Preprocessing, feature extraction and AI-based real-time energy consumption prediction at the edge or in the fog.

Members:	Estrada Pico Rebeca Leonor, Torres Morán Danny Alfredo, Aizaga Villón Xavier Fransico, Farinango Salazar Pedro.
Date:	1 year 6 months
Financing:	Espol

What is it about?

This project focuses on the implementation of a sensor network based on a microcontroller and an energy measurement module, using the MQTT communication protocol for data transmission. Each sensor is designed to measure voltage, current, power, frequency, energy and the power factor of a specific device. Specifically, measurements are carried out on workstations that share similarities with the servers of a Datacenter located in the Information Technology Center (CTI).

A fundamental part of this project involves the evaluation of different regression models, with the objective of selecting the most appropriate one to predict energy consumption from various measurements collected from the workstations over hours of continuous operation. This system allows the optimization and reduction of maintenance costs of computer equipment due to the implementation of monitoring and management of energy resources, contributing to improving efficiency in Datacenter environments, which implies a positive impact both economically and environmentally.

General objective:

Evaluate preprocessing algorithms, feature extraction and regression models, to make real-time predictions of energy consumption, using a sensor network (iot) and fog computing concepts.

Project objectives:

Objective 1: Assemble six single-phase energy consumption metering devices to monitor six workstations at the CTI-ESPOL facilities.
Objective 2: Measure and store energy consumption data from eight workstations, to generate a database that allows training regression models.
Objective 3: Establish the preprocessing and feature extraction algorithms in real time that must be executed on the gateway.
Objective 4: Identify the most appropriate trained models to predict real-time energy consumption from the gateway.

Project methodology:

Design and assembly of energy consumption metering equipment.
Measurement and data storage.
Identification of appropriate regression models.
Integration of algorithms in the IoT-based sensor network.
Validation and optimization of the prediction model.