As a clean and flexible approach to energy utilization, hydropower is playing an increasingly vital role in power systems. However, runoff uncertainty significantly impacts the scheduling effectiveness of cascade hydropower stations. To address this, this paper constructs a short-term optimization scheduling model for cascade hydropower stations and proposes a novel Random Environment Simulation-based Multi-stage Reinforcement Learning (RESMRL) algorithm. The proposed algorithm effectively handles runoff uncertainty while overcoming the limitations of conventional deep reinforcement learning in terms of generalization and safety. RESMRL employs Random Environment Simulation (RES) to introduce a wide range of diverse scenarios into the agent's training environment, enhancing its alignment with complex and dynamic real-world scheduling conditions. Furthermore, RESMRL adopts a multi-stage training strategy that decomposes the short-term scheduling task into multiple learning stages with varying levels of difficulty. This enables the agent to progressively acquire scheduling experience, with each stage focusing on distinct objectives and thereby enhancing its adaptability to complex and diverse RES environments, starting from easier to more challenging scenarios. Additionally, a novel hybrid constraint processing framework is proposed. Integrating the physical characteristics of cascade hydropower stations, this framework accurately and efficiently manages the numerous constraints inherent in cascade hydropower scheduling, thereby ensuring the safety of scheduling decisions made by the agent. Case studies conducted on the cascade hydropower stations in the lower Jinsha River in China demonstrate the effectiveness and superiority of the proposed methodology compared to other benchmark approaches.

This paper aims to elucidate the mechanism of ultra-low frequency oscillation (ULFO) in the pumped storage power plant (PSPP) of dual units sharing one pipeline (DUSOP) and their suppression methods. Based on Hopf bifurcation theory, stability margin analysis, and the time-varying six-coefficient method, nonlinear dynamic models are established for three design scenarios: upstream single surge tank (USST), upstream and downstream double surge tanks (UDDST), and upstream tandem double surge tanks (UTDST). The effects of different surge tank configurations (STCs), governor parameters, and hydraulic parameters on unit frequency stability and oscillation characteristics are systematically analyzed. Results indicate that UTDST achieves smaller amplitude, larger damping, and shorter stabilization under load, outperforming UDDST in dynamic response. Surge tank water level oscillations contain multiple frequency components, while unit rotational speed frequency is primarily influenced by proportional gain. Crucially, the proportional gain significantly increases frequency while inducing non-monotonic damping changes driven by the competition mechanism between mechanical damping and hydraulic negative damping, whereas the integral gain, though raising frequency, weakens damping and may induce instability due to phase lag superposition. Additionally, reducing water flow inertia time constant and increasing mechanical inertia time constant expand the stability domain and improve system stability. Finally, a three-dimensional distribution of frequency and damping ratio with regional partitioning enables visual optimization of key control parameters, delineating a robust ‘safety island’ for the effective suppression of ULFO.

The rapid growth of renewable energy (RE) has introduced challenges for power systems, including output fluctuations from high RE penetration. Conventional cascaded hydropower (CHP) struggles to accommodate wind and photovoltaic (PV) output. Hybrid pumped storage hydropower (HPSH), through bidirectional pumping, stores excess RE and releases it at peaks, improving flexibility. Thus, a multi-objective framework coordinates HPSH-WIND-PV to balance economic benefits and system stability. K-means generates 8 representative scenarios covering typical and extreme outputs. A multi-objective stochastic optimization-based day-ahead scheduling strategy for a complementary system has been developed, which fully embodies HPSH's integrated pumping-generation characteristics and is designed to achieve an optimal balance between system stability and economic returns. To characterize the trade-off surface, an improved normalized normal constraint (INNC) method is employed; relative to NNC and epsilon-constraint, it produces a more uniformly distributed Pareto set, reducing the variance of adjacent-solution distances by 6.21 % and 38.56 %, respectively. In a case, revenue increases by 6.38 %, renewable curtailment decreases by 78.06 %, with curtailment even reduced to zero in some scenarios, and residual-load variance drops by 10.28 %. Results confirm that coordinated HPSH-WIND-PV scheduling balances economic efficiency and operational stability, offering theoretical and technical support for high RE grids.

