In response to the challenges where large-scale renewable energy integration leads to intricate source-network-load-storage elements and surging complexity in new power systems, rendering traditional balancing architectures and hierarchical analysis methods inadequate, theoretical achievements and research technologies regarding hierarchical and partitioned balance architectures are comprehensively reviewed. The adaptability requirements of new power systems for such architectures are elucidated, followed by a summary and comparative analysis of existing hierarchical control and partitioning strategies. Furthermore, layer-zone fusion mechanisms are explored, and existing technical limitations are analyzed from data and modeling perspectives. The results indicate that while renewable energy control pressures can be alleviated by existing strategies, deficiencies remain in handling massive heterogeneous data fusion and precise modeling of complex systems; moreover, high dynamic balance demands are difficult to be met by current layer-zone coordination mechanisms. Future hierarchical and partitioned balance architectures are identified as a critical direction for supporting the operation of new power systems. Notably, a novel technical pathway for achieving safe and efficient operation under the “carbon neutralization and carbon peaking” goals is offered by the introduction of large models and artificial intelligence technologies.

Hydrogen production from new energy is the most promising hydrogen production method under the targets of carbon peak and carbon neutrality. Aiming at the problem that the off-grid new energy hybrid hydrogen production system lacks a cooperative control strategy， this paper proposes a multi-level power replacement strategy based on the expected operation interval of the proton exchange membrane （PEM） electrolyzer. When the power command on the hydrogen production changes， the PEM electrolyzer first makes rapid adjustments to respond to the change. Then， a power replacement strategy is formulated in combination with the expected operation interval of the PEM electrolyzer. By utilizing the large-capacity support capability of the alkaline electrolyzer， the power outside the expected range of the PEM electrolyzer is gradually transferred to the alkaline electrolyzer， which fully exploits the multi-type and multi-time-scale response capabilities. Finally， the strategy proposed in this paper is tested on the self-developed integrated simulation platform for new energy hydrogen production. The operation results show that the strategy proposed in this paper can fully exert the multi-time scale response capabilities of the two types of electrolyzers， not only newly improves the response speed of the power command but also ensures the safety of the hydrogen production system. Moreover， it can meet the stability requirements of off-grid hydrogen production using new energy， and thus has high practical value.

As an efficient hydrogen production technology， alkaline electrolyzer holds significant application prospects in green hydrogen production. For alkaline water electrolysis hydrogen production systems， this paper proposes an accurate and applicable semi-theoretical and semi-empirical alkaline electrolyzer model based on electrochemical principles and thermodynamic analysis. This model treats the electrolyzer’s cell voltage and gas purity as functions of operating pressure， temperature， and current density， and incorporates the influence of Faraday efficiency on hydrogen production rate. Based on literature-reported experimental data of a 50 m³/h alkaline electrolyzer under different operating conditions， this paper determines the model parameters via fitting using a nonlinear regression method. This paper further employs these experimental data to verify the model’s accuracy and analyze its applicability under varying operating conditions. The simulation results show that the proposed model can effectively predict the electrolyzer’s performance parameters， thereby providing a theoretical basis for the optimal design and the control systems of electrolyzers.

In recent years，typhoons have occurred frequently in coastal cities，and the vulnerability of distribution network lines and loads affected by typhoons has been increasing day by day. Therefore，an optimal scheduling model of mobile emergency power supply vehicle based on hierarchical sequence method was proposed. Firstly，the failure rate model under extreme disaster conditions was constructed，and the Monte Carlo method was used to determine the line vulnerability model. At the same time，the synergy of multiple flexible resources such as distributed power sources，energy storage systems and mobile emergency power vehicles under different spatial and temporal scales was considered to formulate a reliable distribution network resilience assessment model combining islanding and reconfiguration. Finally，the original nonlinear problem was convexized into a standard mixed integer second-order cone problem that was easy to solve by the second-order cone relaxation technique. In this study，the hierarchical sequential solution algorithm is used to take the supply rate of important loads as the main goal，while reducing the overall load loss of the distribution network as much as possible as a sub-goal. The effectiveness of the strategy is verified by the example results. Compared with other scheduling methods，the accuracy of the algorithm are further proved.

