Yash957/solar-energy-forecasting

Intelligent Solar Energy Generation Forecasting and Agentic Grid Optimization System

Project Overview

This project aims to design and implement an AI-driven solar energy forecasting and grid optimization system. The system predicts solar power generation using historical and weather-related data and extends this capability into an agentic AI assistant that generates structured, explainable recommendations for grid optimization and energy utilization.

The project is developed in two milestones:

• —Milestone 1: Machine Learning–based Solar Energy Forecasting
• —Milestone 2: Agentic AI Grid Optimization Assistant

The final application is publicly hosted, uses only free-tier/open-source tools, and provides a user-friendly web interface.

Project Demo

Demo Video: Demo video (Google Drive)

Problem Statement & Use Case

Solar energy generation is highly variable due to weather conditions and seasonal patterns. Grid operators require accurate forecasts and intelligent decision support to:

• —Balance supply and demand
• —Reduce energy wastage
• —Improve renewable energy integration
• —Plan storage and load shifting strategies

This project provides:

• —Accurate short-term/long-term solar generation forecasts
• —Automated analysis of variability and risk
• —AI-generated, structured grid optimization recommendations

Overall System Architecture

High-Level Architecture

User (Browser)
   │
   ▼
Streamlit Web UI
   │
   ├── Data Upload & Selection
   ├── Forecast Horizon Selection
   ├── Visualization (Graphs & Metrics)
   │
   ▼
Backend Analytics Layer (Python)
   │
   ├── Data Preprocessing & Feature Engineering
   ├── ML / Time-Series Forecasting Models
   ├── Model Evaluation (MAE, RMSE)
   │
   ▼
Forecast Outputs (Structured Data)
   │
   ├── Visualization Module
   └── Agentic AI System (Milestone 2)
           │
           ├── Forecast Analysis Agent
           ├── Variability & Risk Detection Agent
           ├── Knowledge Retrieval Agent (RAG)
           └── Optimization Recommendation Agent
                   │
                   ▼
           Structured Grid Optimization Report

Note: The agentic AI system provides decision-support recommendations only and does not directly control any grid infrastructure.

Milestone 1: ML-Based Solar Energy Forecasting (Mid-Sem)

Objective

Build a machine learning or time-series forecasting system to predict solar energy generation.

Inputs

• —Historical solar power generation data
• —Weather indicators (irradiance, temperature, cloud cover)
• —Time-based features (hour, day, month, season)

Functional Requirements

• —Data preprocessing and cleaning
• —Feature engineering
• —Solar energy forecasting
• —Trend and seasonality analysis
• —Visualization of predictions

Technical Requirements

• —ML or time-series models such as:
• —Linear Regression
• —Random Forest Regressor
• —ARIMA / SARIMA
• —Prophet (optional)

• —Evaluation metrics:
• —MAE (Mean Absolute Error)
• —RMSE (Root Mean Square Error)

UI Requirements

• —Dataset upload or selection
• —Forecast horizon selection
• —Line plots (Actual vs Predicted)
• —Trend and seasonality visualizations

Milestone 1 Deliverables

• —Problem understanding & use-case description
• —Input–output specification
• —Forecasting pipeline architecture
• —Working application with basic UI
• —Forecast accuracy evaluation

Milestone 2: Agentic AI Grid Optimization Assistant (End-Sem)

Objective

Extend the forecasting system into an agentic AI assistant that reasons about forecast variability and generates structured grid optimization recommendations.

Functional Requirements

• —Analyze forecast outputs and uncertainty
• —Identify variability and risk periods
• —Retrieve renewable energy and grid management guidelines
• —Generate structured optimization recommendations
• —Handle incomplete or uncertain data gracefully

Technical Requirements

• —Open-source or free-tier LLM integration
• —Agentic workflow with explicit state management (LangGraph)
• —Retrieval-Augmented Generation (optional but recommended)
• —Prompt strategies to avoid unsupported claims

Structured Output Report

The generated report includes:

• —Solar generation forecast summary
• —Identified variability and risk periods
• —Grid balancing and storage recommendations
• —Energy utilization optimization strategies
• —Supporting references

