Solar-Panel-Efficiency-Prediction

☀️ Solar Panel Efficiency Prediction

Submission for the Zelestra X AWS ML Ascend Challenge - 2nd Edition
High-accuracy predictive modeling for solar panel efficiency

---

📘 Overview

This project applies a structured machine learning pipeline to forecast solar panel efficiency using environmental and operational data. The final model leverages ensemble learning and advanced feature engineering to achieve high predictive accuracy.

---

🧰 Tools & Technologies

• Language: Python 3
• Environment: Jupyter Notebooks (local setup)
• Libraries:
  • Data Handling: pandas, numpy
  • Visualization: matplotlib, seaborn
  • Machine Learning: scikit-learn, xgboost, lightgbm, catboost, optuna

---

📊 Methodology

1. Data Preprocessing

• Cleaned missing/inconsistent values
• Removed outliers
• Scaled numerical features using StandardScaler / MinMaxScaler
• Encoded categorical variables

2. Feature Engineering

• Derived new features: power ratio, panel efficiency
• Created interaction features from temperature, irradiance, humidity
• Removed multicollinear and low-variance features

3. Model Development

• Baseline models: Linear Regression, Decision Trees
• Advanced models: XGBoost, LightGBM, CatBoost
• Final model: Stacked Regressor combining top-performing models
• Hyperparameter tuning via Grid Search and Optuna

4. Evaluation Strategy

• Metrics: RMSE, MAE, R² Score
• Validation: 5-Fold Cross-Validation
• Target: >90% accuracy on test data
• Leaderboard feedback used for iterative improvements

---

📁 File Mapping

File Name	Description
Solar_V2.ipynb	Intermediate enriched model development
solar_v1_accuracy_89.90197.ipynb	First successful version with ~90% accuracy
Solar_v3_tus (1).ipynb	Final stacked model with performance tuning

---

🚀 Reproduction Steps

To reproduce the results locally:

1. Install dependencies

You can install all required packages using:

pip install matplotlib seaborn numpy pandas lightgbm xgboost scikit-learn catboost scipy optuna