The study explores the risk spillover effects among carbon markets， energy markets， and green finance markets， with a particular focus on the impact of extreme events on these markets. The aim is to provide insights for mitigating risks in the carbon market and promoting its healthy development. Based on an in-depth analysis of the risk spillover mechanisms among carbon， energy， and green finance markets， this research employs time-varying parameter vector autoregressive （TVP-VAR）-Diebold-Yilmaz（DY） and TVP-VAR-Baruník-Křehlík（BK） spillover index methods-and integrates them with a spillover network model. Utilizing historical data from 2018 to 2023 pertaining to China’s carbon market， energy market， and green finance market， this paper conducts an empirical investigation into the time-frequency risk spillover effects across these various markets. The results indicate that： （1） The spillover effects among the carbon-energy-green finance markets exhibit time variability and are susceptible to extreme sudden events. （2） Within different temporal domains and frequency cycles， both the green stock market and new energy market serve as primary sources of systemic risk transmission； conversely， while the carbon market acts as a recipient of risk in the short term， it transitions into a source of risk outflow over the long term. （3） The green stock market and new energy market constitute significant centers within the systemic spillover network.

With the rapid advancement of renewable energy， wind power has emerged as a significant clean energy source. Consequently， the accuracy of wind power forecasting is essential for ensuring both the stability and economic efficiency of the power system. To address the challenges of nonlinearity and non-stationarity in wind power forecasting and to enhance prediction accuracy and reliability， this study proposes a novel wind power forecasting model based on variational mode decomposition （VMD）， K-means clustering analysis algorithm， and TimesNet deep learning model. Firstly， VMD is employed to decompose nonlinear and non-stationary time series signals into multiple intrinsic mode functions （IMF）， facilitating the analysis and extraction of trends and periodic components from wind speed and generation data. Secondly， K-means clustering algorithm is utilized to classify the obtained IMFs， thereby identifying characteristic patterns of fluctuations in wind power. This process effectively enhances the model’s ability to capture variations in power under different wind conditions. Finally， the results processed through clustering are inputted into the TimesNet deep learning model for prediction. Comparative experiments with various existing wind power forecasting models demonstrate that the proposed forecasting model significantly reduces errors in predicting wind power output.

Aiming at the challenges of insufficient scheduling flexibility and rising operational costs in multi-park electric-thermal systems under large-scale renewable energy integration，this paper proposes an optimal scheduling model for tiered dual-time-scale distributed electric-thermal system based on edge computing. Firstly，a three-tier collaborative architecture comprising a physical equipment layer，edge computing layer，and cloud layer is constructed. Edge computing facilitates rapid data processing and distributed decision-making among parks. Secondly，the improved analytical target cascading method is employed with a dual-time-scale strategy： the lower layer optimizes electrical energy interactions at a 5 min granularity，while the upper layer coordinates thermal energy interactions at a 1 h granularity. The augmented Lagrangian method is integrated to decouple and iteratively solve multi-time-scale optimization problems. Finally，a benefit redistribution mechanism based on energy contribution degrees is designed，utilizing an asymmetric mapping function to quantify each park’s contributions to electric-thermal exchanges and renewable energy consumption，ensuring equitable profit distribution. Case studies demonstrate that the proposed model reduces comprehensive operational costs by 34.46% compared to conventional methods，significantly improves renewable energy consumption rates，and achieves convergence within eight iterations. The findings confirm that the integration of edge computing and dual-time-scale strategies effectively addresses spatiotemporal disparities in energy flows，providing theoretical and practical insights for coordinated optimization in multi-energy-coupled systems.

As a new type of power system scheduling mode，virtual power plant （VPP） can realize the efficient utilization of new energy power by aggregating distributed energy resources. However，the traditional scheduling strategy aiming at economy has been unable to meet the needs of current low-carbon development. Based on this，this paper proposes a multi-objective optimization scheduling strategy for VPP considering both economy and carbon emissions. Firstly，a post-combustion carbon capture device was introduced into the VPP system，and combined with a flexible carbon trading strategy，a multi-objective optimization scheduling model considering economic cost and carbon emissions was constructed. Secondly，to obtain the optimal solution of the model，the augmented ε-constraint method was used to solve the Pareto solution set，and the entropy weight - technique for order preference by similarity to ideal solution （TOPSIS） method was used to evaluate the solution set. Finally，multi-case simulation experiments were carried out around different carbon capture and carbon trading strategies to compare and analyze the differences in scheduling results between the single-objective model only considering economy or low-carbon characteristics and the multi-objective model considering both economy and carbon emissions. The experimental results show that when the ladder carbon trading mechanism and the corresponding carbon capture operation mode are adopted，the carbon emissions of VPP reach the lowest level. In addition，compared with the single objective model，the multi-objective optimization strategy considering both economy and carbon emissions can effectively reduce carbon emissions and improve economic benefits.

