کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

Explainable AI-Driven Deep Learning Framework for Short-Term Net Load Forecasting

نویسندگان :

Sina Hossein Beigi Fard¹ ََAmir Hossein Baharvand² Amir Hossein Poursaeed³ Meysam Doostizadeh⁴

1- دانشگاه لرستان 2- دانشگاه لرستان 3- دانشگاه لرستان 4- دانشگاه لرستان

کلمات کلیدی :

short-term net load forecasting،bayesian optimization،explainable artificial intelligence،deep learning،bidirectional long short-term memory،renewable energy resources.

چکیده :

To ensure the effective and sustainable operation of modern energy supply networks, estimating the electricity load is crucial for strategic planning, power generation, and system operation. In order to improve the precision and effectiveness of Short-Term Net Load Forecasting (STNLF), this paper presents an optimized deep neural network. Bayesian optimization of Bidirectional Long Short-Term Memory (BiLSTM) serves as the foundation for the proposed deep learning model. Bayesian optimization method is used to fine-tune the hyperparameters of the BiLSTM model. Moreover, historical time-series data with hourly resolution from 2018 and 2019 Austria are used to demonstrate the effectiveness of the proposed method. This dataset contains variables including temperature, solar energy, wind turbine generation, and actual loads. Additionally, an explainable artificial intelligence technique is used to provide better transparency and alignment with domain knowledge to explain the impact of input factors on the forecasts. The efficacy of the proposed approach is further demonstrated by validation on a different experimental dataset. Interestingly, there is a notable improvement in the evaluation of performance indices compared to the conventional machine learning-based forecasting techniques. This demonstrates the deep learning model's resilience and dependability in improving STNLF accuracy.