Rotating machinery is crucial element in mechanical equipment, and during serving cycle their failure is inevitable because of artificial and non-artificial reasons. Signal processing techniques are available to diagnose the failure. Due to the nonlinearity and simplicity in computation rules and the richness in theoretical system, mathematical morphology (MM) has received significant research attention in this area, and numerous papers had been published in academic journals, conference proceedings, etc. The review paper attempts to overview the morphological framework and to summarize these applications grouped as rolling element bearing and gear. Finally, the relevant discussions on MM are analyzed, and several potential prospects are suggested.

As the global water crisis intensifies, human water extraction activities exert an increasingly profound impact on the sustainability of water resources. The distribution of water user groups in river basins exhibits distinct network characteristics, with pronounced asymmetric externalities in water extraction. However, current research lacks adequate investigation into the impact of these features. This study develops an evolutionary game model on weighted and directed water use networks, using the Hanjiang River Basin in China as a case study to analyze network characteristics and water extraction dynamics. The results reveal that the network exhibits significant strength assortativity and small-world properties. The network topology and player payoff structures drive water extraction dynamics, leading to diverse behavioral patterns. Under initial conditions, cooperative water extraction emerges as the dominant strategy in the network, with an average cooperation ratio of 0.7031 over 100 rounds in noiseless simulations. Cooperation is primarily observed in the midstream and downstream sections of the basin. Key parameters that drive cooperation, including marginal benefit, penalty coefficient, and marginal cost, reflect important institutional and environmental factors shaping users' decisions. As noise intensifies, decision randomness increases, significantly undermining the strategic superiority of the dominant strategy and reducing the amplitude of phase oscillations among nodes. When cooperative extraction is favored under the given network and payoff structure, reducing controllable uncertainty can effectively promote cooperation. The study provides key insights for sustainable river basin management.

The present study investigates the stability and dynamic response characteristics of pumped storage units (PSU) in a primary frequency regulation (PFR) under opening control mode. A comprehensive nonlinear mathematical model has been developed for the pumped storage unit governing system (PSUGS), incorporating hydraulic-mechanical-electrical coupling dynamics, including explicit representation of the synchronous generator excitation system. Derivation of nonlinear state equations is undertaken in order to characterize multi-energy domain interactions within PSUGS. These state equations serve as the fundamental framework for subsequent stability analysis. Employing Hopf bifurcation theory, the stability boundaries and critical oscillation modes are systematically analyzed under grid frequency disturbances, with numerical simulations validating the theoretical bifurcation thresholds. Further investigation reveals that PSUGS exhibits a dual-frequency oscillation phenomenon during PFR: (1) low-frequency oscillations (LFOs, 0.1–2.5 Hz) governed by generator dynamics and (2) ultra-low-frequency oscillations (ULFOs, <0.1 Hz) dominated by hydraulic transients in the penstock-governor subsystem. It is noteworthy that the Hopf bifurcation boundary consists of two distinct components corresponding to these oscillation modes, thereby demonstrating frequency-dependent instability mechanisms. The process of parametric sensitivity analysis is one of quantifying the regulatory impacts of key factors. These factors include hydraulics, mechanical, and electrical parameters. The results of the study demonstrate that ULFOs are particularly sensitive to hydraulic system configurations, while LFOs are predominantly influenced by electrical grid interactions. The findings of the present study provide a theoretical foundation for the optimization of frequency regulation strategies and the suppression of multi-frequency oscillations in pumped storage systems.

Runoff prediction plays a critical role in water resource management and flood mitigation. Traditional runoff prediction methods often rely on single-layer optimization frameworks that process the data without decomposition and employ relatively simple prediction models, leading to suboptimal performance. In this study, a novel two-layer optimization framework is proposed that integrates data decomposition techniques with multi-model combination strategies, establishing a closed-loop feedback mechanism between decomposition and prediction processes. The framework employs the Snow Ablation Optimizer (SAO) to optimize combination weights across both layers. Its adaptive fitness function incorporates three evaluation metrics—Mean Absolute Percentage Error (MAPE), Relative Root Mean Square Error (RRMSE), and Nash–Sutcliffe Efficiency (NSE)—to enable adaptive data processing and intelligent model selection. We validated the framework using observational data from Huangzhuang Hydrological Station in the Hanjiang River Basin. The results demonstrate that, at the decomposition layer, optimal performance was achieved by combining non-decomposition, Singular Spectrum Analysis (SSA), and Complementary Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) (with coefficients 0.4061, 0.6115, and −0.0063), paired with the long short-term memory (LSTM) model. At the prediction layer, the proposed algorithm achieved a 32.84% improvement over the best single decomposition method and a 30.21% improvement over the best single combination optimization approach. These findings confirm the framework’s effectiveness in enhancing runoff data decomposition and optimizing multi-model selection.