Under the goal of “Dual Carbon” strategy，how to realize the flexible interaction between virtual power plants and promote the low-carbon operation of virtual power plants through carbon price is a problem worthy of study. Therefore，the peer to peer （P2P） trading model of virtual power plants is studied based on the carbon flow theory. Firstly，based on the carbon emission flow theory，the distribution characteristics of carbon flow in the network are analyzed，and natural gas is introduced to form a multi-energy network，and a low-carbon economic scheduling model is established. Secondly，considering the privacy of each virtual power plant participating in the trading，a quantitative index method including the trading information of the bid volume and quotation is proposed，and a P2P trading model based on comprehensive priority is established. At the same time，combined with the carbon emission responsibility of virtual power plants in the network，carbon pricing method is introduced into the P2P trading mechanism，and a comprehensive price model of “energy-carbon” based on carbon tax is established. Finally，an example is given to verify that the proposed method can not only reduce the operating cost of the virtual power plant，but also effectively reduce the carbon emission.

Dynamic equivalent modeling of large-scale wind farms is the foundation for studying wind power grid integration，while the clustering-based equivalent model of wind farms cannot fit the dynamic output characteristics with high accuracy，and the poor generalization ability in its application is an inherent defect of clustering based model. Aiming at this problem，this paper proposes a wind farm equivalent modeling method based on particle swarm optimization-long short term memory neural network-error correction model （PSO-LSTM-ECM）. Firstly，K-means clustering algorithm and capacity weighting method are used to cluster wind turbines in wind farms，and a clustering equivalent model of the wind farms is constructed； Then，ECM is constructed based on the transient response errors of the detailed model and the clustering equivalent model，and the correction model is obtained through the LSTM neural network training optimized by PSO，and the output value of the network is compensated to the clustering equivalent model； Finally，a joint simulation is conducted on PSCAD and Matlab platforms to compare and analyze the detailed wind farm model，clustering equivalent model，and the model proposed in this paper. The result proves the effectiveness and superiority of the proposed model.

Integrating demand response into virtual power plants can enhance their flexibility and economic efficiency，but the inherent uncertainty of demand response poses challenges for scheduling and operation. Moreover，research on applying multiple demand response types in virtual power plants remains limited. To address these issues，this paper proposes an incentive-based demand response model incorporating a benefit coefficient that adjusts incentive compensation to reduce payouts under unsatisfactory demand response performance，as well as a replaceable-based demand response model considering customer satisfaction to better reflect its influence on replaceable demand response. Finally，a multi-energy virtual power plants model integrating multiple demand response types is established，considering three demand response strategies to achieve superior optimization. Case studies demonstrate that the incentive-based demand response model considering the benefit-coefficient can improve economic efficiency of the system，and the replaceable-based demand response model considering customer satisfaction can more accurately capture potential load variations to enhance demand response precision，and the coordinated participation of multiple demand response types in virtual power plant scheduling yields optimal overall performance.

The gigawatt-hour level large-capacity energy storage station provides critical support for the absorption of renewable energy， as well as enhancing system safety and operational control capabilities. However， with the significant increase in the number of energy storage units and individual battery cells within gigawatt-hour level battery energy storage stations， the risk of battery inconsistency has also intensified. This poses a serious threat to the safe and efficient operation of these storage facilities. To address this issue， an analysis is conducted on how the consistency of individual batteries and energy storage units affects the charging and discharging capabilities of energy storage stations. A method for calculating and evaluating these capabilities is proposed. A two-tier power distribution model for energy storage stations is developed， taking into account both station-level and virtual subsystem-level considerations. At the station level， based on the state of health （SOH） of each battery unit， the energy storage station is divided into several virtual subsystems. The overall charging and discharging strategy for each subsystem is designed according to their average state of charge （SOC） consistency， allowing for accurate calculation of total power demand across subsystems. At the virtual subsystem level， a power distribution strategy is established with an optimization goal aimed at minimizing SOC variance among stored units. The proposed power distribution method can delay lifespan degradation in low SOH energy storage units while accelerating SOC consistency convergence among them， thereby improving overall utilization rates within the energy storage station. Through simulation experiments conducted alongside analyses from Qinghai’s Togruoge 270 MW/1.080 GW·h battery energy storage project， we evaluated our proposed power distribution and management strategies. Results indicate that this approach demonstrates considerable effectiveness in achieving battery consistency management.

