A machine learning-driven MCDA-TOPSIS framework for wave energy converter selection in the Philippines

Abstract

Wave energy is a promising yet underutilized resource for archipelagic nations such as the Philippines. Wave energy converters hold significant potential for diversifying renewable portfolios. This study introduces a hybrid Multi-Criteria Decision Analysis framework that couples scenario-based TOPSIS with machine learning-derived weights to reduce subjectivity and improve robustness in prioritizing WEC technologies for microgrid integration. The framework is applied to two datasets: Set-1 (general WEC types) and Set-2 (coastal-compatible onshore devices). Scenario-based TOPSIS evaluates ten scenarios with different normalizations and weighting schemes, while ML-TOPSIS derives criterion weights from scenario-based closeness means using eXtreme Gradient Boosting (XGBoost). Validation employs leave-one-out cross-validation with bootstrap confidence intervals, and rank agreement is tested using Spearman's ρ, Kendall's τ, and Top-k Jaccard overlap. Results show that scenario-based rankings remain coherent under balanced and techno-economic weights, whereas entropy weighting yields less stable outcomes. ML-TOPSIS aligns with the scenario-based signal without replicating. Learned importances emphasize cost, efficiency, technology readiness, and deployment feasibility, enabling transparent and adaptive weight generation. Across methods, Bulge Wave, OWC, and Point Absorber Devices (Set-1), and Power Wing and Wave Clapper (Set-2) consistently rank highest, with Point Absorber and Power Wing emerging as robust performers under sensitivity analysis. By mapping criteria to SDGs and aligning with RA9513 and the Philippine Energy Plan 2023–2050, this study provides one of the first ML-driven MCDA applications for marine renewables in a developing archipelagic context. The findings contribute practical value by informing policymakers and developers on sustainable technology choices for pilot deployment in the Philippines' renewable microgrid transition.