This paper aims to study the mechanism of avoiding the S-shaped region (S-shaped region, SFR) during the startup of pumped storage units (pumped storage units, PSUs). Firstly, the state space model of the PSU in frequency mode is built using the transfer coefficient of the pump turbine. Then, according to the characteristics of the SFR, the accurate range of the SFR is determined in the full characteristic curve. Finally, combined with a specific power station, this paper proposes a novel geometric perspective method to reveal the underlying mechanism for avoiding the SFR during the startup of PSUs. The core innovation lies in establishing, for the first time, the precise spatial relationship (positioning and distance) between the no-load operating point and the upper boundary of the SFR, thereby identifying two critical necessary and sufficient conditions for successful startup avoiding instability. Based on this mechanism, the critical state of PSUs entering the SFR and the influence of operation points on the startup stability that the PSU is putting into PID control are analyzed using the Hopf bifurcation principle. The results show that two conditions need to be met when the PSU starts up to avoid the SFR. One is that the system operation point is in the stable region, and the other is that the speed overshoot is less than the critical speed overshoot. The speed overshoot is the direct cause of the unit entering the SFR, leading to startup failure. When the PSU is started up and put into proportional–integral–derivative (proportional–integral–derivative, PID) control, a certain margin of flow and guide vane opening will help reduce the speed overshoot and prevent the unit from entering the SFR.

Abstract The health condition assessment (HCA) of hydropower unit plays an important role in enhancing safe operation of hydropower stations and reducing maintenance costs. Due to the harsh environment, there are some problems of sensor data during the unit operations, including abnormal data, missing data and high-noise data. Also, it is difficult to obtain complete monitoring data during run-to-failure maintenance. These issues definitely cause the low-confidence HCA. Thus, a data-model-interactive HCA for hydropower unit was proposed. First, a high-fidelity 3D mechanism simulation model, interacting with the actual unit, was built using unit drawing data provided by manufacturers. Average values of the volute inlet pressure and draft tube outlet pressure were fed into the dynamics simulation model to obtain the simulated pressure pulsation data under working conditions (water head H and power P). Then, a long short-term memory based healthy condition model was constructed using the power parameters, simulation and normal pressure pulsation. After model construction, difference values between simulated pressure pulsation dataset and degradation state pressure pulsation dataset were calculated to build the performance degradation index (PDI), describing the HCA of units. Finally, the PDI was fed into the convolutional neural networks and long short-term memory model to achieve degradation trend prediction. Validation experiment was conducted to verify the effectiveness of proposed method using actual monitoring data and operating parameters of 6# unit in the hydropower station.

This study proposes an adaptive selection method for hydrological runoff prediction models based on reinforcement learning(RL), specifically using the deep Q learning(DQN) algorithm. This method adaptively selects the optimal data-driven model, including Recurrent Neural Networks (RNN), Long Short-Term Memory networks (LSTM), and Gated Recurrent Unit networks (GRU), to optimize prediction performance. Furthermore, the effectiveness of this method is validated through simulation examples. Compared to single models and traditional combined models, this strategy adaptively selects a combination of prediction models based on the characteristics of the prediction target. Consequently, it enhances the dynamic adaptability of model selection, thereby improving the accuracy and stability of predictions.

Recently, deep learning technology-based neural networks have been adopted for remaining useful life (RUL) prediction of rotating machines. However, there are still some shortcomings: (1) an individual degradation feature cannot sufficiently represent the degradation process, which has an adverse impact on the accuracy of prediction results; (2) most recurrent neural network-based prediction methods have difficulty in quantifying the uncertainty of the forecast results. In this paper, a fusing sensitive degradation features with uncertainty analysis for RUL prediction of rotating machines is proposed. Firstly, the statistical features contained in the vibration signal used to monitor the degradation of rotating equipment are extracted to construct the original feature set. Then, the weight coefficients of the monotonicity, correlation and robustness criteria are determined by the self-adjusting analytic hierarchy process. The sensitive features that describe the degradation process are selected from among the statistical features. Furthermore, the sensitive features are fed into residual networks and gated recurrent unit, and the spatial and temporal correlation of the features are considered to establish the health index (HI). Finally, the fitted HI is input into a Gaussian process regression model, and the prediction results with confidence intervals are obtained. To verify the effectiveness and superiority of the proposed method, two public bearing datasets and three model methods are used for comparative experiments.

