PORTALE DELLA DIDATTICA

Ricerca CERCA
  KEYWORD

Predictive Analysis of Thermal Load in Power Plants using Machine Learning

azienda Tesi esterna in azienda    


Parole chiave ENERGY FORECAST, MACHINE LEARNING, DEEP LEARNING, OPTIMIZATION

Riferimenti GIUSEPPE RIZZO

Riferimenti esterni fabio.esposito@energonesco.it, fabrizio.dominici@linksfoundation.com

Tipo tesi APPLIED RESEARCH

Descrizione Background: Energon is an Energy Service Company (ESCo) and it manages numerous thermal refrigeration and cogeneration/trigeneration power plants across Italy. As part of a developmental project, Energon installed a data acquisition system in some power plants, collecting data via MeterBUS or Modbus protocols. This data, which includes thermal loads and energy consumption metrics, is then centralized for billing efficiency, maintenance analysis, and optimization algorithm development.

Objective: This thesis aims to predict the thermal load of a power plant, in particular the Trezzano plant, using the collected data. It explores whether it's feasible to accurately forecast the thermal load over future time intervals, and if so, the best models for this purpose. The study focuses on analyzing a large dataset comprising various measures like thermal load on different circuits, energy consumption, and timestamps. The focus is on determining which variables impact the thermal load and how to effectively forecast it using heuristic optimization algorithms. The thesis will also explore the integration of external data, such as weather or irradiance, to enhance prediction accuracy. The irregularity in data sampling intervals and the need to fill in missing values present unique challenges. The thesis will address these by rounding timestamps and possibly interpolating data to achieve uniformity and reliability.

Company: There is the possibility to do a thesis together with a curricular internship. Full remote

Conoscenze richieste Basics of Statistics, Data mining
Knowledge of Machine learning and Deep learning, Python language, Relational and NoSQL databases


Scadenza validita proposta 31/12/2024      PROPONI LA TUA CANDIDATURA