With the in-depth promotion of the strategy of "carbon peak and carbon neutrality"，virtual power plant （VPP） has shown significant advantages in integrating and dispatching new energy. It is one of the effective means to improve the operation economy of VPP to deeply tap the huge potential of new energy in the field of reactive power support. Firstly，a set of new energy reactive power capacity evaluation system considering active power output and inverter constraints was constructed. Secondly，in order to improve the solving speed of the reactive power optimization model，according to the nonlinear characteristics of active network loss and power flow constraints，an active network loss estimation method based on power flow iteration and a power flow linearization method based on variable space optimization selection were proposed. A VPP linear power flow reactive power optimization model considering the reactive power potential of new energy sources and based on variable space optimization selection was constructed. Finally，the improved IEEE 33-node active distribution system was taken as an example to verify the effectiveness of the proposed model. The results show that when VPP makes full use of the reactive power support ability of new energy，the system voltage deviation is reduced by 0.673 pu，and the operating cost is reduced by 1 254.9 yuan.

To address weak infrastructure， poor voltage stability， and low renewable-energy utilization in rural areas， this paper proposes a siting-and-sizing model for distributed generation （DG） that simultaneously optimizes voltage quality and economic performance. One objective aims to minimize voltage deviations caused by DG integration， thereby enhancing distribution-network power quality； the other seeks to minimize the levelized cost of energy （LCOE） over the full life cycle of the DG portfolio， accounting for investment， operation and maintenance expenses， and energy yield. The model is solved with a double deep Q-network （DDQN）， yielding a configuration that balances voltage stability and cost. Simulation on a modified IEEE 33-bus rural feeder shows that the DDQN-based scheme markedly improves voltage profiles while reducing upgrade costs. Furthermore， comparative analyses with the deep Q-network （DQN）， non-dominated sorting genetic algorithm II （NSGA-II）， and multi-objective particle swarm optimization （MOPSO） methods verify the superiority of the proposed approach， highlighting the efficiency， adaptability， and robustness of reinforcement learning for complex energy-system optimization.

Compared with salt caverns and artificial cavities，using pipeline steel as above-ground gas storage chambers offers greater advantages for small-scale distributed compressed air energy storage（CAES）systems. This paper establishes a detailed dynamic simulation model of a 10 MW-class distributed CAES system based on AMESIM software. The research investigates key parameters such as discharge duration，above-ground storage chamber volume，system efficiency，and energy storage density under different energy storage durations and different maximum storage pressures of the above-ground storage chambers. In addition，an economic analysis of the system is also conducted. The results show that heat loss of the thermal storage and exchange system is the main cause of energy loss in the CAES system. As the energy storage duration increases，the volume of above-ground storage chambers increases，while the system efficiency remains unchanged and energy storage density increases，meanwhile，the reduction rate of the static payback period gradually slows down. With an increase of maximum ground chamber pressure，the chamber volume decreases，system efficiency declines and energy storage density increases，while the static payback period first declines and then rises. When the maximum pressure of the above-ground chamber rises from 9 MPa to 14 MPa，the system efficiency drops from 67.59% to 54.37%. The minimum static payback period of 8.29 years is achieved at the optimal pressure of 11.8 MPa.

To address the prominent issues of insufficient utilization of user-side flexibility resources and the low degree of energy coupling in park-level electricity-heat-hydrogen integrated energy systems，this paper proposes a low-carbon scheduling strategy incorporating the concept of equivalent energy storage. First，user-side adjustable resources are considered，and the dispersed regulation capabilities among multiple user-side entities are aggregated，thereby introducing the concept of equivalent energy storage（EES）. Second，a multi-mode coordinated operation framework is established for park-level multi-energy systems，which integrates electrical energy storage，hydrogen energy storage，and hydrogen-blended gas combined heat and power units. This framework characterizes the coupling relationships of electricity-heat-hydrogen energy flows，while a stepwise carbon trading mechanism is introduced. Together with EES，user-side adjustable resources are aggregated to reduce the system’s dependence on high-carbon units. Finally，case studies are conducted to validate the effectiveness of the proposed strategy. The results demonstrate that，compared with the case without EES，the proposed method reduces the total operating cost of the system by 13.04% and achieves a 29.62% reduction in carbon emissions under the constraints

With the integration of large-scale， distributed， and diverse distributed resources， virtual power plant （VPP） technology has become a vital tool for effectively managing and optimizing demand-side resources. To better align VPPs with the development needs of China’s new-type power system， this paper proposes an optimal scheduling model for VPPs participating in a green certificate-carbon joint trading mechanism， taking into account uncertainty risks. First， an optimal operation model for a VPP is constructed， consisting of wind turbines， photovoltaic units， gas turbine units， energy storage systems， and flexible load resources on the user side. The objective is to minimize the VPP’s operating cost， considering electricity markets， the green certificate-carbon joint trading mechanism， and incentive-based demand response. Second， multiple uncertainty factors within the VPP， such as generation sources， loads， and demand response， are comprehensively considered， and the conditional value-at-risk （CVaR） theory is applied to quantify the risks associated with these uncertainties. Finally， a case study is introduced to verify the economic and environmental benefits of the proposed model. The inclusion of CVaR also provides a robust decision-making basis for balancing VPP profits and risks.

