand in electrical energy systems
Abdulaziz, 13 October 2023
As the world shifts towards renewable energy sources, grid management encounters unprecedented challenges. The integration of renewables introduces fluctuations in energy production, further complicating the intricate task of maintaining a balance between supply and demand [1].
Machine learning presents a robust solution to address these challenges. Various ML algorithms, including Random Forest, Decision Trees, Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU), have been harnessed for forecasting energy supply and demand. These algorithms adeptly process extensive datasets, yielding remarkably accurate projections [2].
Machine learning models consider a wide array of variables to make accurate forecasts. These range from weather patterns and historical consumption rates to social events and other intermittent factors. Their ability to analyse large and diverse datasets provides grid operators with valuable insights for planning electrical production and distribution [3].
Comparison between actual solar data and prediction obtained with a Machine Learning model
Reinforcement Learning (RL), a specialized subfield of ML, offers significant benefits for real-time adjustments to both supply and demand. By interacting with the environment to achieve a goal, such as maximizing energy output while minimizing waste, RL algorithms can adapt to fluctuating energy production patterns. This is especially valuable when dealing with the variable nature of renewable resources like solar and wind power [4]. As we integrate more renewable energy sources into our grids, machine learning algorithms, particularly those specializing in reinforcement learning, will become essential tools in managing these complex systems.
Figure 1 shows an example of the use of machine learning to forecast solar generation - the model is able to track subtle changes and identify complex patterns. By continuously improving their predictive capabilities with each new piece of data, these algorithms contribute to creating more efficient, reliable, and sustainable electrical grids.
References:
[1] Adeoye, O., & Spataru, C. (2019). Modelling and forecasting hourly electricity demand in West African countries. Applied Energy, 242, 311-333.
[2] Eseye, A. T., Lehtonen, M., Tukia, T., Uimonen, S., & Millar, R. J. (2019). Machine learning based integrated feature selection approach for improved electricity demand forecasting in decentralized energy systems. IEEE Access, 7, 91463-91475.
[3] Grimaldo, A. I., & Novak, J. (2020). Combining machine learning with visual analytics for explainable forecasting of energy demand in prosumer scenarios. Procedia computer science, 175, 525-532.
[4] Cebekhulu, E., Onumanyi, A. J., & Isaac, S. J. (2022). Performance analysis of machine learning algorithms for energy demand–supply prediction in smart grids. Sustainability, 14(5), 2546.
Contact: abdulaziz.alhayd@kcl.ac.uk