IRJET- A Review on Power Line Carrier Communication (PLCC) Systems
Poster
1. Artificial Intelligence Techniques for Load
Forecasting in the Smart Electrical Grid
Xiaohan Ni
(Dr. Sri Kolla, Advisor)
Electronics and Computer Engineering Technology
The need for efficient usage of
traditional energy resources and
the limited renewable energy
resources has led to the evolution
of the Smart Electrical Grid.
There are many factors that can affect the actual
electrical load, such as climate, special events, and
load timing. It is important to forecast the load based
on the previous loads so that generation can be
properly allocated and scheduled.
Load Forecasting
Data Source
• Bowling Green Utility Data: 2004 – 2015
• Short-term Forecasting:
• a. hourly: 2013 Jan 7 - 13 August 5-11
• b. daily : 2013 Jan 7 – 20 April 1 – 14 August 5 – 18
December 2 - 15
• Long-term Forecasting: Monthly, 2006 – 2015
Load t-1
Load t-2
Load t-3
Load t-4
Output
(Present
Load
Prediction)
A N N or
SVM
Prediction Results for both
SVM and ANN methods
Results
The load prediction results of the ANN method are better
than the SVM method for the short term forecasting of the data
considered. The results of this study may assist Bowling Green
and other electric utilities to effectively forecast the electric load
on the grid and appropriately schedule the generation capacity
using traditional and renewable sources.
Conclusion
X-axis : Time (Hours)
Y-axis : Load (MW)
Prediction Results for
the ANN
X-axis : Time (Hours)
Y-axis : Load (MW)
_ _ _ _ _
Prediction Results
Actual Load
_ _ _ _ _
Prediction Results (ANN)
Actual Load
_ _ _ _ _ Prediction Results (SVM)
The smart grid is a high level concept that infuses
Information and communication technologies with
electrical grid to increase performance and provide
new capabilities
Smart Grid
Abstract
This project applies Artificial Intelligence (AI)
techniques based methods for load forecasting in the smart
electrical grid. For the Bowling Green utility data, this project
successfully used Artificial Neural Networks and Support
Vector Machine methods for the electricity load forecasting.
ANN Training using
MATLAB for Bowling
Green Utility Hourly Data
Prediction Results for the
SVM
X-axis : Time (Hours)
Y-axis : Load (MW)
_ _ _ _ _
Prediction Results
Actual Load
Artificial Neural Networks (ANN)
ANN are mathematical models inspired by biological neural
networks, used to estimate or approximate functions that
depend on a large number of generally unknown inputs . A
feed forward layered neural network model is used for load
forecasting in this project.
Support Vector Machine (SVM)
SVM are statistical learning theory based models with
associated learning algorithms that analyze data and
recognize patterns used for classification and regression
analysis.
(From Abhisek Ukil ISBN 978-3-540-73169-6 Springer)
Oj =Tj (Sj )
Sj = Xi
i=1
n
å Wi
(From Abhisek Ukil ISBN 978-3-540-73169-6 Springer)
(From Home Power Magazine)
(From Hitachi Smart Grid Site)
ε is the insensitive
measure of error