Study Helps Optimize EV Charging Stations with Integrated PV Forecasting

By Muhammad OsamaReviewed by Laura ThomsonAug 12 2024

A recent study published in the journal Forecasting comprehensively explored how the accuracy of photovoltaic (PV) and electric vehicle (EV) load forecasts affects the operating costs of a microgrid-based EV charging station. The researchers developed a two-layer energy management system (EMS) that uses forecasting models to optimize microgrid operations and lower overall costs. They aimed to show the impact of combining PV and EV forecasts on costs, emphasizing the need to assess the entire system's performance instead of just focusing on individual forecast accuracy.

Background

The transportation sector is a major contributor to greenhouse gas emissions, making the shift to EVs essential for meeting decarbonization targets. This transition requires the development of reliable charging infrastructure, especially public charging stations (CSs). Integrating these stations with renewable energy sources (RESs) like PV and battery energy storage systems (BESSs) can improve energy efficiency, reduce reliance on fossil fuels, and enhance grid stability.

Microgrids are becoming essential for integrating distributed energy resources (DERs), such as EVs, PV systems, and BESSs. They can operate independently or alongside the main grid, offering greater flexibility and control over energy production, distribution, and consumption. However, efficient microgrid operation depends on accurate forecasts of EV demand and PV generation to optimize energy management and reduce operating costs.

About the Research

In this paper, the authors assessed the impact of EV forecasting accuracy on the operating costs of a microgrid equipped with a charging station, PV system, and BESS. They used a "predict-then-optimize" framework, where forecasting models predicted future EV demand and PV generation. These forecasts were then fed into an optimization model to determine the best energy management strategy. In addition to this, experiments were conducted at the Multi-Good Microgrid Laboratory (MG²Lab) at Politecnico di Milano, alongside offline simulations. The MG²Lab integrates various DERs, including solar PV, combined heat and power, BESS, and hydrogen storage, allowing the EMS to be tested under real-world conditions.

The EMS used a two-layer approach: the first layer applied a mixed-integer linear programming (MILP) optimization model to determine the best dispatch of the BESS over 24 hours, considering EV load and PV generation forecasts. The second layer used a heuristic method to make real-time adjustments, correcting discrepancies between forecasts and actual values to ensure system balance and efficiency.

The researchers examined two EV load forecasting methods: a long short-term memory (LSTM) neural network and a persistence approach. PV generation forecasts were obtained using a physical hybrid artificial neural network (PHANN) model. The accuracy of these forecasts was evaluated using statistical error metrics such as root mean squared error (RMSE), mean absolute error (MAE), and symmetric mean absolute percentage error (SMAPE).

Research Findings

Online experiments at the MG²Lab facility showed that the developed EMS and forecasting experiments at the MG²Lab showed that the developed EMS and forecasting models worked well in real-world settings, as the experimental results closely matched the offline simulations, validating the approach.

The offline simulations, which included different seasons, revealed that the accuracy of combined PV and EV forecasts significantly influenced microgrid operating costs, with potential cost variations of up to 10% compared to ideal scenarios with perfect forecasts.

PV and EV forecasts contributed equally to the cost variations, each accounting for about 5%. However, the interaction between PV and EV forecasts sometimes led to unexpected results, either amplifying or reducing overall inaccuracies.

While the LSTM model provided more accurate EV load predictions than the persistence approach, using the LSTM-based EMS with the PHANN PV forecast resulted in higher operating costs. This outcome suggests that forecast accuracy alone is not enough. The timing of electricity prices should also be considered when training forecasting models.

Applications

This research has important implications for designing and operating microgrid-based EV charging stations. The proposed EMS, which integrates forecasting models and optimization techniques, can improve the efficiency and cost-effectiveness of these systems. By understanding how forecast accuracy impacts operating costs, system designers can make informed decisions about the trade-offs between forecast complexity, computational demands, and overall system performance.

The insights from this study can also guide the development of customized loss functions for training forecasting models, focusing on the specific needs of downstream decision-making processes, such as energy management optimization, rather than relying only on generic statistical metrics.

Conclusion

Combined PV and EV forecasting effectively estimated microgrids' operating costs. The authors emphasized the importance of considering the time-dependent nature of electricity prices and the interactions between different forecasting models when designing and optimizing these systems.

Future work should focus on creating customized loss functions for training forecasting models tailored to the specific needs of energy management optimization. Incorporating more advanced uncertainty quantification techniques, such as robust optimization or stochastic programming, could further improve the resilience and reliability of energy management strategies.

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.