Machine Learning-Driven Multi-Objective Energy Management and Optimal Power Flow of Grid-Connected Photovoltaic–Battery Energy Storage Systems for Enhanced Grid Resilience and Renewable Energy Integration

Abstract

Grid-connected photovoltaic-battery energy storage systems face a persistent operational problem because variable photovoltaic generation, fluctuating electricity demand, battery state-of-charge and degradation constraints, and network power-flow requirements must be coordinated simultaneously, while existing research has frequently examined forecasting, battery scheduling, optimization, optimal power flow, and grid resilience as separate functions. This study aimed to quantitatively evaluate how machine learning-based predictive energy management, multi-objective energy optimization, intelligent battery energy storage management, and machine learning-assisted optimal power flow contribute to Grid Resilience and Renewable Energy Integration Performance in selected grid-connected PV-BESS case environments. A quantitative, cross-sectional, case-study-based design was employed, using purposive sampling and a structured five-point Likert-scale questionnaire administered to technically experienced professionals associated with power systems, renewable energy, battery storage, grid operations, energy management, and machine learning. Of 330 questionnaires distributed, 305 were returned and 296 usable responses were retained, yielding an 89.7% usable response rate. Data analysis incorporated reliability testing, descriptive statistics, Pearson correlation, and multiple regression modeling. The overall instrument demonstrated strong reliability, with Cronbach’s alpha of .93. Descriptively, Grid Resilience and Renewable Energy Integration Performance Recorded M = 4.18, SD = 0.54, while Multi-Objective Energy Optimization achieved the highest independent-variable mean of M = 4.16, SD = 0.55. All four predictors were positively and significantly related to performance: MLPEM, r = .68; MOO, r = .72; IBESS, r = .70; and MLOPF, r = .75, all p < .001. The regression model explained 69.4% of the variance in performance, R² = .694, Adjusted R² = .690, F (4, 291) = 165.02, p < .001. MLOPF emerged as the strongest predictor, β = .31, followed by MOO, β = .28, IBESS, β = .24, and MLPEM, β = .22. These findings imply that utilities, renewable-energy developers, and grid operators can strengthen resilient renewable-energy integration by coordinating predictive intelligence, multi-objective optimization, intelligent storage control, and network-aware power-flow management as an integrated operational architecture rather than isolated technological capabilities.