Table 3 Quantitative and qualitative comparison of proposed fuzzy reactive power control with existing reactive-power-based and neural-adaptive control methods.
From: Fuzzy logic-based reactive power control for power factor enhancement in EV drives
Feature/metric | Conventional FOC (vector control) | Neural network adaptive reactive power control | Proposed fuzzy reactive power control (FRC) |
|---|---|---|---|
Need for flux estimator / observer | Required | Required (often neural flux estimator) | Not required |
Computational complexity (operations per cycle) | ~ 8,000–10,000 ops/cycle | ~ 20,000–35,000 ops/cycle (due to NN inference) | ~ 10,000–14,000 ops/cycle (5 × 5 MF fuzzy logic) |
Real-time feasibility on low-cost DSPs | Good | Limited (requires high-end DSP/FPGA/GPU) | Excellent |
Dynamic torque response time | 3.5–4.2 ms | 3–3.8 ms | 3–3.5 ms (comparable or faster) |
Adaptability during EV driving cycles | Moderate (fixed flux) | High (learning-based) | High (reactive power + fuzzy inference) |
Power factor performance under variable loads | 0–5% improvement | 8–12% improvement | 10–20% improvement (validated experimentally) |
Robustness to parameter variations | Moderate (depends on Rr, Ls tuning) | Highly sensitive to model mismatch unless retrained | High (no estimator, rule-based logic) |
Training / tuning requirements | None | Requires dataset generation & online/offline training | None (rule-based) |
Memory requirement | Low | High (NN weight storage) | Low |
Ease of implementation on EV motor drive | Medium | Difficult: requires neural architecture support | Simple and practical |
Overall suitability for real-time EV applications | Good | Limited by hardware constraints | Very high |
