IntroductionThe realization of China’s dual-carbon goals relies on accurate and fine-grained carbon emission modeling of the new-type power systems. Conventional carbon calculation methods are designed for traditional power systems and rely on deterministic power flow calculations. This makes it difficult to capture the impacts of source–load uncertainty on the carbon flow rate under high renewable penetration.MethodsThis paper proposes a carbon flow rate prediction model based on Edge Conditioned Convolution (ECC) to account for source–load uncertainty. First, a carbon flow rate calculation framework is established by incorporating the full lifecycle carbon emissions of renewable generators and a network loss allocation mechanism. This mechanism combines upstream and downstream power flow tracing to allocate network losses between the generation and consumption sides using an allocation coefficient. Second, an ECC-based carbon flow rate prediction model is developed to incorporate branch attributes into information aggregation through edge-conditioned convolution weights. A physics-informed loss function is adopted to encourage the conservation of the carbon flow rate. Furthermore, by using Sobol sequences and Inverse Transform Sampling (ITS) algorithms to generate multi-scenario data sets, the application scenarios of the prediction model are extended to power systems with high renewable-energy penetration. This also enhances the model’s generalization ability under uncertainty.ResultsCase studies show that the proposed model achieves a mean absolute error of approximately 0.144 tCO2/h, confirming practical prediction accuracy. Further analysis reveals that under different renewable penetration levels, the method more accurately characterizes carbon flow rate distributions and adapts better to source–load uncertainty than conventional carbon flow tracing methods.DiscussionThese results support the use of ECC to predict carbon flow rates under source–load uncertainty and the specified network-loss allocation scheme, providing a potential basis for low-carbon power system dispatch.
To address the substantial computational burden of high-order spatial derivatives, the difficulty of residual optimization near material interfaces, and the optimization stiffness associated with direct space-time training when conventional strong-form physics-informed neural networks (PINNs) are applied to heterogeneous multigroup neutron diffusion problems, a PINN method based on weak-form discrete physical residuals is developed. Finite-element shape functions are used to construct the flux interpolation field and assemble the discrete physical operators, while the neural network predicts nodal fluxes and neural source iteration drives the coupled convergence of the fluxes and effective multiplication factor. P1 triangular elements are adopted for the two-dimensional problem and Q1 hexahedral elements for the three-dimensional problem, eliminating the need to evaluate high-order spatial derivatives of the flux by automatic differentiation during training. The method is progressively verified using the TWIGL and IAEA-3D PWR benchmarks. For the steady-state TWIGL problem, the PINN yields an effective multiplication factor of 0.9132141, differing from the finite element method (FEM) result by 0.04 pcm, while the root-mean-square errors (RMSEs) of both fast- and thermal-group fluxes are below 2.4 × 10−4. When this steady-state flux is used to initialize the FEM/Crank–Nicolson (C-N) transient calculation, the RMSEs of relative power for the step and ramp perturbations are 5.27 × 10−4 and 4.57 × 10−4, respectively, relative to the pure FEM/C-N results; the maximum absolute relative deviations from the published POLCA-T reference values are both below 0.1%. For the IAEA-3D PWR problem, the PINN produces an effective multiplication factor of 1.0289669, differing from the same-mesh FEM result by 3.67 pcm. The RMSEs of the fast- and thermal-group fluxes are 3.008 × 10−3 and 1.487 × 10−3, respectively. For the assembly relative power, the mean and maximum absolute relative errors with respect to published reference values are 0.1779% and 0.4647%, respectively. These results demonstrate that the proposed method avoids the dependence of strong-form PINNs on high-order spatial automatic differentiation and achieves accurate solutions of heterogeneous two-dimensional P1 and three-dimensional Q1 multigroup core-diffusion problems within a unified framework, providing a common methodological basis for differentiable solution of complex reactor-physics problems.
