Energy Digest

MambaDSSE is a model-free data-driven framework that incorporates Koopman-theoretic probabilistic filtering with selective state-space models to estimate distribution system states. The proposed method outperforms machine learning baselines on scalability, resilience to DER penetration levels, and robustness to data sampling rate irregularities. It captures long range dependencies from data, improving performance in distribution system state estimation.

A data-driven predictive control framework called Koopman Predictive Control (DKPC) is proposed to regulate power system frequencies using inverter-based resources with complex or uncertain dynamics, operating without requiring a parametric model. The method lifts nonlinear system dynamics into a higher-dimensional space where they can be approximated as linear, allowing for explicit constraints on control input and output. Numerical results demonstrate DKPC's effectiveness compared to a benchmarked approach.

A unified framework for attaining optimal synchronization in directed multiplex networks composed of phase oscillators is presented, featuring a multiplex synchrony alignment function that integrates structural properties and dynamical characteristics of individual directed layers. The framework yields two classes of frequency distributions that outperform conventional distributions, and numerical simulations demonstrate their effectiveness on various directed duplex topologies. Optimization through link rewiring and swapping algorithms also reveals correlations between node frequencies, including positive relationships with out-degree and negative correlations between neighboring frequencies.

The article characterizes the exact Pareto front in average-cost multi-objective Markov decision processes (MOMDPs), showing that it is a continuous, piecewise-linear surface lying on the boundary of a convex polytope. The Pareto front can be exactly computed without approximations, and each edge is realized as a convex combination of policies at its endpoints. This allows for efficient solutions to certain non-convex MDPs by leveraging the geometry of the Pareto front.

A new distributed model predictive control (DiMPC) methodology leverages multiparametric programming and facet properties to compute explicit control laws offline, reducing computational effort and communication load. This approach achieves comparable control performance to centralized methods while significantly reducing communication overhead and computation time by up to 98% compared to classic iterative DiMPC methods. The methodology is well-suited for real-time control applications with tight latency and computation constraints.

Dynamic coordination of distributed energy resources (DERs) can increase grid hosting capacity to support more renewables, storage, and electrified load growth by up to 22 times. Dynamic operation and DER interactions enhance capacity and power flows, reducing solar curtailment and improving reliability and power quality. Batteries emerge as the most critical technology for supporting dynamic resource coordination, enabling up to 200% solar penetration.

A phase-locked loop (PLL)-based adaptive sub-/super-synchronous resonance damping controller is proposed to address existing suppression methods for D-PMSG wind farms, offering a simple structure with easy parameter tuning and flexible adaptability. The PLL parameter is critical to SSO suppression, and only one key parameter needs to be tuned due to the avoidance of phase compensation. The controller was verified through CHIL tests under various operating conditions, addressing concerns about frequency estimation, computational efficiency, and potential impacts on PLL.

A physics-informed Reinforcement Learning framework combines semi-Markov control with a Gibbs prior to control topologies in power grids, balancing control quality and computational efficiency. The approach reduces exploration difficulty and online simulation cost while preserving policy flexibility. It achieves strong performance across benchmark environments, outperforming traditional methods by up to 255% in reward and 284% in survived steps.

A balancing market model integrated into Model Predictive Control (MPC) using a convex neural network has been developed to improve decision quality in implicit balancing, addressing shortcomings of previous models by capturing uncertainties and incorporating attention-based input gating mechanisms. The proposed model was evaluated on Belgian data, showing improvements in MPC decisions and reduced computational time. This approach enhances the accuracy of grid stability maintenance and profit earning for transmission system operators.

A new power transformer differential protection system is proposed that eliminates the need for an inrush blocking module, reducing delays in fault detection and removal. The system uses a data-aided approach with a neural network to distinguish between inrush and non-inrush current waveforms, extracting the fundamental component from the non-inrush part. This allows for more sensitive and rapid detection of internal faults hidden within inrush currents.