With the continuous advancement in the construction of new power systems， the issues related to insufficient new energy consumption capacity in Northwest China and electricity supply shortages in eastern regions have become increasingly prominent. Therefore， transporting surplus new energy from Northwest China to developed eastern regions has emerged as an effective means to achieve inter-regional energy balance. In response to the demand for long-distance and large-capacity new energy delivery， this paper proposes a collaborative approach utilizing both electricity and hydrogen for new energy transmission. First， we establish a framework for a new energy delivery system that encompasses both electric power transmission and hydrogen transport channels. Next， considering cost and loss factors associated with equipment at the source end， transmission end， and load end during the new energy delivery process， while imposing constraints on supply-demand balance， losses incurred during transportation capacity limits as well as delivery time， we formulate an optimization model aimed at minimizing total costs and losses within this collaborative planning context. Finally， we conduct simulation analyses using four cases from Gansu Province’s new energy delivery projects. The results indicate that implementing this collaborative planning scheme reduces overall system costs by approximately 5.72% to 7.74%， thereby validating the effectiveness and rationality of the proposed model.

The installed new energy capacity of new energy bases is usually equal to or slightly smaller than the capacity of collection engineering. Considering the randomness，fluctuation and intermittency of new energy units’ output，the capacity of collection engineering has not been fully utilized in most of the time. The potential scale of new energy that can be accepted by collection engineering could be further excavated from the perspective of national economy and rational new energy utilization rate. This paper studies reasonable accepted new energy scale of new energy bases in various voltage levels based on optimal levelized cost of electricity considering the characteristics of wind and solar power stations and regulation capability of power grid. The research of collection scene in north China region shows that the new method proposed in this paper can accept 50% larger capacity of wind and photovoltaic power and effectively reduces the levelized cost of electricity for new energy bases while meeting the requirements of new energy utilization rate. In addition，the acceptable new energy capacity of complementary collection stations can be significantly larger than pure wind or photovoltaic collection stations. Research in this paper can provide reference for acceptable new energy potential scale of collection engineering in subsequent new energy bases.

In response to the challenge of early warning for abnormal oil sump temperatures in wind turbine gearboxes, a fault warning method based on supervisory control and data acquisition (SCADA) data is proposed to enhance the operational reliability of the turbines. Firstly, by integrating wind speed-power distribution characteristics, an outlier detection approach utilizing the interquartile range and longitudinal filtering based on data dispersion is employed to eliminate anomalous power points. Subsequently, key input features influencing oil sump temperature are identified using a random forest algorithm, leading to the development of a temperature prediction model based on categorical boosting (CatBoost). The hyperparameters of this model are optimized using tree-structured parzen estimator (TPE). Finally, dynamic warning thresholds are established through statistical process control based on residual distributions. In a practical case study from a specific wind farm, this model issued effective warnings approximately 5 hours prior to gearbox failure; notably, the time points at which residuals exceeded control limits closely aligned with the progression of faults.The proposed method demonstrates significant efficacy in identifying abnormal conditions related to oil sump temperatures and possesses strong early warning capabilities along with substantial engineering application value.

In response to the supply-demand imbalance faced by intelligent buildings under the dual uncertainties of photovoltaic output on the source side and electricity demand on the load side， this study aims to reduce energy storage investment and electricity costs while enhancing the economic viability and robustness of shared energy storage systems. To achieve this goal， we develop a bi-level optimization model for shared energy storage based on hybrid game theory. In this model， energy storage operators and building users form a leader-follower game relationship； operators act as leaders setting internal transaction prices while users respond as followers through demand response strategies. Additionally， cooperative game theory is employed among buildings to fairly allocate costs using bilateral Shapley value methods.The uncertainty in source-load dynamics is characterized by constructing fuzzy sets for photovoltaic output using Wasserstein distance and incorporating conditional value at risk （CVaR） to depict investment risks arising from load fluctuations. The Karush-Kuhn-Tucker （KKT） conditions are utilized to transform the bi-level model into a single-layer mixed-integer linear programming problem for solution. Simulation results based on an intelligent building cluster in Jiangsu demonstrate that the proposed strategy effectively reduces redundant energy storage capacity by 12.3% and lowers average electricity costs across buildings by 8.7%， while simultaneously increasing operator profits and shortening payback periods for investments. Compared with traditional robust optimization methods and deterministic approaches， our method significantly enhances economic performance without compromising system robustness. The proposed hybrid game optimization strategy can collaboratively address dual uncertainties in sources and loads， facilitating efficient utilization of shared energy storage resources while achieving mutual benefits for all parties involved. This approach provides an effective pathway toward low-carbon operational efficiency for clusters of intelligent buildings.

