 ICACT20250070 Slide.01
[Big Slide]
|
Chrome  Click!! |
 |
Main slide |
| ICACT20250070 Slide.02
[Big Slide]
|
Chrome Click!! |
 |
Agenda |
| ICACT20250070 Slide.03
[Big Slide]
|
Chrome Click!! |
 |
Accurate load forecasting is crucial for the optimized operation and planning of smart power grids.
Traditional centralized forecasting models struggle with data privacy concerns, communication overheads, and lack of model adaptiveness.
This paper proposes a federated learning-based framework for short-term load forecasting in smart grids.
Local machine learning models are trained on distributed private datasets across different stations of the grid and only the model parameters are communicated to a central server to create an aggregated global model for load forecasting, without exchanging any raw private data.
|
| ICACT20250070 Slide.04
[Big Slide]
|
Chrome Click!! |
 |
Smart grids, the backbone of today's electricity infrastructure, use advanced metering, control, and coordination to optimize power generation, distribution, and consumption [1-4].
Federated learning has quickly gained pull as a distributed machine learning approach [5-8], promising a secure and scalable solution [9-10].
The use of decentralized edge devices to train models locally eliminates the need to transmit raw private data. [11-14]
|
| ICACT20250070 Slide.05
[Big Slide]
|
Chrome Click!! |
 |
Combining smart infrastructure and federated learning opens up new possibilities for environmentally friendly and consumer-friendly power grids.
Recently, various ML techniques such as RNNs, CNNs, and LSTM networks have been used for load forecasting, capturing spatial and temporal electricity demand dynamics across grid hierarchies [16-17].
|
| ICACT20250070 Slide.06
[Big Slide]
|
Chrome Click!! |
 |
Proposed Model |
| ICACT20250070 Slide.07
[Big Slide]
|
Chrome Click!! |
 |
Federated learning is a way to train AI models without anyone seeing or touching your data.
Federated learning (also known as collaborative learning) is a machine learning technique that trains an algorithm across multiple decentralized edge devices or servers holding local data samples, without exchanging them.
|
| ICACT20250070 Slide.08
[Big Slide]
|
Chrome Click!! |
 |
Graphical Model |
| ICACT20250070 Slide.09
[Big Slide]
|
Chrome Click!! |
 |
Mathematical Model (Client Side)
feedforward, activation function, (i/h layyer)
feedforward, activation function, (H/O layyer)
Error |
| ICACT20250070 Slide.10
[Big Slide]
|
Chrome Click!! |
 |
Mathematical Model (Client Side)
Chain rule |
| ICACT20250070 Slide.11
[Big Slide]
|
Chrome Click!! |
 |
Mathematical Model (Client Side)
Chain rule Change in weights between H/O layer |
| ICACT20250070 Slide.12
[Big Slide]
|
Chrome Click!! |
 |
Mathematical Model (Client Side)
Chain rule (bewteen I/H layer |
| ICACT20250070 Slide.13
[Big Slide]
|
Chrome Click!! |
 |
Final Equaction Mathematical Model (Client Side)
|
| ICACT20250070 Slide.14
[Big Slide]
|
Chrome Click!! |
 |
ith client FML pseudo code. |
| ICACT20250070 Slide.15
[Big Slide]
|
Chrome Click!! |
 |
Mathematical Model (Server Side) LM , BR |
| ICACT20250070 Slide.16
[Big Slide]
|
Chrome Click!! |
 |
Mathematical Model (Server Side) SC,
Case1: If d1=d3=d5 and d2=d4=d6
|
| ICACT20250070 Slide.17
[Big Slide]
|
Chrome Click!! |
 |
Mathematical Model (Server Side) SC,
Case2: If d1≠d3 ≠ d5 and d2 ≠ d4 ≠ d6
|
| ICACT20250070 Slide.18
[Big Slide]
|
Chrome Click!! |
 |
Mathematical Model (Server Side) |
| ICACT20250070 Slide.19
[Big Slide]
|
Chrome Click!! |
 |
FML algorithm pseudo code at Server Side |
| ICACT20250070 Slide.20
[Big Slide]
|
Chrome Click!! |
 |
Graphical Model |
| ICACT20250070 Slide.21
[Big Slide]
|
Chrome Click!! |
 |
Simulation and results |
| ICACT20250070 Slide.22
[Big Slide]
|
Chrome Click!! |
 |
Performance Evaluation of the Proposed System Incorporating a Federated Learning Approach for Load Prediction in a Smart Grid |
| ICACT20250070 Slide.23
[Big Slide]
|
Chrome Click!! |
 |
Comparison of the Proposed System with State-of-the-Art Methods. |
| ICACT20250070 Slide.24
[Big Slide]
|
Chrome Click!! |
 |
Conclusion
This paper presented a decentralized architecture for privacy-preserving electricity load forecasting across smart power grids.
We customize a layered federated learning approach to arrange an integrated predictive model between distributed grid components.
Proposed federated model achieves 98.60 % accuracy and 1.4% miss-rate.
Proposed Federated Learning Approach gives more accurate results as compared to previous published approaches like GA, ANN etc.
|
| ICACT20250070 Slide.25
[Big Slide]
|
Chrome Click!! |
 |
References |
| ICACT20250070 Slide.26
[Big Slide]
|
Chrome Click!! |
 |
References |
| ICACT20250070 Slide.27
[Big Slide]
|
Chrome Click!! |
 |
References |
| ICACT20250070 Slide.28
[Big Slide]
|
Chrome Click!! |
 |
References |
| ICACT20250070 Slide.29
[Big Slide]
|
Chrome Click!! |
 |
References |
| ICACT20250070 Slide.30
[Big Slide]
|
Chrome Click!! |
 |
Thanks |
