sensor_impact_FDD

This is an NREL public repository used for sensor impact evaluation and verification project funded by DOE. Specifically, the function of repository can evaluate the sensor impact, including sensor accuracy and sensor selection, on fault detection and diagnostics (FDD) performance.

The run of the modules in this repo relies on the fault building simulation data that calibrated on Oak Ridge National Laboratory's Flexible Research Platform. Check the Data section for download information.

This project is a work-in-progress and is temporarily in a personal repository, which in the future will be moved to https://github.com/NREL/sensor_impact_FDD after the account is set up by NREL's Information Technology Support team.

Authors: Liang Zhang, Matt Leach, National Renewable Energy Laboratory, January 18, 2021

Installation

Download and install the latest version of Conda (version 4.4 or above) Create a new conda environment:

$ conda create -n <name-of-repository> python=3.8 pip

$ conda activate <name-of-repository>

(If you’re using a version of conda older than 4.4, you may need to instead use source activate .)

Make sure you are using the latest version of pip:

$ pip install --upgrade pip

Install the environment needed for this repository:

$ pip install -e .[dev]

Data

This module is developed based on the fault building simulation data that calibrated on Oak Ridge National Laboratory's Flexible Research Platform. The data include 22 fault types data under seven weathers from different climate zones. The details of physics-based modeling of the faulty buildings are introduced in this paper.

Downloading data is mandatory to use the classes in the repo,

The data can be downloaded here. Please contact Liang.Zhang@nrel.gov if there is any download issues.

Modules

The base.py and SIE.py provide modules to realize sensor impact evaluations. This repo contains three sub-classes that realize three modules used for sensor impact evaluation in FDD.

Module 1: Sensor Accuracy Impact on FDD Performance

Sub-Class Name: sensor_accuracy_impact_FDD or SAIF

Module 2: Sensor Selection and Impact on FDD Performance

Sub-Class Name: sensor_selection_impact_FDD or SSIF

Module 3: Sensor Accuracy Impact on Sensor Selection and FDD Performance

Sub-Class Name: sensor_accuracy_impact_sensor_selection or SAISS

Machine Learning Algorithms

Linear Models

Classifier using Ridge regression

Logistic regression

Stochastic Gradient Descent - SGD

Passive Aggressive Algorithms

Perceptron

Linear and Quadratic Discriminant Analysis

Linear Discriminant Analysis

Quadratic Discriminant Analysis

Support Vector Machines

C-Support Vector Classification

Nu-Support Vector Classification

Linear Support Vector Classification

Nearest Neighbors

Classifier implementing the k-nearest neighbors vote

Nearest centroid classifier

Gaussian process classification (GPC) based on Laplace approximation

Naive Bayes

Gaussian Naive Bayes

Naive Bayes classifier for multivariate Bernoulli models

Naive Bayes classifier for multinomial models

Decision Trees

Decision Tree Classifier

Ensemble methods

Random Forest Classifier

Extra-Trees Classifier

AdaBoost classifier

Histogram-based Gradient Boosting Classification Tree

Gradient Boosting for classification

Neural Network

Multi-layer Perceptron classifier