Underwater compressed air energy storage（UWCAES）is vital for balancing power supply-demand fluctuations but faces challenges of instantaneous overpressure and pressure oscillations in flexible balloons during deep-sea operation and dynamic charging/discharging. This paper proposes a fuzzy PID（proportional integral derivative）- based method to suppress these pressure fluctuations. First，a dynamic pressure transmission model incorporating underwater environmental parameters is established for the balloon. Then，a fuzzy PID control algorithm is developed，utilizing the pressure error and its rate of change as inputs. This algorithm constructs membership functions and a fuzzy rule base to dynamically adjust PID parameters in real-time，optimizing the valve opening adjustment rate. Finally，case studies confirm algorithm robustness under dynamic conditions like vortex-induced shock. By achieving a 26.7% reduction in pressure standard deviation（to 30.8 kPa ）over PID control，the proposed strategy effectively mitigates overpressure and fluctuations，advancing the deployment of underwater flexible compressed air energy storage.

Driven by the “carbon neutrality and carbon peaking” goal，gravity energy storage has become an important support technology for new power systems due to its advantages of environmental protection，no self-discharge and flexible layout. Aiming at the power fluctuation problem caused by mass block scheduling in the charging and discharging process of gravity energy storage and the active/reactive power response demand of the grid，this study takes the ramp gravity energy storage as the object，and respectively constructes the simulation model of the gravity energy storage system including two types of motors（electrically excited synchronous motor and doubly-fed motor）and three typical control strategies（vector control，direct power control and sliding mode control）. The corresponding mathematical model and power coordination control strategy are established. The simulation results show that the electrically excited synchronous motor system with sliding mode control has the best dynamic response and steady-state accuracy in terms of active/reactive power regulation performance. The doubly-fed motor combined with sliding mode direct power control strategy also shows good adjustment ability and robustness.

To address the issue of low prediction accuracy in photovoltaic power generation caused by the intermittency and volatility resulting from weather changes， this paper proposes a multi-level short-term photovoltaic power forecasting method. The method is based on collaborative clustering using self-organizing map and K-means algorithm （S-Kmeans）， and an improved artificial lemming algorithm （IALA）-optimized variational mode decomposition （VMD）， combined with a temporal convolutional network （TCN） and bidirectional gated recurrent unit （BiGRU）. First， key meteorological factors are selected through correlation analysis， and photovoltaic data is classified into three typical weather conditions - sunny， cloudy， and rainy by using the S-Kmeans co-clustering method. Then， the IALA is employed to adaptively optimize the VMD parameters， enabling optimal decomposition of the photovoltaic power series and capturing local signal features more effectively. Finally， a TCN-BiGRU model is constructed for each subsequence， and the prediction result is obtained through component forecasting and global reconstruction， thereby improving prediction accuracy. Experimental results show that the proposed model outperforms the comparison models across all performance metrics under various weather conditions， validating its effectiveness in short-term photovoltaic power forecasting.

To effectively identify and eliminate abnormal data in the measured data of wind turbines, an anomaly detection algorithm based on manifold learning is proposed through the analysis of high-dimensional measured data from wind turbines. Firstly, the k-nearest neighbor mutual information algorithm is employed to select feature variables for the wind turbine. Subsequently, an optimized t-distributed stochastic neighbor embedding (t-SNE) algorithm is utilized. This optimized algorithm replaces the sample distance metric with a weighted sum of the Euclidean distance and the local principal component analysis (LPCA) difference, enabling the extraction of low-dimensional features with inherent patterns from the high-dimensional manifold data. This facilitates the distinct separation of data with different distribution characteristics in a visualized two-dimensional space. Furthermore, the density-based spatial clustering of applications with noise (DBSCAN) algorithm is applied to cluster the data within this two-dimensional space. The results demonstrate that, compared to the principal component analysis (PCA) algorithm, locally linear embedding (LLE) algorithm, and the original t-SNE algorithm, the proposed method can effectively achieve visual separation and clustering for data under various complex operating conditions, successfully identifying and eliminating abnormal data.

