Constraint-Driven Warm-Freeze for Efficient Transfer Learning in Photovoltaic Systems

TL;DR

The paper introduces Constraint-Driven Warm-Freeze (CDWF), a resource-efficient transfer learning framework that dynamically allocates training capacity to critical model components, enabling robust PV cyberattack detection on edge devices.

Contribution

CDWF is a novel, budget-aware adaptation method that combines a warm-start importance assessment with constrained optimization to optimize model fine-tuning under hardware limitations.

Findings

CDWF retains 90-99% of full fine-tuning performance.

It reduces trainable parameters by up to 120x.

Effective for PV cyberattack detection and standard vision benchmarks.

Abstract

Detecting cyberattacks in photovoltaic (PV) monitoring and MPPT control signals requires models that are robust to bias, drift, and transient spikes, yet lightweight enough for resource-constrained edge controllers. While deep learning outperforms traditional physics-based diagnostics and handcrafted features, standard fine-tuning is computationally prohibitive for edge devices. Furthermore, existing Parameter-Efficient Fine-Tuning (PEFT) methods typically apply uniform adaptation or rely on expensive architectural searches, lacking the flexibility to adhere to strict hardware budgets. To bridge this gap, we propose Constraint-Driven Warm-Freeze (CDWF), a budget-aware adaptation framework. CDWF leverages a brief warm-start phase to quantify gradient-based block importance, then solves a constrained optimization problem to dynamically allocate full trainability to high-impact blocks while efficiently adapting the remaining blocks via Low-Rank Adaptation (LoRA). We evaluate CDWF on standard vision benchmarks (CIFAR-10/100) and a novel PV cyberattack dataset, transferring from bias pretraining to drift and spike detection. The experiments demonstrate that CDWF retains 90 to 99% of full fine-tuning performance while reducing trainable parameters by up to 120x. These results establish CDWF as an effective, importance-guided solution for reliable transfer learning under tight edge constraints.