Prediction of electrical load demand using combined LHS with ANFIS

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Data Science

Background:

• Accurate load demand prediction is critical for power system management, especially in sensitive sectors like healthcare.
• Traditional methods struggle with the complex, nonlinear patterns in energy consumption data.

Purpose of the Study:

• To enhance the predictive accuracy of electrical load demand using a novel hybrid machine learning methodology.
• To address limitations of existing models, such as overfitting and adaptability to diverse data.

Main Methods:

• A hybrid approach combining Latin Hypercube Sampling (LHS) for stratified input variable sampling with Adaptive Neuro-Fuzzy Inference Systems (ANFIS) was developed.
• The methodology involved simulating energy demand patterns over 1000 iterations and evaluating performance using Mean Squared Error (MSE).
• Comparative analysis included ANFIS alone and ANFIS combined with the Monte Carlo (MC) method.

Main Results:

• The ANFIS-LHS model demonstrated superior predictive performance, achieving higher accuracy and robustness compared to ANFIS alone and ANFIS-MC.
• The proposed method showed a 96.42% improvement in accuracy over the standalone ANFIS model.
• Sensitivity analysis and risk assessment were incorporated, further enhancing predictive capabilities.

Conclusions:

• The ANFIS-LHS hybrid model offers a significant advancement in electrical load demand prediction.
• This methodology effectively overcomes limitations of previous approaches, providing more reliable energy management solutions.
• The findings contribute to more adaptive and accurate energy forecasting in power systems.