In regional power grids with multiple high-voltage direct-current (HVDC) infeeds, renewable-generation variability, variations in HVDC power transfers, and network constraints are strongly coupled, while planning solutions derived from DC power-flow models may not ensure AC operational feasibility. To address these issues, we propose an AC-security-feedback-based method for the siting and sizing of pumped-storage power plants. The proposed method integrates zonal source–load data with data from multiple HVDC infeeds to construct a joint scenario-feature set incorporating residual load and its ramping characteristics. Representative scenarios are then extracted through block-wise standardization, principal component analysis, and K-medoids clustering. Subsequently, a two-stage mixed-integer programming model is formulated by jointly considering candidate-site resource potential, engineering constraints, and grid-interconnection conditions. The model accounts for integer unit configurations, multi-scenario operation, operational constraints of conventional generating units, and reserve requirements. To overcome the inability of DC power flow to accurately represent active-power losses, reactive power, bus-voltage magnitudes, and branch apparent power, the resulting planning scheme is subjected to rigorous AC power-flow validation for every scenario and time period. The resulting security information is fed back into the planning model to iteratively refine the solution. Case-study results demonstrate that the proposed method preserves the computational efficiency of mixed-integer programming while effectively eliminating overloads on critical branches. All operating points satisfy the prescribed security constraints, thereby substantially improving the AC operational feasibility of the planning solution.
Stability and power quality assessment are critical for the effective and reliable operation of hybrid renewable power grids, where the intermittency of solar and wind generation can lead to significant voltage and frequency fluctuations. This study proposes a framework that integrates an Unscented Kalman Filter (UKF) with a Graph Neural Network (GNN) to simultaneously estimate system states, evaluate dynamic stability, and monitor power quality. The UKF provides accurate voltage and frequency estimation under various disturbances, while the GNN predicts the operating states: stable, marginal, and unstable. The framework is validated under multiple scenarios, including nominal operation, high-load conditions, renewable intermittency, faults, measurement noise, and extreme disturbances. The results demonstrate low estimation errors, with voltage RMSE ranging from 0.008 to 0.06 pu, and frequency RMSE ranging from 0.01 to 0.08 Hz, depending on the operating conditions. The GNN achieves classification accuracy above 94% in most scenarios, reaching up to 98%–99% under nominal conditions. Furthermore, multi-scenario heatmaps and standard power quality indices, such as voltage sag/swell, frequency deviation, and total harmonic distortion (THD), are employed to verify robustness and consistent performance under simulated conditions of the proposed approach. THD remains below 2% under nominal conditions and increases to 5%–5.5% under disturbances, while voltage sag and swell events occur within the ranges of 5%–10% and 3%–6%, respectively. Moreover, the stability index ranges from 0.85 to 0.88 under nominal conditions, decreases to 0.70–0.78 under high-load and renewable-intermittency conditions, and approximately 0.20 under extreme disturbances. Overall, the UKF–GNN framework provides an effective framework for stability and power quality assessment under simulated conditions in renewable-energy-based power systems.
A modified lithium-ion cell integrated with a hollow-core fiber photothermal spectroscopy (HCF-PTS) sensor was developed for operando monitoring of gas evolution during electrochemical cycling. The platform enabled simultaneous measurements of cell voltage, internal pressure, and gas-phase concentrations of C2H4 and CO2 in a sealed LiFePO4/graphite cell. The HCF-PTS system with limits of detection of 1.2 ppm for C2H4 and 1.7 ppm for CO2 were obtained. Measurements from two independently assembled modified cells showed similar voltage, pressure, and gas-evolution profiles, demonstrating good reproducibility under the present experimental conditions. Comparison with a conventional coin cell further showed similar initial charge–discharge voltage characteristics. In addition, CH4 detection was separately demonstrated with a limit of detection of 1.1 ppm, illustrating the scalability of the sensing architecture toward additional gas species. The proposed platform provides a minimally invasive approach for correlating electrochemical behavior with gas-phase reaction dynamics in working lithium-ion cells.
