Co-optimization of photovoltaic systems and electric vehicles using federated learning

In addressing the privacy and voltage issues in coordinated optimization of the distribution grid, the research first employs a federated learning framework with distributed photovoltaic (PV) sources for cooperative optimization. Through the use of constraint strategies, the distributed solution is optimized. Subsequently, the battery energy management of electric vehicles is utilized, coupled with adjustments to the power output of the distributed PV arrays, to actively control the voltage in the distribution grid. The results indicate that, in terms of privacy protection, the research algorithm based on federated learning avoids local data exchange, safeguarding privacy in each region. With the incorporation of distributed PV control, the situation of voltage exceeding the upper limit is completely resolved during the time period from 11 to 15. The collaborative involvement of electric vehicles and distributed PV in distributed voltage control not only compensates for voltage drops caused by peak loads but also alleviates voltage increases resulting from reverse power flow. The research algorithm demonstrates a favorable balance between ensuring good algorithm performance and providing strong privacy protection, thus possessing high practical value.