Tutorial T3 (Full day)
Computational Intelligence in Power System Applications
Organizers:
- Kalyanmoy Deb, Michigan State University, kdeb@egr.msu.edu
- Marco Mussetta, Politecnico di Milano, marco.mussetta@polimi.it
- Emanuele Ogliari, Politecnico di Milano, emanuelegiovanni.ogliari@polimi.it
Abstract:
Module A: Machine Learning for power forecasting
The variability of renewable energy represents a huge challenge in the integrated electricity systems: power production forecasts can help reducing the amount of operating reserves needed for the system, finally reducing the balancing costs. While physical predition methods strongly rely on the accuracy of the weather forecast, Artificial Neural Networks are based on the learning process of the underlying models and are commonly referred to as a “data-driven” or “black box” approaches. In fact, they need historical data that, after being collected, are used to infer a general trend and behavior in order to predict future output of the power plant. Hybrid methods, consisting in any combination of the physical-based approach and Machine Learning can guarantees the highest level of accuracy when adopted to the power forecast of RES.
Module B: Evolutionary Multi-Criterion Optimization with Case Studies on Power Dispatch Problem Solving
Evolutionary optimization methods, proposed in early sixties and used in practice since eighties, are population-based algorithms which are easily customizable to suit different problem-solving tasks. Evolutionary multi-criterion optimization (EMO) algorithms, proposed since early nineties, revolutionized the solution of problems having multiple conflicting objectives. Starting with two and three-objective problems, EMO researchers have devised algorithms for solving up to 15-objective problems and applied to many engineering and practical problems. In this tutorial, we shall present a step by step account of the growth of EMO field by describing the principles of multi-criterion optimization, some key algorithms, and recent advances in the field. Case studies on power dispatch problem for single and multiple criteria aspects and its static and dynamic versions will be presented.
Room:
See Tutorials program page.
Registration:
See Registration page.
Structure:
| 08:30-09:00 | Registration |
| 09:00-10:30 | Module A: Machine Learning techniques for power forecasting (Marco Mussetta) |
| 10:30-10:45 | Coffee break |
| 10:45-12:15 | Module A: Hybrid methods for power forecasting (Emanuele Ogliari) |
| 12:15-12:30 | Questions and Answers |
| 12:30-13:30 | Light lunch |
| 13:30-15:00 | Module B: Evolutionary Multi-Criterion Optimization with Case Studies on Power Dispatch Problem Solving – part 1 (Kalyanmoy Deb) |
| 15:00-15:15 | Coffee break |
| 15:15-16:45 | Module B: Evolutionary Multi-Criterion Optimization with Case Studies on Power Dispatch Problem Solving – part 2 (Kalyanmoy Deb) |
| 16:45-17:00 | Questions and Answers |
Short Biographies:
Kalyanmoy Deb
Marco Mussetta
