Machine learning has showed some interesting results in power system applications, e.g. in detecting challenging faults, forecasting loads, etc.
Does anybody have any experience with practical implementation of machine learning applications in power system?
Wind power prediction, price forecasting, power system security assessment and power management strategy for PV/battery.
look to link
https://www.redpixie.com/blog/examples-of-machine-learning
Artificial neural networks are widely used for fault detection, classification and even fault location. Additionally, artificial neural networks can effectively be used to detect and classify the high impedance faults that present a special challenge for traditional protection devices.
You can take a look at the following research paper:
Article Approach for identification and classification of HIFs in me...
Also in network optimization and anti-islanding detection, in a AVR control and other controls systems-.
Dear Amin
It goes without saying, a wide variety of applications can be considered for using machine learning in power systems. According to valid sources of information, one of the most popular apps is forecasting the consumption or demand of load, STLF, and of course analysis and modeling the electricity markets. You ought to take a glance at one of my recent works about modeling the electricity price published in IEEE CONFERENCE 2017 :
https://www.researchgate.net/publication/323715832_Using_artificial_neural_networks_and_system_identification_methods_for_electricity_price_modeling?_sg=71d8lLChNDhciJ2rQ4LBfWzbDlhfTuJRel-0rKnHi0CyE2ffcYLaD6jjnfzRFYF0xEPnJxfI3t47w1i64cf009csNZPwkk9RodTfeb-2.G8mxTtUrjTgta1JgKJfpz0JzT7tYNoP_Sb_iQ434aRRnatyTWU9YX--yTBCG_nwBlCwdr0eUQ-bs_0EkzKlIGA
If you would have any questions, I'm glad to respond to you.
It can use inside the applications for lightening detection system which can be of interest for transmission company. Large number of data can be analyzed with such methodology.
I suggest the following book by Dr. @reza arghandeh.
https://www.amazon.com/Big-Data-Application-Power-Systems/dp/0128119683
It has many examples from different aspects of power system and is coauthored by many experts in the field.
I suggest this article for anybody interested in applying machine learning to power system protection. It has all the elements of a machine learning pipeline:
Article Sustainable Deep Learning at Grid Edge for Real-time High Im...
Question is related to practical application. We are dealing with couple of projects related to PMU data analytics where the target application is "health monitoring of equipment" and " event classification"... Both the applications are offline. Online application such as fault detection...I am not sure if the utilities are ready at this point of time...It may take time and proper justification that machine learning is the last resort to protect a line/equipment with highest dependability (Note: protection engineers are traditionally bit conservative).
@ManasJena: Agreed. Real-time power system protection does not seem to add that much value since we already have deterministic approaches that do very well. And that is what protection engineer like. On the other hand, we also have faults such as high impedance fault that are causing problems, and there hasnt been a deterministic solution. Such small issues here and there can be resolved using machine learning.
https://orbi.uliege.be/bitstream/2268/81333/1/Agregation-Thesis-LW.pdf
Dear Amin, we use simple machine learning methods in combination with system identification methods in order to find faults and gradual malfunction in LVDC grids. See e.g. Conference Paper Refined Fault Detection in LVDC-Grids with Signal Processing...
Dear Amin,
Please see our latest research article:
Article DA-DCGAN: An Effective Methodology for DC Series Arc Fault D...
Machine learning based methods are increasingly being used for fault diagnosis applications. However, the performance of such detection algorithms will degrade because of variations between the source domain data used during the development and the target domain data encountered in operation in the field (i. e. from PV emulator-based laboratory to real PV systems). Furthermore, the faults data in the target domain for model training are usually not available.
Therefore, we propose a machine learning methodology (DA-DCGAN) based on domain adaptation, convolutional neural networks (CNN), and generative adversarial network (GAN) for DC series arc fault diagnosis in PV systems.
DA-DCGAN firstly learns an intelligent normal-to-arcing transformation from the source-domain data. Then by generating dummy arcing data with the learnt transformation using the normal data from the target domain and employing domain adaptation, a robust and reliable fault diagnosis scheme can be achieved for the target domain without using target-domain fault data. Ultimately, a lightweight CNN is trained and implemented in a embedded controller to achieve real-time series arc fault diagnosis for target-domain.
This methodology might also have potentials for other fields.
The use of machine learning technology in detecting defects, faults, errors in systems, in data processing processes, etc., is a potential large development opportunity for this type of applications of this technology.
An interesting issue. I am interested in continuing discussions in this topic.
- Badges
- Science topic