A novel fuzzy supervisor for online weight synthesis in hybrid parallel control of DFIG-based wind energy conversion systems

Abstract

The Hybrid Parallel Controller (HPC) combining Proportional–Integral (PI), Sliding-Mode (SMC), and Backstepping (BSC) through a convex weighted sum outperforms pure, cascaded, and switched/selector hybrids controllers on DFIG-based wind energy conversion systems (WECS) with offline weights. Online weight tuning could extend these gains further, but no prior fuzzy method applies: Type-1, Interval Type-2 (IT2), and ANFIS produce scalar outputs, whereas HPC weight synthesis requires simplex-valued α ∈ Δ 3 , rule consequents on Δ 3 , a defuzzification preserving Σ αᵢ = 1, and closed-loop stability for arbitrary admissible α(t) ∈ Δ 3 . This paper develops a Type 1 Mamdani fuzzy supervisor (HPC-FLC) for online HPC weight synthesis, establishing a mapping Φ: [− 1, 1] 2 → Δ 3 from the tracking error and its rate. The 25-rule base is first established through manual tuning grounded in the complementary roles of PI, SMC, and BSC, and is then verified and refined against the Genetic Algorithm-optimal anchors of (Chahbi et al., 2026a) to enhance performance under different operating conditions. A Uniform Ultimate Boundedness theorem is proved for arbitrary α(t) ∈ Δ 3 , exploiting only the plant’s affine input structure and the convexity of Δ 3 without Lipschitz constraints on Φ. On a 1.5 MW DFIG–WECS at different scenarios, HPC-FLC improves upon HPC-GA by 28–36% in ITAE at higher inertias and reduces ITAE by 98.7–99.8% relative to Type-1 and IT2 primary-loop controllers tuned to their published settings; response time falls to 3–7 ms (70–87% faster than HPC-GA). Processor-in-the-loop (PIL) execution on an STM32 Cortex-M confirms an 18.3 µs control-cycle latency at 10 kHz. These figures are reported for the tested operating envelope and benchmark tunings rather than as unconditional guarantees.