The load fluctuation and flow rate change caused by the tail water pressure and head fluctuation of the hydropower unit will generate external perturbation to the speed control system of the hydropower unit，which makes the speed control of the hydropower unit have nonlinear dynamic characteristics，interferes with the normal low-frequency oscillation suppression，and fails to effectively reduce the amplitude of the low-frequency oscillation of the hydropower unit. Therefore，a low-frequency oscillation suppression method based on the backstepping sliding mode algorithm is proposed for the speed control of hydropower units. First of all，the external disturbing factors such as water current speed change and load fluctuation are quantitatively analyzed to identify the nonlinear dynamic speed regulation state of the hydropower unit. Low-frequency oscillation parameters such as oscillation amount，oscillation amplitude，and oscillation frequency are detected in the nonlinear dynamic speed regulation state as the initial data for oscillation suppression. Then，taking the identification results of the speeding state of the hydroelectric unit and the detection results of the low-frequency oscillation parameters as input values，the “sliding mode variable structure” technology is introduced into the speed regulation state based on the backstepping sliding mode algorithm，so as to make the speeding state converge to the predetermined sliding mode surface within a limited time，and generate the backstepping sliding mode control law，which can be used to compensate for the nonlinear state of the unit and the external disturbances，and to reduce the low-frequency oscillations to suppress nonlinear disturbances. Finally，the speed control parameters are adaptively adjusted to realize the low-frequency oscillation suppression of hydroelectric unit speed control. The test results show that the method can significantly reduce the low-frequency oscillation amplitude of the hydropower unit with and without external disturbances，and the fluctuation coefficient of the low-frequency oscillation amplitude oscillation attenuation rate can be controlled to be less than 0.1 under the influence of external disturbances，which means that it has a better effect of disturbance suppression and nonlinear dynamic adaptation.

High-penetration renewable energy systems exhibit pronounced uncertainty. As an emerging long-duration physical energy storage technology，advanced adiabatic compressed air energy storage（AA-CAES）provides valuable support for enhancing system flexibility and regulation capability. However，conventional robust planning typically adopts conservative configurations across all scenarios，making it difficult to accurately characterize the risk of power and energy limit violations in storage operation. To address this gap，this study proposes an AA-CAES capacity optimization method that incorporates wind-photovoltaic uncertainty and achieves an effective trade-off between economic performance and operational risk through chance constraints. First，a chance-constrained model is developed to bound the violation probabilities of AA-CAES charging/discharging power and energy capacity at prescribed confidence levels，and binary variables combined with a big-M linearization strategy are employed to reformulate the problem as a mixed-integer linear program（MILP）. Second，a multi-scenario stochastic planning framework is constructed to represent the temporal variability of renewable resources. Finally，simulation studies and confidence-level sensitivity analyses are conducted. The results demonstrate that，compared with stochastic planning without chance constraints，the proposed method effectively controls violation risk while maintaining superior system cost performance，thereby enhancing both reliability and economic efficiency.

The LCL filter is usually constructed by centrally deploying filter capacitors on the AC side of the megawatt multi-module converter. The stability of the system is affected by the communication delay between the control device and the power module and the number of power modules，and the traditional active damping design method is difficult to ensure stable operation. Therefore，a new design method of active damping was proposed. In this method，the equivalent active damping control model of the converter considering the communication transmission delay was established in the discrete domain，and the equivalent transfer function root locus method was introduced. The active damping feedback coefficient and quasi-resonant controller parameters were designed under different communication delays and power modules，and their influence on the selection and design of converter parameters was analyzed. The effectiveness of the proposed method was verified by simulation and experiment. The active damping design can quickly obtain the control parameters required for the stable operation of the high-power converter，thereby improving the efficiency of equipment development.

With the continuous increase in the grid-connected capacity of distributed photovoltaics in rural distribution networks， it has become extremely urgent to achieve the flexible transformation of distributed photovoltaics， enabling them to be measurable， adjustable， and controllable. Based on this backdrop， this paper proposes a rolling approximate dynamic programming （ADP） model for rural distribution networks under the process of distributed photovoltaic flexible transformation. Firstly， taking into account the process of distributed photovoltaic flexible transformation， based on the concept of “multi-stage planning and first-stage implementation”， a rolling multi-stage stochastic programming （MSSP） model for rural distribution networks is established. The MSSP model comprehensively considers various factors such as the construction progress of distributed photovoltaics and the dynamic changes in power demand in rural areas at different stages. Secondly， using the ADP algorithm based on the Markov decision process （MDP） as the core， a rolling ADP algorithm is developed. This algorithm can iteratively optimize the decision-making process in each stage， thereby realizing the rolling solution of the proposed planning model. Through this approach， the model can adapt to the changing scenarios in the process of rural distribution network planning and distributed photovoltaic development. Finally， the improved IEEE 33-bus typical system is employed to validate the proposed model and algorithm. The simulation results demonstrate that the proposed model can effectively address the challenges in the process of distributed photovoltaic flexible transformation in rural distribution networks. It can obtain an optimal configuration plan that not only offers better economic benefits but also exhibits a stronger ability to handle the continuously emerging long-term uncertainties.

