Energy Efficiency Enhancement using Neural Networks

This study looked into assessing the heating load and cooling load requirements of buildings (that is, energy efficiency) as a function of building parameters. The project was done as part of Machine Learning class at the University of Texas at Dallas. The entire summary of the project can be found in the Jupyter Notebook

Table of contents

• General info
• Technologies and Tools
• Setup
• Process
• Code Examples
• Features
• Status
• Contact

General info

The goal of the study is to understand various machine learning classification algorithms and their performance. We have added the cooling load and heating load, which can define the overall load of the apartment. We have studied the trend of the overall load by dividing it into three classes.

Technologies and Tools

• Python 3.5
• TensorFlow

Setup

The dataset used and its metadata can be found in the dataset. The jupyter notebook can be downloaded here and can be used to reproduce the result. Installation of TensorFlow would be required to run all the models. You can find the instructions to install TensorFlow in installation guide.

Process

• We have added the cooling load and heating load which will define the overall load of the apartment.
• We studied the trend of overall load and divided it into three classes, low efficient, high efficient and average efficient.
• We first tried multioutput machine learning regressor models like KNN, logistic regression, random forest, decision tree, linear & kernelized SVM.
• After that we created a neural network classification model to further increase the accuracy of the classification to 97%.

Code Examples

Some examples of usage:

# Linear SVM

param_grid = {"C": [1e0, 1e1, 1e2, 1e3, 1e4],"gamma": np.logspace(-2, 1, 2, 3, 5)}
grid_search_svm = MultiOutputRegressor(GridSearchCV(SVR(kernel='linear'), param_grid, cv=10,return_train_score=True))

grid_search_svm.fit(X_train, y_train)
train_r2_score=r2_score(y_train,grid_search_svm.predict(X_train))
test_r2_score=r2_score(y_test,grid_search_svm.predict(X_test))
output = output.append(pd.Series({'model':'SVM Linear', 'train_r2_score':train_r2_score,'test_r2_score':test_r2_score}),ignore_index=True )
output

# Bagging ensembler using Decision Tree Regressor as base model

from sklearn.ensemble import BaggingRegressor


param_grid = {'max_samples':[5,10], 'max_features':[1,2,3,4,5,6,7]}
bagging_DT = MultiOutputRegressor(GridSearchCV(BaggingRegressor(DecisionTreeRegressor(),n_estimators=750), 
                                                          param_grid,cv= 10,return_train_score=True))

bagging_DT.fit(X_train, y_train)
train_r2_score=r2_score(y_train,bagging_DT.predict(X_train))
test_r2_score=r2_score(y_test,bagging_DT.predict(X_test))
output = output.append(pd.Series({'model':'Multi Output DT Bagging', 'train_r2_score':train_r2_score,'test_r2_score':test_r2_score}),ignore_index=True )
output

#Ada Boosting on Linear SVM regressor¶

param =  { "n_estimators": [100,500,1000] }

base_svr=SVR(kernel='linear')
ada_svr = AdaBoostRegressor(base_estimator=base_svr,learning_rate = 0.7,random_state=10)
adaboost_svr = MultiOutputRegressor(GridSearchCV(ada_svr,param_grid=param,n_jobs=-1),n_jobs=-1)
adaboost_svr.fit(X_train, y_train)
train_r2_score=r2_score(y_train,adaboost_svr.predict(X_train))
test_r2_score=r2_score(y_test,adaboost_svr.predict(X_test))
output = output.append(pd.Series({'model':'Adaboost_LinearSVM', 'train_r2_score':train_r2_score,'test_r2_score':test_r2_score}),ignore_index=True )
output

# Neural Network Classification model

param_grid = {'epochs':[50, 200] , 'batch_size':[10, 50, 100]}

model = KerasClassifier(build_fn = model_classifier , verbose = 0)

grid_search_Keras_Class = GridSearchCV(model , param_grid , cv =10)

grid_search_Keras_Class.fit(X_train_class, y_train)

print('Best parameters for efficiency classification {}'.format(grid_search_Keras_Class.best_params_))

from sklearn.metrics import accuracy_score
print('The Train Accuracy score is',accuracy_score(y_train_class, grid_search_Keras_Class.predict(X_train_class)))
print('The Test Accuracy score is',accuracy_score(y_test_class, grid_search_Keras_Class.predict(X_test_class)))

Features

The training and testing scores af various models are as listed below:

sr no	model	train_r2_score	test_r2_score
0	Linear Regressor	0.902539	0.899354
1	KNN Regressor	0.935276	0.904540
2	Random Forest Regressor	0.997856	0.984413
3	SVM Linear	0.897939	0.896891
4	SVM RBF	0.992589	0.986182
5	Multi Output DT Bagging	0.875499	0.880045
6	Adaboost_DecisionTree	0.999908	0.983499
7	KNN Adaboost	0.972943	0.929250
8	Adaboost_LinearSVM	0.896437	0.893522
9	Gradient Boosting Regressor	0.999541	0.996787
10	Logistic Reression	0.875	0.880208
11	Linear SVC	0.875	0.880208
12	SVM rbf	0.970486	0.984375
13	Random Forest Classifier	1.0	0.96875
14	Gradient Boosting Classifier	1.0	0.979166
15	Neural Network Regression model	0.812056	0.799832
16	Neural Network Classification model	0.963541	0.963541

Status

Project is: finished

Contact

Created by me and my teammate Siddharth Oza.

If you loved what you read here and feel like we can collaborate to produce some exciting stuff, or if you just want to shoot a question, please feel free to connect with me on email, LinkedIn, or Twitter. My other projects can be found here.