All classes Power Quality Disturbance Classification Wavelet Syllabus · Episode 6 of 6

Class Wavelet Quality Power Disturbance Classification #6: Noise Robustness Test & Method Comparison

0 views

Tutorial on noise robustness testing for power quality disturbance classification: SNR, noise augmentation, comparison of SVM, Random Forest, XGBoost. Episode 6, the closing episode of the Wavelet Power Quality Disturbance Classification series.

Software: Python (PyWavelets, scikit-learn, XGBoost)
Level: Advanced
Duration: ± 6 min
Series total: 40 min · 6 videos
Published: October 25, 2026

Episode 6 of 6 · view syllabus

Download PDF (5 PDF pages)

0 downloads

Download Project File· 5 MB Unlock All Premium Classes ▶ Watch the video lesson0 plays

▶ Video Lesson

Watch every step above performed live, with English narration.

0 plays

Class book

Page 1 / 5

Class Wavelet Quality Power Disturbance Classification #6: Noise Robustness Test & Method Comparison: Cover: Class Wavelet Quality Power Disturbance Classification #6: Noise Robustness Test & Method Comparison

Class Wavelet Quality Power Disturbance Classification #6: Noise Robustness Test & Method Comparison: Before you start, steps list, and Step 1

Class Wavelet Quality Power Disturbance Classification #6: Noise Robustness Test & Method Comparison: Step 2: Clean fragile training model

Class Wavelet Quality Power Disturbance Classification #6: Noise Robustness Test & Method Comparison: Step 3: Noise augmentation strength

Class Wavelet Quality Power Disturbance Classification #6: Noise Robustness Test & Method Comparison: Done: Final Result and Further Practice

The end

Cover: Class Wavelet Quality Power Disturbance Classification #6: Noise Robustness Test & Method Comparison

Click a page or use ◀ ▶ to turn · swipe on mobile · to enlarge

Want to go deeper?

Join one of our courses or ask us directly on WhatsApp.

Download PDF· 0 downloads View the full 6-episode syllabus View Power Quality Courses Ask on WhatsApp

Recommended classes for you

Picked from the software, topic, and level of “Class Wavelet Quality Power Disturbance Classification #6: Noise Robustness Test & Method Comparison”.

Distributed Generation Optimization PSO-GWO

6 episodesFree preview

Python · Intermediate–Advanced

Distributed Generation Optimization PSO-GWO

6 video lessons with PDF guides, 1 h in total. Starts with: Radial IEEE 33-Bus Distribution System: Line and Load Data.

1 h 0 0

View class

6 episodesFree preview

Python · Intermediate–Advanced

Microgrid Energy Management MPC

6 video lessons with PDF guides, 54 min in total. Starts with: Microgrid Components & PV, Wind, Load, and Electricity Tariff Profiles.

54 min 0 0

View class

6 episodesFree preview

Python · Intermediate–Advanced

PID & FOPID Metaheuristics

6 video lessons with PDF guides, 52 min in total. Starts with: Control System: Transfer Function, Step Response, and Performance Specifications.

52 min 1 0

View class

10 episodesFree preview

Cycle-Tempo · Beginner–Advanced

Cycle-Tempo

10 video lessons with PDF guides, 1 h 14 min in total. Starts with: Simple Rankine Cycle (Basic Steam Power Plant).

1 h 14 min 52 0

View class

MATLAB: Optimizing PV and Inverter Capacity for the Lowest Hydrogen Cost (LCOH)

Free

MATLAB R2025b · Intermediate

MATLAB: Optimizing PV and Inverter Capacity for the Lowest Hydrogen Cost (LCOH)

A MATLAB class from an empty project: downloading PVGIS data, modeling an on-grid rooftop PV system and inverter hour by hour (8,760 hours), then finding the capacity with the lowest hydrogen cost (LCOH) using Particle Swarm Optimization, validated by an Exhaustive Grid Search, with sensitivity analysis and assumption checks. 19 code files written and explained line by line.

1 h 8 min 12 0

View class

MATLAB: The Water Cycle Algorithm (WCA) for Hybrid Microgrid Optimization

Premium

MATLAB R2025b · Intermediate

MATLAB: The Water Cycle Algorithm (WCA) for Hybrid Microgrid Optimization

A MATLAB class from an empty project: applying the Water Cycle Algorithm to size the solar panels, wind turbines, batteries, and diesel generator of a hybrid microgrid. Includes validation against the theoretical optimum, a 30-run stability test, and the effect of population size. 11 code files written and explained line by line.

27 min 5 0

View class

3 episodesFree preview

LINGO · Beginner

LINGO

3 video lessons with PDF guides, 22 min in total. Starts with: Linear Programming for Beginners.

22 min 13 0

View class

2D Wave Flume Simulation with a Breakwater in Ansys Fluent

Premium

ANSYS 2026 R1 (SpaceClaim, Meshing, Fluent) · Intermediate

2D Wave Flume Simulation with a Breakwater in Ansys Fluent

A 2D wave flume simulation in Ansys Fluent 2026 R1: a 100 m flume with a submerged breakwater, meshing, VOF with Airy waves, a numerical beach, and the water surface profile.

15 min 10 0

View class

Need more than a video?

Other ways we can help with Python.

