KINETIC

BSc (Hons) Artificial Intelligence and Data Science | Data Science Group Project

Rajapaksha Subasin Pathiranage Saniru Uthsara

Dimbulkumbure Gedara Sasiri Akalanka

Obada Kankanamge Dewmi Tharunya Sulakshani

Payagala Waduge Gangamini Harshitha

Sri Lanka aims to achieve 70% renewable energy by 2030, yet residential solar adoption remains critically low. The primary barrier is financial uncertainty, our survey found that 59.4% of homeowners lack reliable tools to assess whether solar investment will deliver returns, especially given unpredictable monsoon weather, unknown grid connectivity, and volatile electricity tariffs. To address this gap, this project develops an integrated AI-driven Solar ROI Prediction and Recommendation System. The system combines four core components: (1) solar generation forecasting using Gradient Boosting, achieving a Mean Absolute Error of 18.33 kWh/kW—a 72.4% improvement over traditional physics-based methods; (2) geospatial transformer suitability analysis using Random Forest with 100% accuracy; (3) electricity consumption forecasting using LSTM networks with clustering-based similar household matching, achieving a test MAE of 58.70 kWh with 93.8% of predictions within ±150 kWh; and (4) risk-aware financial modeling using Monte Carlo simulation running 2,000 iterations to generate probabilistic Return on Investment and payback period estimates with P10/P50/P90 confidence intervals. The system accepts user location and just 3 months of electricity bills, then outputs a comprehensive solar investment report including 12-month generation and consumption forecasts, transformer headroom analysis with curtailment risk assessment, ROI with confidence intervals, and downloadable PDF reports. Validated by domain experts from LECO and the energy sector, this system empowers Sri Lankan homeowners to make confident, data-driven solar investment decisions, supporting the nation’s renewable energy transition goals.