In islanded low-voltage microgrids, the parallel operation of inverters using traditional droop control strategies often results in imbalanced output impedances among inverters due to variations in line impedance. This imbalance prevents the equitable distribution of reactive power according to the designated droop coefficients. To address this challenge, this paper proposes an improved droop control strategy. First, the virtual impedance control method is added to the strategy of traditional droop control, and then an adaptive weight particle swarm optimization algorithm (APSO) is introduced in the voltage outer loop control module to optimize the parameter setting of the PID controller. Finally, a comparative analysis was conducted on the output results of the improved droop control before and after introducing the algorithm. This proposed strategy is systematically analyzed through modeling and simulation in the MATLAB/Simulink environment to assess its impact on output voltage stability, reactive power sharing, and circulating current magnitude among parallel inverters. The simulation results indicate that the improved droop control with the introduction of virtual impedance effectively achieves equal reactive power sharing among parallel inverters. Moreover, by employing the APSO algorithm, the stability of the inverter output voltage is significantly enhanced. The maximum voltage deviation between two inverters is reduced to below 0.03%, while the maximum reactive power deviation decreases from over 3.5% to less than 2.5%. These improvements effectively reduce the circulating currents between two paralleled inverters, ensuring that the maximum current remains below 1.5 A and the minimum stays above −1 A.

In the parallel operation of islanded microgrids, line impedance mismatches and random load fluctuations, along with the dynamic response lag and difficulty in multidimensional parameter tuning of traditional control strategies, lead to power sharing imbalances and instability in frequency and voltage. To address these issues, this paper proposes a hierarchical cooperative control strategy based on consistency that integrates load-observation-based dynamic reference feedforward (LODRF) and adaptive particle swarm optimization (APSO). First, an improved adaptive virtual impedance (IAVI) strategy based on consistency is introduced into the virtual synchronous generator control framework. Second, an LODRF mechanism is applied at the secondary control layer to actively reconstruct the power baseline by observing the load status at the point of common coupling (PCC) in real time. Furthermore, an APSO algorithm utilizing the integral of time-weighted absolute error (ITAE) as a global performance index is constructed to optimize key proportional–integral controller parameters cooperatively. Simulation results from a four-unit heterogeneous parallel system in MATLAB/Simulink demonstrate that the IAVI strategy enables stable convergence of frequency and voltage and proportional power sharing. Compared with the system without LODRF, the proposed strategy reduces maximum frequency and voltage dynamic deviations under load disturbances by 78.5% and 53.3%, respectively, and shortens effective recovery times by 0.01 s and 0.09 s, respectively. Moreover, compared with the standard PSO algorithm, the APSO-optimized system reduces maximum frequency and voltage deviations by 3.1% and 36.4%, respectively. Additionally, average active and reactive power sharing errors in the steady state are kept below 0.9%, verifying the significant advantages of the strategy in improving dynamic disturbance rejection and steady-state precision.

Accurate remaining useful life (RUL) estimation of insulated gate bipolar transistors (IGBTs) is challenging because degradation records are nonlinear, noisy, and sensitive to model configuration. This study proposes AH-PKO-MHA-BiLSTM, in which a two-layer bidirectional long short-term memory (BiLSTM) network encodes each observed turn-off peak of the collector–emitter voltage (V_CE) history window, four-head multi-head attention (MHA) reweights the resulting hidden-state sequence, and an adaptive hybrid pied kingfisher optimizer (AH-PKO), an enhanced variant of the pied kingfisher optimizer (PKO), searches the hidden size, learning rate, dropout, and L2 weight decay. The protocol uses a chronological 305/38/39 split, causal smoothing, and min–max scaling fitted on the training windows only, with validation-only model selection. The supervised target is an offline-defined normalized degradation index-based RUL proxy. Across ten paired final seeds, AH-PKO-MHA-BiLSTM achieved a root mean square error (RMSE) of 0.013509 ± 0.000356, a mean absolute error (MAE) of 0.010744 ± 0.000645, and a coefficient of determination (R²) of 0.8900 ± 0.0058. Its mean RMSE was 62.20% lower than that of a validation-selected V_CE cumulative-trend baseline, 32.30% lower than that of unoptimized MHA-BiLSTM, and 14.36% lower than that of standard PKO-MHA-BiLSTM. All paired neural-model contrasts remained significant after the Holm correction. Under an equal budget of 600 fitness evaluations, AH-PKO also attained the lowest mean final validation RMSE and the shortest search time across five search seeds. These findings support improved RUL-proxy prediction within a single-device accelerated-aging trajectory; cross-device, variable-load, and online generalization require further validation.