Optional Extensions

• —Battery storage optimization analysis
• —Multi-site solar forecasting
• —PDF export of optimization report
• —Scenario-based energy planning

Dataset

• —Solar Energy Power Generation Dataset
• —Source: Kaggle
• —Link: https://www.kaggle.com/datasets/stucom/solar-energy-power-generation-dataset
• —File: data/spg.csv
• —Records: 4,213 entries
• —Features: 21 columns including:
• —Weather indicators: temperature, humidity, pressure, precipitation, cloud cover
• —Solar radiation: shortwave radiation backwards surface
• —Wind data: speed and direction at multiple altitudes (10m, 80m, 900mb)
• —Solar geometry: angle of incidence, zenith, azimuth
• —Target variable: generated_power_kw

Tech Stack & Tools

Core Technologies

No paid APIs are used in this project.

Project Setup Guide

1️⃣ Clone the Repository

git clone https://github.com/Yashsingh045/Intelligent-Solar-Energy-Generation-Forecasting.git
cd Intelligent-Solar-Energy-Generation-Forecasting

2️⃣ Create Virtual Environment

python -m venv venv
source venv/bin/activate  # Linux / Mac
venv\Scripts\activate     # Windows

3️⃣ Install Dependencies

pip install -r requirements.txt

Key Dependencies:

• —streamlit - Web application framework
• —pandas - Data manipulation and analysis
• —matplotlib - Data visualization
• —numpy - Numerical computing
• —seaborn - Statistical data visualization
• —scikit-learn - Machine learning library
• —jupyter - Interactive notebook environment

4️⃣ Run the Application (Local)

streamlit run app.py

The application will open in your browser at http://localhost:8501

5️⃣ Explore the Data (Optional)

jupyter notebook notebooks/data.ipynb

This notebook contains exploratory data analysis including:

• —Data quality checks
• —Correlation analysis
• —Feature visualization
• —Statistical summaries

How to Use the Application

1. 1.Launch the App

   streamlit run app.py

1. 1.Upload Dataset
2. 2.Click "Browse files" or drag and drop your CSV file
3. 3.Maximum file size: 10MB
4. 4.Supported format: CSV with UTF-8 encoding

1. 1.Explore Your Data
2. 2.View dataset preview (first 20 rows)
3. 3.Check dataset shape and column names
4. 4.Select any numeric column from the dropdown
5. 5.View interactive line plots

1. 1.Analyze Patterns
2. 2.Use the notebook for deeper analysis
3. 3.Examine correlations between features
4. 4.Identify key predictors of power generation

Deployment

The application is deployed using Hugging Face Spaces (Streamlit):

• —Free-tier hosting
• —Publicly accessible URL
• —Automatic build from repository

Deploying to Hugging Face Spaces

1. Handle the Large Model File with Git LFS

The trained model (models/random_forest_model.joblib, ~58MB) exceeds GitHub/HF's 25MB file limit. Use Git LFS:

# Install Git LFS (once)
git lfs install

# Track the large file
git lfs track "models/random_forest_model.joblib"

# Commit the .gitattributes file
git add .gitattributes
git add models/random_forest_model.joblib
git commit -m "Track large model file with Git LFS"
git push

2. Set the GROQAPIKEY Secret

The AI Assistant tab requires a Groq API key (free at console.groq.com):

1. 1.Go to your HF Space → Settings → Variables and secrets
2. 2.Click New secret
3. 3.Name: GROQ_API_KEY, Value: your key from Groq console
4. 4.Save — the Space will restart automatically

3. Push to HF Spaces

# Add HF Spaces as a remote (replace <your-username> and <space-name>)
git remote add space https://huggingface.co/spaces/<your-username>/<space-name>
git push space main

Future Improvements

• —Integration of real-time weather data (free APIs)
• —Advanced deep learning models (LSTM, Temporal CNNs)
• —Enhanced uncertainty quantification
• —Real-world grid simulation scenarios

✨ This project demonstrates the intersection of machine learning, renewable energy analytics, and agentic AI systems to support sustainable power grid operations.