Introduction to Machine Learning with Examples in Power Systems
A six-hour Imperial College Summer School contribution for 42 undergraduate students, combining guest lectures and group-project tutorials.
Summary
This six-hour Summer School contribution combines lectures on machine-learning fundamentals and their applications in power systems with facilitated group-project tutorials. The cohort comprised 42 undergraduate students from South China University of Technology, Hunan University, and Xi’an Jiaotong University. The teaching moves from problem formulation and core methods to model evaluation, time-series foundation models, and the practical role of data.
Course materials
2026 lecture notes
Open PDF
Python tutorial
Open the GitHub repository
Learning aims
- Build an accessible foundation in machine learning for students working on energy and power-system problems.
- Connect model choices to engineering questions, available data, and appropriate performance measures.
- Help project groups turn an initial idea into a defensible machine-learning workflow.
Indicative topics
- Electric power-system background and machine-learning applications in the energy sector
- Supervised learning: linear regression, logistic regression, neural networks, and support vector machines
- Unsupervised learning: principal component analysis and k-means clustering
- Model evaluation, underfitting and overfitting, regularisation, hyperparameters, and validation
- Time-series foundation models, including Amazon Chronos
- Data preparation, data quality, and open energy-system data sources
Group-project tutorials
The tutorials support student teams as they translate a project question into a practical machine-learning workflow. Discussion and feedback focus on:
- defining the engineering question, inputs, and target outputs;
- identifying suitable data and checking its quality;
- selecting a task formulation, baseline, and modelling approach;
- planning validation and interpreting generalisation performance; and
- communicating assumptions, results, and limitations clearly.