Learn with a mentor

Scheduled Simulation & Engineering courses with live guidance and a certificate.

Course

Course/Training Ansys ~ Heat Transfer and Heat Flux Analysis

★ 5.0 (10)

Course

Course/Training MATLAB — MATLAB App Design: The Ultimate Guide to MATLAB Apps

★ 5.0 (10)

Course

Course/Training Ansys — Airflow Simulation of a Car Radiator with ANSYS

★ 5.0 (9)

See all in Simulation & Engineering →

Have it built for you

No time to learn Python? Our team can do the project for you.

Service

Matlab Development Service — Complete Neural Signal Processing and Analysis Zero to Hero

★ 5.0 (59)

Service

Ansys Maxwell Development Service

★ 5.0 (50)

Service

Petrel Development Service

★ 5.0 (14)

See all in Simulation & Engineering →

Get notified about new Python classes

At most one email a week. Unfollow anytime with one click.

Transcript

The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.

Show transcript

Introduction

Here is the final result, the model with 99% accuracy on clean signals actually drops below 50% when the signal is noisy. A simple trick during training brings the accuracy back to around 90%. Let's get started. Hello, welcome to the final episode of the Wavelet-based Power Quality Disturbance Classification series. Voltage recordings in the field always contain noise from sensors, converters, and interference. A model tested only on clean signals can be very misleading. Today we are testing the robustness of the three models against noise and looking for ways to make them resilient.

Step 1: Noisy test data & augmentation

Create a test file named underscore noise dot py. We will use the same train and test split as in episode 5. The test data is created in four versions: clean, and with Gaussian noise at SNR 40, 30, and 20dB. The smaller the SNR, the higher the noise level, 20dB means noise power is 1 per 100 signal power. The training data is also created in two versions: clean, and with random noise between 20-50dB. This is NICE augmentation.

Step 2: Fragile clean model training

We will train the three models with both versions of the training data, then test them on all four versions of the test data. Run it, feature calculation and training will take a few seconds. The results are surprising. Models trained on clean signals are very fragile. At SNR 40dB, SVM accuracy drops to 71%, and at 20dB it drops to 23%. XGBoost also only reaches 51%. The cause is that noise increases the energy of high-frequency detail coefficients, so that normal signals appear like transients or NOC.

«— Step 3: Noise augmentation strength

Now look at the models trained with NICE augmentation. Accuracy at SNR 40 and 30dB is around 97-98%, and at 20dB around 86-88%. XGBoost with augmentation is the most balanced. 99% on clean signal and remains high on noisy signal. The left graph shows the accuracy of the clean trained model which drops sharply, and the right graph shows the model with augmentation which is much more stable.

Done!

Done, and the series is also completed. We have built a complete research on the classification of power quality disturbance. Model for 9 disturbances, dataset according to IEEE-159, in straightaway flat transform, feature extraction of 32 features, classification with SVM, random forest, and XGBoost, up to noise robustness test with augmentation. For further research, try wave-flat denoising before feature extraction, real recorded data, more complex disturbances, or one-dimensional CNN directly from the signal. Thank you for following this series. Currently working on a thesis, dissertation, or research in the field of power quality and signal processing? Cerdas Inovasi Digital helps with machine learning implementation, from generating disturbance signals according to IEEE-159, wave-flat transformation, Feature extraction, up to classification with SVM, random forest, and XGBoost, as well as opening a Python programming course for electrical engineering. Want to master the entire series, from synthetic signals up to noise resistance testing? All 6 episodes, complete with step-by-step PDF guides and source code, are available for premium tutorial subscribers. Subscribe starting at Rp69.000 per month, save 28% with the 12-month package, or pay once Rp999.000 for lifetime access. Want just one class? Buy single classes with lifetime access. Subscribe now on the premium page cerdasinovasidigital.com. Contact us via WhatsApp, or visit cerdasinovasidigital.com. Thank you for watching, see you in the next episode.

Ask the class material

Type a question or keyword; we find the part of this episode that explains it and jump the video there.

My notes

Notes are saved with the current video time and are only visible to you.

Questions & answers

Stuck on a step? Ask here; the instructor and other learners can answer.

No questions yet.

Your instructor

Cerdas Inovasi Digital Team

Software developers, BIM & 3D engineers, and AI practitioners

53 video classes1,296 episodes

We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.

See our clients →

Memuat

▶ Video Lesson

Recommended classes for you

Need more than a video?

Transcript

0:03Introduction

0:45Step 1: Noisy test data & augmentation

2:24Step 2: Fragile clean model training

3:25«— Step 3: Noise augmentation strength

4:40Done!

Ask the class material

My notes

Questions & answers

Your instructor

▶ Video Lesson

Recommended classes for you

Need more than a video?

Transcript

0:03Introduction

0:45Step 1: Noisy test data & augmentation

2:24Step 2: Fragile clean model training

3:25«— Step 3: Noise augmentation strength

4:40Done!

Ask the class material

My notes

Questions & answers

Your instructor

Introduction

Step 1: Noisy test data & augmentation

Step 2: Fragile clean model training

«— Step 3: Noise augmentation strength

Done!

Introduction

Step 1: Noisy test data & augmentation

Step 2: Fragile clean model training

«— Step 3: Noise augmentation strength

Done!