Implementation of Solar Tracker Controller Using Artificial Neural Network ECE 539 Course Project By Ray Tang
Motivation Alternative Energy - Solar Power Sunlight Gives Energy and the Quality of Light That No Other Source Can Replace Sunlight Varies its Direction From Day to Day and Moment to Moment Practice "On-line Training/learning" Compete with Sunflower
Project Details Design and Build a Solar Tracker Design and Implement the ANN Assess the Difficulty of the Ann Approach
Solar Tracker Two Axis for Tilt and Rotate Two Maxon Type A Gear Motor 0.2E Resolution 15 cm x 15 cm Square Solar Panel 12VDC at 60mA Max Output Five Photo Sensors as Input Colour Filters to Reduce Noise and Enhance Directional Sensitivity Analogue Output
The ANN Multi-layer Perceptions Using a Configuration BP Training Algorithm Off-line Training with 499 Artificial Data On-line Training with Live Data Learning Rate at 0.3, Momentum at 0.8 Training with Tolerance at 0.1
The Set-up Serial Port Correction STK500 Sensor Reading Analogue to Digital for Sensor Inputs Servo Driver for Positioning Using Sensor Reading as Input Decision Sends Back to the Solar Tracker
Result 24th November Voltage (V) Neural Net Decision Time (x30 sec.) Correction
Future Improvements Improve On-line Training Add Current Sensor to Solar Panel Investigate Other Approaches Fuzzy Logic Control Time Series Prediction
Conclusion "Unsuccessful" On-line Training Acceptable Performance Stable System
Reference Annual Report 24, Climate Monitoring & Diagnostics Laboratory CMDL, Western Power Corporation, Kalbarri Photovoltaic System, Gordon, M.. and Wenger, H., Central-Station Solar Photovoltaic Systems: Field Layout, Tracker, and Array Geometry Sensitivity Studies, Solar Energy, Vol. 46, No. 4, pp , April ey., Panico, P., Garvison, P., Wenger, H., and D. Shugar, Backtracking: A Novel Strategy for Tracking PV Systems, IEEE Photovoltaic Specialists Conference, Las Vegas, NV, October Rogers, Jo (1997) Object - Oriented Neural Networks in C++, Academic Press, Inc., New York.