As solar energy systems scale up globally, predicting performance degradation and panel failures is critical for minimizing downtime and maximizing energy output. This project applies a stacked regression ensemble using sensor and environmental data to forecast solar panel efficiency, enabling predictive maintenance and operational optimization.

• Training File: data/train.csv (20,000 rows × 17 columns)
• Test File: data/test.csv (12,000 rows × 16 columns)
• Sample Submission: data/sample_submission.csv
• Target Variable: efficiency
• Index: id

• Numerical: temperature, irradiance, humidity, panel_age, maintenance_count, soiling_ratio, voltage, current, module_temperature, cloud_coverage, wind_speed, pressure
• Categorical: string_id, error_code, installation_type

• Data Preprocessing: pandas, numpy, seaborn, matplotlib, scikit-learn
• Machine Learning Models:
  • RandomForestRegressor
  • LightGBMRegressor
  • CatBoostRegressor
  • ANN using TensorFlow and SciKeras
• Model Tuning: GridSearchCV
• Ensemble: Manual stacking with Ridge Regression as meta-model

• Missing value imputation:
  • Mean imputation for continuous variables.
  • Mode imputation for categorical variables like maintenance_count.
• Label encoding for categorical features.
• Added engineered feature: power_output = voltage * current.
• Log transformation for voltage to reduce skew.
• Scaled numerical features with RobustScaler.

1. Base Models:
   • Random Forest
   • LightGBM
   • CatBoost
   • ANN using TensorFlow (via KerasRegressor)
2. Stacking:
   • OOF (Out-of-Fold) predictions generated for training meta-model
   • Final meta-model: Ridge Regression

• ✅ Validation RMSE: 0.10610
• ✅ Mean CV RMSE: 0.10213
• ✅ OOF and CV-based stacking strategy used for generalization
• ✅ Submission generated using predictions on test set

• Enables smart monitoring systems to proactively maintain PV panels.
• Reduces energy loss and improves sustainability in large-scale solar farms.
• Supports intelligent energy grid design by forecasting panel health trends.

git clone https://github.com/rohankharche34/Solar-panel-performance-optimization.git cd Solar-panel-performance-optimization

python -m venv venv source venv/bin/activate # on Windows use `venv\Scripts\activate`

pip install -r requirements.txt

Python 3.10

Scikit-learn 1.4.2

XGBoost 2.0.3

LightGBM 4.3.0

CatBoost 1.2.3

Tensorflow==2.16.1

Scikeras==0.12.0

Pandas, NumPy, Matplotlib, Seaborn