Smart Grid Stability

1. Presented By
MURTADHA BAJI EIDAN
ANWAR LIWA’A SHAKER
Electronic and Communication Department
M.Sc. 2022 - 2023
Smart Grid Stability with
Machine Learning
Machine Learning 1 Dept. of ECE, University of Kufa,

2. Smart Grid is an electricity grid network enabling two-way flow of electricity
and data, including smart appliances, renewable and efficient resources.
Smart Grid
Machine Learning 2 Dept. of ECE, University of Kufa,
This system allows for monitoring, analysis, control and communication
within the supply chain to help improve efficiency, reduce energy
consumption and cost.

3. Modelling Grid Stability
As the whole process is time-dependent, dynamically estimating grid
stability becomes not only a concern but a major requirement.
The work focuses on Decentral Smart Grid Control (DSGC) systems,
a methodology strictly tied to monitoring particular property of the grid,
it’s Frequency, hence frequency is time factor
Machine Learning 3 Dept. of ECE, University of Kufa,

4. Addressing simplifications in the model
the original DSGC model was run to generate a set of inputs and outputs
that a 'learning machine' can process and make predictions from. so, The
need of a tool to predict grid stability would have been met, and the
Classification ("stable" versus "unstable") problem must be solved.
In other words, machine learning is used in the following way:
1. A given set of input parameters is fed into the original DSGC model
2. The DSGC model process this Data and returns a binary output - the
grid stability for that particular set of inputs ('stable' or 'unstable' - a
binary classification!);
3. Using classifiers to fit and predict the data, the select the best Model
4. Improving the accuracy of the selected model on the data
Machine Learning 4 Dept. of ECE, University of Kufa,

5. The Data
The Dataset chosen for this machine learning has a synthetic nature and
contains results from simulations of grid stability for a reference
4-node star network imported from “ UCI Machine Learning Repository “
https://archive.ics.uci.edu/ml/datasets/Electrical+Grid+Stability+Simulated+Data+#
Machine Learning 5 Dept. of ECE, University of Kufa,

6. The Data
independent and Dependent Variables:
Machine Learning 6 Dept. of ECE, University of Kufa,

7. The Data Representation
Machine Learning 7 Dept. of ECE, University of Kufa,
The data contains 14 column, 12 for the features, and 2 for the response. And it
have 60,000 dataset as represented.

8. Data Analysis
Machine Learning 8 Dept. of ECE, University of Kufa,
It is important to verify from correlation graph that the data has fewer
correlative features that can’t be dimensionally reductive.

9. Machine Learning Model
Selection
Machine Learning 9 Dept. of ECE, University of Kufa,
Data is well behaved and in general uniformly distributed, we will use
two classifiers to fit the smart grid stability data and then predicting
these data using the two classifiers, the using ROC metric to show the
performance of these classifiers on the data.
1- Using ROC Score for implementing Support Vector Machine (SVM)
Classifier Predicting Accurcy:
ROC_Curve = roc_auc_score(y_test, y_svm) = 0.7854264760130516
2- Using ROC Score for implementing Random Forest Classifier
predicting Accurcy:
ROC_Curve = roc_auc_score(y_test, y_forest) = 0.9383804622719695
From the ROC metric, Random Forest classifier perform well in about
93%
for the specific data.
So, Random Forest classifier perform well for the smart stability data
and we

10. Machine Learning
Implementation
Machine Learning 10 Dept. of ECE, University of Kufa,
For implementing the performance of the model, we use the Confusion
matrix
to describe the data predicted from Random Forest Classifier

11. Future directions for this work
The future directions for this work focused on modelling a machine
learning approach enhance flexibility of dealing with any new set of data
that affect the grid stability, also using advanced Machine learning
Algorithms to provide more accuracy and give us a good prediction of
future demand process with Smart grid systems
Machine Learning 11 Dept. of ECE, University of Kufa,

12. Any Question ?