From March 2021 to November 2023, I had the privilege of being a member of the Electrical Engineering Lab at Chongqing University, where I worked on a research project advised by Professor Duo Liu aimed at identifying and predicting meteorologically sensitive loads in power supply station areas. This project involved the development of advanced machine learning models to analyze municipal grid power load data and forecast both long- and short-term power load curves. The experience significantly deepened my understanding of power systems. It reinforced my skills in applying machine learning to real-world energy problems.
Key Achievements
-
Algorithm Development:
- Developed a hybrid model combining Graph Neural Networks (GNN), Support Vector Machines (SVM), and Long Short-Term Memory (LSTM) networks for power load identification and prediction. This model was designed to analyze large-scale municipal grid power load data and predict short-term and long-term power load curves. The hybrid approach allowed for capturing spatial dependencies (via GNN) and temporal patterns (via LSTM), enhancing the model’s accuracy and robustness in predicting meteorologically sensitive loads.
-
Model Training and Optimization:
- Focused on fine-tuning the model’s parameters, ensuring that the hybrid neural network could handle the complexity of the power load data. Using various real-world data, we tested the model’s predictive capability under different conditions, improving its generalizability and accuracy.
-
Research Basics:
- Besides algorithm development, I contributed to producing technical reports and presentations for the project’s mid-term and final defenses. I actively prepared the PPTs for project reviews, ensuring that the presentation communicated our progress and findings to a diverse audience, including academic peers and industry partners.
- I also proofread and helped structure the project’s technical reports’ main framework, content, and format. This experience honed my ability to communicate complex technical concepts clearly and concisely, a vital skill for presenting research results to professional stakeholders.
What I Learned
-
Interdisciplinary Application of Machine Learning:
- This project gave me valuable insights into the intersection of machine learning and power systems. By combining GNN, SVM, and LSTM, I gained hands-on experience applying advanced AI techniques to real-world energy problems. This interdisciplinary approach helped me appreciate the nuances of energy systems and machine learning, particularly how each technique can complement the others in solving complex prediction problems.
-
Power Systems Analysis:
- Working with municipal grid power load data gave me an in-depth understanding of how meteorological factors affect power consumption. This experience highlighted the critical importance of accurate forecasting in energy management, wildly when demand fluctuates significantly due to weather patterns.
-
Technical Communication and Report Writing:
- One of the critical skills I developed during this project was the ability to communicate research findings in a structured and professional manner. Whether preparing technical reports or crafting presentation slides for project defenses, I learned to ensure clarity, precision, and accuracy when presenting complex technical material. These communication skills are essential for collaborating with researchers and presenting findings in academic and industry settings.
Conclusion
This research project at Chongqing University was instrumental in developing my expertise in both machine learning and power systems analysis. The hybrid GNN-SVM-LSTM model we created for predicting meteorologically sensitive loads has the potential to improve the efficiency of power grid management significantly. As I continue my academic journey, I am excited to apply the insights gained from this project to further research in machine learning, power systems, and energy optimization.
Through this experience, I have developed a strong foundation in the interdisciplinary application of machine learning techniques to energy systems. I look forward to exploring more innovative solutions to modern power grids’ challenges.