Energy Digest

A novel simulation-based, model-free framework is proposed for coordinated optimization of inverter-based resource (IBR) control parameters to enhance grid transient dynamic performance, eliminating the need for explicit mathematical models of complex nonlinear system dynamics. The framework uses a high-fidelity power system simulator and a projected multi-point zeroth-order optimization algorithm with adaptive moment estimation. Extensive simulations demonstrate the effectiveness of the approach in optimizing IBR control parameters to improve grid transient frequency response under large disturbances.

A new dissipativity-based distributed controller synthesis framework has been introduced for networks with heterogeneous, nonlinear agents and diverse performance objectives, leveraging the Network Dissipativity Theorem and iterative convex overbounding to enable network-wide consensus on dissipativity variables while keeping sensitive information locally. This approach addresses scalability and information privacy requirements, solving longstanding challenges in controller synthesis. It is applied to full-state feedback controller synthesis for networks with heterogeneous agents.

GeoDistNet is an open-source tool that generates synthetic distribution network structures from publicly available geographic information, addressing the lack of access to detailed utility feeder data. The tool constructs a candidate graph, synthesizes radial topology, assigns electrical parameters, and exports a usable network for power-flow analysis. It provides a reproducible workflow to bridge GIS data with simulation-ready distribution models when utility networks are unavailable.

Real-time surrogate modeling using convolutional neural networks (CNN) and Light Gradient Boosting Machine (LightGBM) has been developed to predict microgrid behavior with high accuracy. The models demonstrate performance changes depending on the variable being predicted, but a combined CNN+LightGBM model achieves stable performance across all variables. This approach offers significant computational efficiency gains, achieving speeds of over $1000\times$ for some models and near real-time performance.

Capturing the complex, non-linear dependencies between EV charging variables, including arrival times, durations, and energy demand, is crucial for accurate modeling. Traditional statistical methods often fail to capture these dependencies, but new frameworks such as Vine copulas and Copula Density Neural Estimation (CODINE) have shown promise in addressing this gap. These models outperform established parametric families and remain competitive against state-of-the-art benchmarks in predicting EV charging events and generating synthetic charging data.

Pulse charging can boost battery charging speed but introduces intermittent currents that challenge charger stability and energy supply. A coordination method is proposed to exploit off-time intervals of pulse current to charge other loads, mitigating fluctuation and amplitude of charging current by grouping and coordinating loads optimally through mathematical models. The proposed method has been applied and evaluated in two scenarios, showing improved results compared to random charging.

Optimization algorithms jointly design charging infrastructure and electric vessel operation to reduce costs, making electrification of marine transport more viable. A mixed-integer linear programming framework is presented for scheduling ferry operations, charging infrastructure, and ship battery size, reducing total costs by 7.8% compared to traditional scenarios. This approach has potential for large-scale applications like the China Zorrilla electric ferry, set to operate between Buenos Aires and Colonia del Sacramento in 2025.

A novel Fisher information matrix formulation compatible with hybrid systems is presented, taking into account discrete events that alter parameter influence propagation, using a saltation matrix to encode sensitivity transformations induced by these events. The resulting "salted Fisher information matrix" unifies continuous and discrete information updates in hybrid systems. This approach provides a unified framework for analyzing hybrid power systems.