This paper proposes a novel real-time adaptive under voltage load shedding (UVLS) method by using optimized fuzzy logic controllers to address fault-induced delayed voltage recovery (FIDVR) and short-term voltage instability (STVIS) problems. An intelligent multi-input multi-output fuzzy logic load shedding controller is developed based on a short-term voltage recovery criterion (STVRC). For the controller, the magnitude deviation and the recovery time deviation of voltage trajectory are extracted as inputs, while the load shedding amount at the current load shedding stage and the time delay of the next stage are taken as outputs. Particle swarm optimization (PSO) technique is further employed for the tuning of the controller parameters to realize the voltage recovery requirement at the lowest cost of load shedding. The proposed UVLS method is formed by a group of the fuzzy logic controllers deployed at the load buses, each monitoring its own bus voltage and shedding load in closed-loop. Responding to the severity of voltage variations, the proposed method is able to adaptively determine the location, time and amount of load shedding in real-time to make the voltage satisfy the STVRC. The effectiveness of the proposed method compared to other methods is testified by simulations on the Nordic test system.

To ensure modern power systems with the ability to ride through frequency excursions after a power deficit, a refined and strict frequency recovery criterion (FRC) is formulated by grid codes. According to the FRC, this paper proposes a real-time adaptive load shedding method for frequency recovery control in receiving-end power systems. A co-active neuro-fuzzy inference system (CANFIS) based load shedding controller is developed for intelligent load shedding control. The magnitude deviation and the recovery time deviation of bus frequency are extracted based on the FRC and taken as feedback signals, while the shedding amount at current shedding round and the time delay of the next round are used as control signals. To further promote the control performance of the CANFIS controller, a reinforcement learning (RL) based deterministic policy gradient (DPG) algorithm is proposed, making the frequency recovery meet the FRC with the lowest cost of load shedding and enhancing robustness. Then, a decentralized load shedding strategy is constructed by CANFIS controllers deployed at load stations. Responding to the severity of frequency condition, the strategy can adaptively determine the time, location, amount and round of load shedding in real-time to make frequency recovery satisfy the FRC. Through simulations on a provincial receiving-end power system, the effectiveness and adaptability of the proposed method are demonstrated.

Power flow (PF) is the basis of steady-state analysis and control of power systems. The conventional model-driven PF formulated by a set of implicated nonlinear equations is solved iteratively by using Newton-Raphson method. But, the speed and convergence of PF computation are influenced by proper initial values and the efficiency of iterative process. The data-driven PF regression method can overcome the above issues by learning an explicit mapping function from PF data set. However, it simply achieves the nonlinear map from PF input to output, overlooking the physical rules in the PF calculation, which may result in poor accuracy and generalization. This paper presents a physically jacobian-informed encoder-decoder neural networks (NNs) for PF nonlinear regression. Based on the forward and inverse PF model, a multi-task learning method with encoder-decoder NNs is constructed for data-driven PF regression. To introduce PF physical characteristics, the corresponding jacobian information is embedded into the regression model to improve the accuracy and generalization. The performance of high accuracy and generalization of the proposed method is validated by IEEE test systems.