Full text
| |
| Author(s): |
Bruno Pissinatto Cancian
Total Authors: 1
|
| Document type: | Master's Dissertation |
| Press: | Campinas, SP. |
| Institution: | Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Elétrica e de Computação |
| Defense date: | 2024-03-21 |
| Examining board members: |
Walmir de Freitas Filho;
João Paulo Abreu Vieira;
Madson Cortes de Almeida
|
| Advisor: | José Carlos Garcia Andrade; Walmir de Freitas Filho |
| Abstract | |
In electric energy compensation systems (net metering), in which a single meter is utilized per customer, i.e., there is no separated meters for load and generation to reduce the costs (e.g., in Brazil), the information given to utilities is the net power curve of each customer. This type of generation connection is known as behind-the-meter. Therefore, there are difficulties to determine the installed capacity of distributed generation and the customer power consumption separately, turning tasks as system expansion planning, power demand prediction, losses analysis (technical and non-technical) and even the viability of new distributed generators more complex. Additionally, errors can occur in the utility database, since the large amount of generation connection requisitions can lead to mistakes during the data registration process, e.g., incorrect values of distributed generators capacities. From the customer perspective, load and generation disaggregation enables non-intrusive monitoring of the generator performance, facilitating, in the case of photovoltaic (PV) systems, the detection of internal defects, the necessity to clean the solar panels, non-predicted shading and even the inoperancy of the whole PV system. With that in mind, three energy disaggregation methodologies for BTM systems are developed in this MSc project. Each methodology considers different scenarios of data availability, making use of techniques as simple as linear regression or more intricate ones, such as generative adversarial networks. The first presented method detects and estimates the installed capacity of BTM PVs systems using native demand (load) and PV generation historical data from observable customers, i.e., with a dedicated meter to the PV, being able to detect PVs systems with accuracy, recall and precision values higher than 90%, and to estimate the installed capacity of these same systems with percentual absolute error smaller than 11,5%. The second developed method disaggregates BTM PV generation using curves of nearby observable PVs systems, reaching median values of daily Normalized Mean Average Error (nMAE) and energy error of 7% and 5%, respectively. Finally, the third method executes energy disaggregation using only net demand curves and low-resolution meteorological data, reaching median values of daily nMAE and energy error lower than 11.5% (AU) | |
| FAPESP's process: | 22/11692-0 - Methodologies for load and generation disaggregation in distribution systems through the utilization of smart meters |
| Grantee: | Bruno Pissinatto Cancian |
| Support Opportunities: | Scholarships in Brazil - Master |