Runoff forecasting is crucial for planning and managing water resources. As hydrological data becomes more available, more data-driven models are being employed to enhance the effectiveness of runoff forecasting. To accurately and quantitatively assess the whole runoff forecast process, from model construction to application, a whole life cycle forecasting evaluation index system was established in this study to evaluate the data, factors, sample integrity, model construction, and forecast result. The analysis included quantitative evaluation criteria to comprehensively consider data quality, forecasting factor characterization, sample representativeness, model generalization, and result quality. An evaluation of forecast results from 7 river basins and their 85 hydrological stations in China showed that the proposed index system could accurately reflect the performance of the forecasting process. The overall performance of the forecasting model and process can be evaluated quantitatively based on the Euclidean distance, and the pathways to improve the forecasting effectiveness of the model can be identified based on the evaluation results. The validity of the proposed index is also experimentally demonstrated. The proposed index system can be applied to the evaluation of data-driven forecasting processes in various fields.

In daily life, tasks such as choosing a mode of transportation in traffic, diagnosing diseases and selecting medications in healthcare, as well as recommending products in e-commerce, can all be fundamentally classified as multi-classification tasks. Currently, effective approaches to solving multi-classification tasks include the behavior modeling-based Integrated Choice and Latent Variable (ICLV) model and the machine learning-based Multinomial Logit Model (MNL). The former slightly outperforms the latter in multi-classification tasks due to its ability to identify latent variables and integrate the selection process. However, if certain shortcomings of the MNL model, such as the assumption of independence from irrelevant alternatives, linearity assumption, and lack of hierarchical structure, can be addressed, MNL could outperform the ICLV model in some datasets. Graph Neural Networks (GNNs), which treat the entire feature set as a graph and consider the relationships between features, break the linearity assumption and offer a more flexible and hierarchical structure. This indicates that GNNs are able to effectively alleviate the limitations of MNL. Therefore, we propose an innovative GNN MNL composite model: first, GNN is employed to efficiently extract features from the dataset, and then the extracted features are used as input to train the MNL model. Finally, the trained MNL is utilized to classify new samples. The model’s accuracy was enhanced by incorporating Generative Adversarial Networks (GANs) for data augmentation during the training process. Through validation on three datasets, including modeChoiceData, we demonstrated that the GNN MNL composite model indeed achieves higher accuracy, confirming its feasibility. Future research could explore the generalizability of the GNN MNL model in other classification domains.

Graph data-driven methods have swept the field of machine fault diagnosis by merits of modeling relationships between samples. Their performance is highly affected by the constructed graphs quality. Compared to the single-sensor data, multi-sensor data can provide more information, so as to construct higher-quality graphs. However, existing graph data-driven diagnosis methods using multiple sensors still have two limitations. Firstly, heterogeneous multi-sensor data are mainly processed as homogeneous data, ignoring the heterogeneity of heterogeneous multi-sensor data. Secondly, the heterogeneous graph is often with a complex graph structure, and consumes much computational cost to learn. To overcome these limitations, A meta-path graph-based graph homogenization framework for machine fault diagnosis is proposed. Heterogeneous multi-sensor data are converted into the heterogeneous graph, modeling the heterogeneity of heterogeneous multi-sensor data. Further, instead of directly inputting the heterogeneous graph into graph deep learning model, a heterogeneous graph homogenization framework is designed to generate a meta-path graph, reducing the complexity of graph structure and improving the graph quality. Finally, a graph convolutional network is used for graph feature learning, obtaining the diagnosis results. Verification experiments show that the proposed method performs better than machine learning-based and graph deep learning-based methods. In addition, discussive experiments show that the meta-path graph is with lower complexity in graph structure and a higher clustering accuracy than single-sensor data-based K-nearest neighborhood graph.

