Abstract
With the development of distributed power sources for the community microgrid, an increasing number of energy consumers possessing local power generation abilities will gradually transform into prosumers, balancing the dual identities of power producers and consumers. Peer-to-peer (P2P) energy trading has the potential to reduce the total cost of prosumers in community microgrids. A P2P energy trading scenario with photovoltaic (PV) systems is designed in this paper. To substantiate the effects of accurate power generation and load forecasts on this scenario, an ensemble method integrating forecasting and P2P trading is proposed. Finally, as the accuracy of power generation and load forecasts improves, bills for individual prosumer and total cost within community microgrid will approach reality. The results of proposed method can be followed by participants within community microgrid.
Introduction
As distributed energy resources, particularly rooftop photovoltaics (PV), become more prevalent, traditional energy consumers are evolving into prosumers. These entities possess the dual capability of producing and consuming electricity. However, managing surplus energy efficiently remains a significant challenge. Selling excess power back to the central grid is often financially disadvantageous, while purchasing from the grid during peak demand is costly.
Peer-to-Peer (P2P) energy trading within local community microgrids offers a promising alternative. By allowing neighbors to trade energy directly, the community can optimize local consumption and reduce reliance on the external grid. The success of such a system relies on two pillars: high-accuracy forecasting and equitable pricing strategies.
Methodology
This research proposes an integrated framework that combines a sophisticated forecasting ensemble with a strategic trading mechanism.
1. Ensemble Forecasting: The "CoF" Model
To achieve high prediction accuracy, the study employs a Combination of Forecasts (CoF) approach. Instead of relying on a single algorithm, the system integrates multiple machine learning models using a weighted ensemble method designed to minimize the sum of squared errors:
- Power Generation Forecasting: Utilizes Random Forest (RF), XGBoost (XGB), and Long Short-Term Memory (LSTM) networks.
- Load Demand Forecasting: Utilizes RF, XGB, and Deep Neural Networks (DNN).
2. P2P Trading Strategy: The "MMR" Method
The actual trading is governed by the Mid-Market Rate (MMR) method. In this model, the P2P transaction price is set at the midpoint between the grid’s buying and selling prices. This ensures a "win-win" scenario: buyers pay less than the grid rate, and sellers receive more than the grid's buy-back rate.
Experimental Setup and Data
The framework was validated using real-world data from Gansu, China, including generation data from a local PV station and residential load data. The models were trained on various meteorological features, such as surface solar radiation, temperature, humidity, and cloud cover. Notably, the study found that surface solar radiation was the primary driver for generation, while surface pressure and cloud cover significantly influenced residential demand.
Key Results
The proposed integrated method demonstrated substantial improvements in both forecasting precision and economic outcomes:
- Enhanced Accuracy: The CoF model achieved a Mean Absolute Error (MAE) of 0.210 for power generation, significantly outperforming individual models like XGBoost (0.320) and Random Forest (0.301).
- Economic Efficiency: Implementing the MMR trading strategy reduced the community's total energy costs from 96,064 yuan (under a conventional grid-only model) to 79,272 yuan—a total saving of approximately 20%.
- Market Stability: The MAE for P2P buying and selling prices remained remarkably low (0.004 and 0.003, respectively), indicating that the predicted trading prices closely mirrored actual real-time market requirements.
Conclusion and Future Work
This study confirms that accurate forecasting is the financial foundation of efficient P2P energy markets. As forecasting accuracy improves, the resulting P2P bills more closely reflect physical reality, ensuring market fairness. Future research will explore the integration of Energy Storage Systems (ESS) to allow for temporal energy shifting and investigate more complex game-theory-based pricing mechanisms to enhance scalability in larger, more diverse community microgrids.
Summary updated