MMC-integrated solar PV with ANFIS MPPT and ANN for grid overload mitigation

This study details the design, modeling, and simulation of a 100-kW grid-connected solar photovoltaic (PV) system, incorporating advanced control and forecasting techniques to alleviate peak load issue at the Bahir Dar 15 kV distribution substation. The significant increase in load demand and dependence on diesel generators have elevated operational expenses and placed considerable stress on Ethiopia's grid infrastructure. An Artificial Neural Network (ANN) was developed to forecast peak load using 44 months of domestic customer data, predicting substation overloading within 22 months of 2023. A two-stage power conversion system is proposed, consisting of an Adaptive Neuro-Fuzzy Inference System (ANFIS)-based boost converter for Maximum Power Point Tracking (MPPT) and a five-level Modular Multilevel Converter (MMC) utilizing Voltage Oriented Control (VOC) for integration with the AC grid. The economic analysis indicates that the proposed system, which employs 41 series and 8 parallel panel configuration, generates over 170,000 kWh annually. The total installation cost is $39,013, with a short payback period under 2.5 years, thereby demonstrating its financial viability in comparison to diesel-based systems. Simulation results in MATLAB/Simulink indicated effective peak load shaving, precise maximum power point tracking under varying irradiance and temperature conditions, and efficient grid power injection with minimal harmonics. This study bridges the gap between predictive load forecasting and real-time control by integrating ANN-based grid overload prediction with ANFIS-MPPT and MMC, validated via a 100-kW case study in Ethiopia.