Fuzzy Clustering-Based Deep Learning for Short-Term Load Forecasting in Power Grid Systems Using Time-Varying and...

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Data Science

Background:

• Accurate short-term load forecasting (STLF) is crucial for power grid reliability and efficiency.
• Deep neural networks (DNNs) show promise for STLF due to their ability to model complex time-series data.
• Existing DNNs for STLF primarily utilize time-varying features, neglecting valuable time-invariant user characteristics.

Purpose of the Study:

• To propose a novel fuzzy clustering-based DNN for enhanced STLF.
• To integrate both time-varying and time-invariant user features for improved forecasting accuracy.
• To develop a simpler and more effective DNN model by leveraging fuzzy clustering.

Main Methods:

• A fuzzy clustering algorithm is employed to group users based on similar time-invariant features (e.g., building characteristics).
• Deep neural network (DNN) models are subsequently developed for each cluster, focusing on time-varying features.
• The proposed model combines fuzzy clustering with DNNs to perform STLF using both feature types.

Main Results:

• The fuzzy clustering-based DNN demonstrated superior performance in STLF compared to standard DNNs.
• The integration of time-invariant features through fuzzy clustering led to more accurate load predictions.
• The proposed method outperformed commonly used models like Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNNs).

Conclusions:

• Integrating time-invariant user features significantly enhances STLF accuracy.
• Fuzzy clustering provides an effective mechanism to incorporate these features, simplifying DNN models.
• The proposed approach offers a more effective and accurate solution for short-term load forecasting in power systems.