Accurately predicting the low output of wind power is the key to ensuring the power supply security of a high-proportion new energy power system. To this end， a parallel prediction method for wind power low output based on gate recurrent unit -denoising auto encoder（GRU-DAE）-DLinear is proposed. The unsupervised learning method is adopted to characterize the typical fluctuation characteristics of low output， and the prediction accuracy is improved through targeted modeling. Firstly， a low-output event classification method based on GRU-DAE is proposed， and the sequential data denoising， induction and reconstruction capabilities of temporal neural networks are utilized to identify typical low-output events. Then， a parallel prediction model for low-output events based on DLinear is established， independently modeling the timing characteristics of different types of low-output events， thereby improving the overall prediction accuracy. Finally， the effectiveness of the proposed method is verified based on the actual operation data of a wind farm in northern China.

Shared energy storage can effectively address the issues of low utilization and high costs caused by individual energy storage configurations by regulating resources across multiple regions. To further exploit the potential of shared energy storage in demand-side resources，this paper introduces electric vehicles and ice storage air conditioning，both with flexible energy storage characteristics，to construct a generalized shared energy storage model for the coordinated optimization of energy usage in smart building clusters. In response to the uncertainty of photovoltaic（PV）output on the energy input side，a time generative adversarial networks（TimeGAN）is employed to simulate a large number of PV output scenarios. By combining daily irradiance data，the static and dynamic features of these scenarios are mined，and typical scenarios are identified using K-medoids clustering. Additionally，a tiered carbon trading mechanism is introduced to limit the carbon emissions of the energy system. An optimization scheduling model for smart buildings is established，considering operational costs，carbon emissions，and user comfort，and is solved using CPLEX. Case studies demonstrate that the proposed method can generate high-quality PV output scenarios，improve regional PV consumption rates，and effectively balance user comfort and costs.

With the continuous development of the power system and the increasing awareness of environmental protection， the proportion of renewable energy generation in the power system is constantly increasing， and the scheduling of single thermal power generation units has become a coordinated scheduling mode for multi-energy generation. To solve the scheduling optimization problem of multi-energy power systems with energy storage devices， this paper establishes a multi-energy power system scheduling model of wind-solar-thermal-energy storage battery-pumped storage with the goal of minimizing system generation costs and pollution emissions. This paper introduces an adaptive strategy based on the number of iterations to optimize position update factor. Gaussian mutation is used to perturb the algorithm population， and the elite strategy in the ant lion algorithm is combined with the grasshopper algorithm to solve the proposed scheduling model using an improved multi-objective grasshopper algorithm. Real examples are simulated and analyzed on the Matlab platform， and an optimal multi-objective power system scheduling scheme is proposed. Through simulation analysis of test functions and simulation examples， the superiority of the improved algorithm and the rationality of the established model are verified.

Aiming at the problem of control interference and equipment loss caused by high frequency power electronic switching action when reconfigurable battery energy storage system participates in the frequency modulation process of power grid， a frequency modulation control strategy based on coordinated topology structure is proposed. Firstly， the operation control method of the reconfigurable battery energy storage system is designed to improve the cycle service life， flexibility and security of the battery energy storage system. Secondly， the virtual synchronous generator control is used to provide frequency modulation service. In order to reduce the influence of high-frequency power electronic switching， a reconfigurable battery energy storage system is proposed to participate in frequency modulation control strategy to ensure frequency stability. Finally， the effectiveness of the proposed control strategy is verified by simulation modeling.

To address the issue that distributed energy storage is difficult to meet the online real-time dispatching requirements of aggregators due to its large quantity，geographical dispersion，and strong uncertainty in responding to the demands of multiple user entities，a clustering-based aggregation optimization dispatching strategy for distributed energy storage is proposed. Firstly，the holographic state model of energy storage is established by considering the physical state parameters and electrical location information of energy storage，based on which the energy storage is clustered by K-Means++ algorithm. Secondly，the frequency modulation，peak regulation，distributed energy trading and voltage regulation demands of multi-user subjects are constructed，and the corresponding service consolidated indicators of energy storage clusters are designed to determine the collection of energy storage with excellent performance to be optimised for each demand. Subsequently，taking into account the revenue of the aggregator participating in demand response of multi-user subjects and the cost of leasing energy storage，the energy storage to be optimised is optimally scheduled with the objective of optimal economic benefit of the aggregator. Finally，the feasibility of this paper’s optimal scheduling strategy for aggregation is verified through a simulation experiment.