TL;DR
This paper presents a low-cost, non-intrusive management framework for retrofitting existing homes with photovoltaic (PV) systems to maximize self-consumption. By using simple sensors and rule-based logic, the system effectively coordinates PV production with heating and hot water demands, achieving up to 89% self-consumption and significantly reducing grid reliance.
What Problem Does This Paper Address?
The increasing adoption of PV systems in homes is a crucial step towards decarbonization, but simply installing panels isn't enough. Many existing homes have legacy heating and hot water systems that operate independently, leading to suboptimal energy use. This "integration gap" means that even with PV, homes often rely on the grid, especially during peak demand or when PV production is low. Furthermore, integrating diverse, multi-vendor equipment into a smart system is challenging due to proprietary interfaces and the high cost and complexity of advanced control methods like Model Predictive Control (MPC). This paper addresses the need for an affordable, adaptable, and minimally invasive solution to maximize PV self-consumption in existing residential buildings.
Methods
The core of the proposed solution is a rule-based control strategy that leverages readily available Commercial-Off-The-Shelf (COTS) sensors and actuators. The framework monitors battery voltage and grid power exchange to intelligently manage loads like heat pumps and hot water cylinders. A key innovation is the introduction of three "service classes" (Low, Medium, High Quality of Service - QoS) that define different levels of heating and hot water provision. These classes are activated based on priority and background guards, which use battery voltage and spare PV power as triggers. This approach allows for concurrent activation of loads, meaning the heat pump can run while the battery is charging, unlike traditional sequential methods that prioritize battery charging first. The system is designed to be retrofitted without requiring modifications to existing equipment, relying on simple sensor readings and relay-based actuation.
Datasets
The paper utilizes data collected from a real-world residential testbed in Northern Italy over a full calendar year (2025). This dataset includes PV production, total consumption, heat pump usage, space heating, domestic hot water (DHW) production, cooling, standby loads, grid feed-in, grid draw, and the resulting self-consumption rate. The authors also present simulated results based on this data and compare them against experimental findings.
Key Insights and Ideas
- Concurrent Activation is Key: The most significant insight is that allowing loads (like heat pumps) to operate concurrently with battery charging, triggered by battery voltage, dramatically improves self-consumption compared to sequential activation. This prevents PV power from being unnecessarily sent to the grid or waiting for batteries to fill.
- QoS-Based Service Classes Offer Flexibility: Defining distinct service classes allows for a user-centric approach, balancing energy efficiency with thermal comfort. This makes the system adaptable to different user needs and preferences.
- Minimal Intrusiveness is Achievable: The use of COTS components and simple rule-based logic demonstrates that effective smart home integration for PV self-consumption can be achieved without expensive custom hardware or complex firmware modifications. This significantly lowers the barrier to adoption for existing buildings.
- Battery Size Matters, But Less Than Expected: Ablation experiments showed that even with reduced battery capacity (down to 5 kWh), the automated system still significantly outperformed baseline control, highlighting the effectiveness of the intelligent management strategy.
Key Results
- The proposed framework achieved up to 89% PV self-consumption in the residential testbed.
- Daily grid usage was kept below 0.5 kWh.
- The system demonstrated a 20% increase in Load Matching Index (LMI) compared to baseline configurations.
- The total cost for the IoT automation layer was approximately EUR 250.
- The framework achieved 88% of the performance of an "oracle" strategy that had perfect foresight, while avoiding the complexity and hardware requirements of such advanced methods.
- Thermal comfort was maintained, with Category II compliance (recommended for new buildings) increasing from 15% to 19% compared to sequential RBC.
Limitations and Future Work
- Seasonal and Geographic Scope: The experiments were primarily conducted in mid-March in a heating-dominated climate zone. Further data collection across different seasons and climates is needed for more generalized guidelines.
- Single Testbed: Results are based on a single residential building. Generalizing to other building typologies (apartments, commercial buildings) requires additional deployments.
- Software Reliability: While the MTBF was weeks to months, improving software reliability is an ongoing engineering priority.
- Future Work: The authors plan to:
- Integrate learning algorithms for PV production and user comfort prediction.
- Extend the framework with user localization capabilities.
- Integrate electric vehicle (EV) charging policies.
- Develop a verification tool for the IoT management framework.
- Investigate security and fault-tolerance concerns.
- Explore data-driven extensions to close the remaining performance gap with predictive load shifting.
Summary updated