A hybrid ANFIS-transformer framework tuned by enhanced HawkFish optimization for voltage and load balancing in smart grids

Abstract

This paper presents a novel hybrid framework that integrates an Adaptive Neuro-Fuzzy Inference System (ANFIS) with a Transformer model, optimized through an Enhanced HawkFish Optimization Algorithm (EHFOA), to enhance voltage regulation and load balancing in smart grid environments. The proposed system leverages the temporal modeling capabilities of Transformers for accurate load and voltage prediction, while ANFIS enables adaptive, rule-based control in dynamic operating conditions. EHFOA, incorporating strategies such as Lévy flight, energy-aware movement, and elite memory, is designed to fine-tune the hyperparameters of both ANFIS and the Transformer for optimal performance. Simulation results using a real-time load monitoring dataset from Kaggle show that the proposed method significantly outperforms traditional ANFIS-only and Transformer-only models. In simulation on a 5 kW PV-battery system, the proposed model achieved a voltage-forecasting RMSE of 1.24 V and MAE of 0.96 V (1.50% MAPE), and a load-forecasting RMSE of 4.15 kW and MAE of 3.22 kW (3.10% MAPE), outperforming standalone Transformer (2.65 V RMSE, 8.12 kW RMSE) and ANFIS (3.43 V RMSE, 9.65 kW RMSE) benchmarks. In grid-control tests, the hybrid controller reduced system energy loss to 3.10%, a 54.4% improvement over the Transformer-only case (6.80% loss). EHFOA delivered rapid convergence—best fitness 0.0094 in 58 iterations—with a runtime of 46.3 s and a final RMSE of 1.24 V, surpassing GA, PSO, and GWO optimizers. These results demonstrate the framework’s ability to deliver high-accuracy forecasting, significant energy-loss reduction, and efficient optimization for next-generation smart-grid deployment.