Table 16 MPPT algorithms computational complexity Comparison.
From: MPPT algorithms for grid-connected solar systems including deep learning approaches
Algorithm | Typical Hardware | Inference Latency (µs) | Memory Usage | Real-Time Deployability |
|---|---|---|---|---|
P&O | Low-end MCU (8–16 bit) | < 10 | Very Low (KB-level) | High |
INC | Low-end MCU (16–32 bit) | < 15 | Very Low (KB-level) | High |
PSO | 32-bit MCU/DSP | 40–150 * | Medium | Medium |
GWO | 32-bit MCU/DSP | 35–140 * | Medium | Medium |
FL | 32-bit MCU/DSP | 80–300** | Medium | Medium |
ANN | DSP/ARM Cortex-M4/M7 | 40–150 | High (tens of KB) | Low-Medium |
LSTM | ARM Cortex-A/Embedded GPU | 200–800 | High (hundreds of KB-MB) | Low |
BiLSTM | ARM Cortex-A/Embedded GPU | 400–1500 | High (MB-level) | Low |
ANFIS | DSP/ARM Cortex-M7 | 100–350 | High (MB-level) | Low |
