In the dynamic landscape of energy distribution, a novel approach to managing photovoltaic (PV) energy and voltage-regulating devices is making waves. Tolulope David Makanju, from the University of Johannesburg’s Department of Electrical and Electronics Engineering, has spearheaded a study that could redefine how we integrate distributed energy resources (DERs) into power grids. Published in the journal *Energies*, the research introduces a dual strategy approach that leverages machine learning (ML) to optimize the coordination of PV smart inverters (PVSIs), onload tap changers (OLTCs), and distribution static synchronous compensators (DSTATCOMs).
The study addresses a critical challenge in modern power systems: the need for efficient voltage regulation and seamless integration of renewable energy sources. “The proactive involvement of PV smart inverters in grid management facilitates voltage regulation and enhances the integration of DERs within distribution networks,” Makanju explains. However, achieving optimal control of these devices has been a complex puzzle. Makanju’s research tackles this by developing a method that forecasts the optimal setpoints for these devices, ensuring stable voltage profiles even under uncertain conditions.
The approach involves two main steps. First, a centralized AC optimal power flow (CACOPF) is run to determine the optimal active and reactive power of PVSIs, the setpoint of DSTATCOMs, and the optimal tap setting of OLTCs. This is based on hourly PV output power and load demand data. Second, machine learning models are trained to act as controllers, determining the reactive-power setpoints for PVSIs and DSTATCOMs, as well as the optimal OLTC tap position for voltage stability.
The effectiveness of this method was rigorously tested on a modified IEEE 33 bus system with high PV penetration. The results were promising: deep neural networks (DNNs) outperformed other ML models in mimicking the coordination method based on CACOPF. Moreover, the DNN-based controller proved to be more computationally efficient than the optimizer approach under varying loading and PV conditions.
This research holds significant implications for the energy sector. By enabling predictive control in power systems, it empowers system operators to make informed decisions under uncertain PV energy and loading conditions. “This approach allows predictive control in power systems, helping system operators determine the action to be initiated under uncertain PV energy and loading conditions,” Makanju notes. This predictive capability is crucial for addressing the computational inefficiencies that arise from contingencies in power systems, which often require running optimal power flow (OPF) multiple times.
The commercial impacts of this research are substantial. As the world shifts towards renewable energy sources, the ability to efficiently integrate and manage these resources within the grid becomes paramount. Makanju’s approach offers a scalable and efficient solution, potentially reducing operational costs and enhancing grid stability. This could pave the way for more widespread adoption of distributed energy resources, accelerating the transition to a more sustainable energy future.
In the ever-evolving energy landscape, Makanju’s research stands as a beacon of innovation. By harnessing the power of machine learning and optimal control strategies, it offers a glimpse into the future of grid management. As the energy sector continues to grapple with the challenges of integrating renewable energy sources, this research provides a promising path forward, one that could shape the future of power distribution and management.