Luminous TechnoX Hackathon Submission 2024
SolarWise is a comprehensive AI-driven energy management solution that integrates IoT devices, cloud computing, and machine learning to provide real-time monitoring, forecasting, and optimization of energy consumption. Our system focuses on maximizing solar energy utilization while minimizing costs through intelligent scheduling and real-time alerts.
##THE USERNAME & PASSWORD IS luminous
- Solar Battery Levels Monitor
- Power Usage Analytics
- Energy Consumption Predictions
- Monthly Savings Calculator
- Device Status Overview
- Real-time Tariff Rates
- Anomaly Detection Alerts
- Live solar power generation tracking
- Power consumption monitoring
- Connected IoT device status
- Battery level monitoring
- Cost savings calculations
- Time-of-Use (TOU) tariff tracking
-
User Consumption Prediction (Linear Regression)
- R-squared value > 90%
- Lightweight model for efficient computation
- Quick insights generation
-
TOU Tariff Prediction (LSTM)
- 92% accuracy in rate predictions
- Captures long-term dependencies
- Adaptive learning capabilities
-
Anomaly Detection (Z-Score)
- 95% precision rate
- Computationally efficient
- Real-time monitoring capabilities
-
Device Scheduling (MILP)
- 100% optimal scheduling solutions
- Efficient constraint handling
- Cost-effective energy usage planning
-
Data Ingestion & Processing
- Amazon Kinesis for real-time streaming
- EKS-deployed stream processors
- Custom protocol-specific data connectors
-
Storage Solutions
- Amazon S3 for raw data
- ElastiCache for quick access
- DynamoDB for device states
- Aurora for structured data
- OpenSearch for analytics
-
Processing & Analytics
- AWS Athena for serverless queries
- Custom APIs on Amazon EKS
- ML model deployment infrastructure
- Real-time anomaly detection
-
User Interface & Control
- Real-time dashboards
- Push notifications
- Device control interface
- Savings analytics
- Local PostgreSQL for edge processing
- Neon Database for cloud storage
- Grafana for real-time visualization
Model Type | Accuracy | Use Case |
---|---|---|
LSTM | 92% | Consumption Prediction |
Linear Regression | 90%+ | Rate Prediction |
Z-Score | 95%+ | Anomaly Detection |
MILP | 100% | Device Scheduling |
- AWS Account with appropriate permissions
- IoT devices with WiFi capability
- Smart meters compatible with the system
- Internet connectivity
- EC2
- EKS
- Kinesis
- S3
- ElastiCache
- DynamoDB
- Athena
- OpenSearch
- AWS Glue
- Modular Architecture
- Scalable across regions
- Handles large data volumes
- Adaptable to various markets
-
User Flexibility
- Remote monitoring
- Anywhere access
- Real-time control
-
Intelligent Optimization
- Continuous learning
- Adaptive to user behavior
- Market-aware scheduling
-
Robust Infrastructure
- 99.5% uptime
- Scalable architecture
- Secure data handling
To minimize costs associated with real-time compute on AWS SageMaker, we implemented models and pre-computed values for 7 days in advance. This setup emulates fetching these values from Grafana using aggregated database tables while ensuring a realistic user experience.
- TOU Tariff Data: Data generated using the Tariff of Time of Use (ToU) Indian Power System Dataset from Mendeley Data, normalized for commercial rates per unit. The system loops these values to simulate upcoming tariff rates.
- Solar Power Generation & Consumption: Cleaned data provided by Luminous after the initial Q&A session, looped to mimic continuous solar power generation and consumption.
Our ML models were trained using the above split and augmented data, validated against actual user consumption data from InAnalytics. This serves as a robust baseline for real consumer data integration.
We have utilized the Pycarat library for quick prototyping and trial and error on over 20 ML models for each prediction feature
This project is licensed under no License. Please dont reuse this file.
Note: This project was developed as part of the Luminous TechnoX Hackathon 2024