A recent study published in Engineering presents a novel cross-time-scale control framework for integrated energy systems (IES) in residential buildings, aiming to address the challenges posed by photovoltaic (PV) prediction errors and optimize energy management. The research, conducted by Ziqing Wei, Xiaoqiang Zhai, and Ruzhu Wang from Shanghai Jiao Tong University, offers a practical solution to enhance the efficiency and flexibility of residential IES operations.

The study introduces a comprehensive control strategy that combines day-ahead optimal scheduling with on-the-fly flexible control, leveraging the thermal flexibility of buildings and the adaptability of heat pumps. The framework involves three key steps: solar irradiance prediction using a convolution-supported long short-term memory neural network (CNN-LSTM), day-ahead optimal scheduling of energy storage, and intra-day flexible control of the heat pump. This approach is validated through a high-fidelity residential building model in Frankfurt, Germany, using actual weather and energy usage data.

The researchers utilized a CNN-LSTM model to predict solar irradiance, which is crucial for effective day-ahead planning of energy storage systems. The model was trained on extensive historical data, including weather forecasts, solar radiation metrics, and other environmental factors. The day-ahead scheduling optimizes the charging strategies for batteries and domestic hot water (DHW) tanks based on predicted solar irradiance, time-of-use tariffs, and operational constraints. The intra-day flexible control adjusts the heat pump’s operation in real-time to compensate for deviations in PV output, ensuring stable system performance despite prediction errors.

The results demonstrate that the proposed method significantly enhances the economic viability and operational efficiency of residential IES. Compared to the day-ahead schedule alone, the flexible control strategy limits the cost increase to just 2.67%, whereas without flexible control, the cost could rise by 7.39%. The study also highlights the computational efficiency of the approach, with the transformation of the mixed-integer programming (MIP) problem into a nonlinear programming (NLP) problem reducing the computation time from 1542 seconds to 3.7 seconds for day-ahead scheduling.

The study’s findings emphasize the importance of accurate PV prediction and the need for flexible control mechanisms to manage renewable energy fluctuations. The proposed framework not only optimizes energy usage but also enhances the system’s adaptability to real-time changes, ensuring both economic benefits and thermal comfort. The research provides valuable insights for the operation and management of residential IES, supporting the broader goals of sustainability and energy efficiency in the building sector.

Future work could focus on further improving the accuracy of PV prediction models and exploring additional flexible control strategies to enhance system resilience and efficiency. The study’s innovative approach offers a promising direction for the development of smart and sustainable residential energy systems.

The paper “Optimal Scheduling and On-the-Fly Flexible Control of Integrated Energy Systems for Residential Buildings Considering Photovoltaic Prediction Errors,” is authored by Ziqing Wei, Xiaoqiang Zhai, Ruzhu Wang. Full text of the open access paper: https://doi.org/10.1016/j.eng.2025.04.021. For more information about Engineering, visit the website at https://www.sciencedirect.com/journal/engineering.