Modified Quantum Long-Short Term Memory with Variational Quantum Circuits for PV Power Forecasting

Table of Contents

• Modified Quantum Long-Short Term Memory with Variational Quantum Circuits for PV Power Forecasting
  • Introduction
  • Requirements
  • Configuration
  • Train
  • Logging
  • Demo App
  • License
  • Contact

Introduction

This repository presents a hybrid AI model that integrates Quantum Machine Learning (QML) and Deep Learning to deliver fast, cost-efficient, and highly accurate photovoltaic (PV) power forecasting. By embedding Variational Quantum Circuits (VQCs) within a modified Long Short-Term Memory (LSTM) framework, we propose a model that not only improves prediction accuracy but also significantly reduces computational resources and time compared to classical ML and deep learning counterparts.

Requirements

Project is tested with Python >= 3.10

pip install -r requirements.txt

Configuration

Most of the configuration for this project can be changed in /qlstm/configs/:

• train.yaml for training
• callbacks.yaml for logging and training behavior

To change training model:

1. Use or create new model class in qlstm/models
2. Import and define your model at line 42 in qlstm/train.py.
3. Change model config in qlstm/configs/train.yaml

For datasets with difference structure:

1. Create a custom data module at qlstm/modules/data.py inherited from CustomDataModule class.
2. Import and define your custom data module at line 39 in qlstm/train.py.
3. Change data config in qlstm/configs/train.yaml

Train

Modify the configurations in qlstm/configs/train.yaml then run:

python qlstm/train.py

With CLI:

python qlstm/train.py -h

Example:

python qlstm/train.py trainer.accelerator=cpu optimizer=0.001

Logging

Note: Use full path

tensorboard --logdir /home/USER/.../QLSTM/lightning_logs

Demo App

Built with Gradio.

Launch local:

python app/app.py

Launch public:

python app/app.py --share
# or
python app/app.py -s

License

This project is licensed under the MIT License. See LICENSE for more details.

Contact

Open an issue: New issue

Mail: pthung7102002@gmail.com

---