Multivariate signal processing methods are becoming more prevalent as sensors and modern science and technology improve. However, most existing multivariate signal decomposition methods suffer from the following challenges: (i) requiring prior knowledge of multivariate signal modes; (ii) limiting to narrowband signal analysis; (iii) presenting mode mixing. To overcome the challenges mentioned above, this study proposes a novel method, named adaptive multivariate chirp mode decomposition (AMCMD). The method captures time-varying joint modes one-by-one in a recursive framework without knowing the precise number of modes. Specifically, a multivariate chirp mode (MCM) is modeled first based on AM–FM signals, with the constraint that there is a joint frequency component between all signal channels. Furthermore, demodulation techniques are used to entail the wideband multivariate signal mode exhibit narrowband characteristics. Finally, the objective function is established and the modes of each channel signal are estimated one by one. The efficacy and superiority of the method are verified by a series of numerical examples. In addition, the analysis of real-world time-varying vibration signals also confirms the practicality of the method. The findings demonstrate that the proposed method can converge faster to the same satisfactory results as existing state-of-the-art methods at a faster rate, even without knowing the precise number of modes.

Water conflict is evolving into one of the most vital social and environmental issues as the global water crisis intensifies in recent years. Previous studies mostly focus on the game between water users but overlook the fact that water resource institutions (WRIs), which are increasingly involved in water conflicts, are also essential players. In this study, a tripartite evolutionary game model that considers WRIs, demand-side users, and supplyside users is developed. Seven potential evolutionary stable strategies and nine evolutionary scenarios are deduced. The theoretical findings are validated by simulating the water conflict of two provinces (i.e., Hubei and Henan) in the Hanjiang basin, China. The results indicate that non-cooperation among water users and ineffective intervention by WRIs are the long-term outcomes of this water conflict case. To foster cooperation, the game structure needs to be altered by affecting the external variables that determine the net benefits of excess water intake and water rights transfer. Owing to the insufficient long-term incentives for WRIs, cooperation among water users cannot rely on the intervention strategy. Nonetheless, even if external conditions are immature, the intervention strategy can act as auxiliary means to foster cooperation in advance but at the cost of some of the WRIs' interests. The outcomes of the present study can provide managers of water resources with contributing information for cooperation promoting among water users.

Power system dispatch (PSD) greatly depends on load forecast (LF) with high accuracy. However, since load curve evolving along time axis is affected by long-term trendy and short-term stochastics electricity consumption modes, it is not easy to forecast load under accuracy requirement of PSD especially for high-proportion renewable energy power system. Therefore, a novel dispatch adaptation load feature mapping network with coordinated memory (LFMN-CN), which can forecast multi-timestep load values in future dispatch span at a time, is proposed. It adopts layered mapping network structure: 1) load features based on periodicity are mapped into two-dimension input matrix in the first layer; 2) nodes in the hidden layer with long-term and current memory, which denote long-term trendy and short-term stochastics electricity consumption modes respectively, are fully connected to construct recurrent LF network; 3) multi-timestep LF values in outer layer are obtained by LF output vector adaptive to timesteps of dispatch span. It has advantages of improved accuracy and dependence only on historical load series. LF results in power system of China show that the proposed model can perform multi-timestep LF more accurately than single-timestep LF.

Based on the joint scheduling model of cascade reservoirs and a dynamic programming (DP) algorithm, this paper studies the optimal control of the yearly drawdown level of an overyear regulation reservoir considering the influence of inflow uncertainty. An innovative dynamic control method has been put forward, and the corresponding technical route is provided. In case study, the seven reservoirs of the Yalong River are used as the research object, the proposed dynamic control method is verified by a detailed case study, and yearly drawdown level dynamic control bounds of the Lianghekou reservoir under two inflow series are constructed. Based on a long series of historical inflows, the simulation calculation and detailed comparative analysis are carried out. It is found that the dynamic control bound constructed by the selected inflow series has little impact on the fluctuation of scheduling results and can well cope with the impact of inflow uncertainty on the scheduling results. In addition, compared with the traditional fixed-yearly-drawdown-level control mode, the proposed dynamic control method can consider the interannual difference of inflow, which can increase the total power generation of the cascade system by more than 94 billion kWh at maximum and realize 63.4%~76.3% of the benefits of the lifting space of yearly drawdown level optimization.