Energy storage fast compensation method and system based on track catenary voltage fluctuation feature recognition

Abstract

The application discloses a kind of energy storage fast compensation method and system based on track contact network voltage fluctuation feature identification, it is related to track traffic traction power supply technical field.The method includes: synchronous acquisition contact network voltage and current signal, extract the composite feature vector containing disturbance root attribute and transient change rate simultaneously;Based on the vector, through the prediction model of fusing dynamic phasor analysis and feedforward neural network, real-time output covers the transient energy shortage sequence of future;According to the sequence, in combination with the principle of nearest compensation and voltage recovery state, generate the collaborative control strategy of fusion overshoot and active damping to drive energy storage unit along line;Strategy is executed in advance and compensation energy is injected, finally based on voltage residual error carries out double closed-loop correction.The application solves the problems of response lag, unclear disturbance identification, poor collaboration and lack of adaptive ability in the prior art, realizes the fast suppression of contact network voltage fluctuation in advance, accurately, collaboratively and self-optimizing.

Description

Technical Field

[0001] This invention relates to the field of rail transit traction power supply technology, specifically to a rapid energy storage compensation method and system based on the identification of voltage fluctuation characteristics of the rail contact network. Background Technology

[0002] Rail transit, especially electrified railways, is characterized by highly volatile, random, and massive traction loads. During startup, acceleration, and transitions through phase-breaking zones, electric locomotives generate enormous power surges within milliseconds to seconds, causing rapid and severe voltage drops or oscillations in the traction contact network. These voltage fluctuations not only affect the stable operation of the locomotive itself but, in severe cases, can trigger protective devices, leading to train interruptions and adversely impacting other sensitive equipment in the power grid.

[0003] To address the aforementioned issues, existing technologies typically incorporate energy storage systems into traction power supply systems for reactive or active power compensation. However, traditional energy storage compensation methods are mostly based on passive responses to current or existing disturbances. For example, triggering compensation by detecting drops in the effective voltage value inherently introduces a response delay, making it difficult to handle instantaneous voltage surges on the microsecond to millisecond scale. Furthermore, traditional methods often treat voltage fluctuations as a single phenomenon, failing to differentiate between their underlying physical causes, resulting in a lack of targeted compensation strategies and poor compensation effectiveness. Simultaneously, in systems containing multiple distributed energy storage units, the lack of efficient collaborative control mechanisms can easily lead to unreasonable allocation of compensation power, significant energy transmission losses, and even the potential for new voltage oscillations due to asynchronous compensation actions. The compensation process also lacks adaptive exit and correction mechanisms, posing a risk of overcompensation or undercompensation, making it difficult to achieve rapid and stable voltage recovery.

[0004] Therefore, the main shortcomings of existing technologies can be summarized as follows: slow response, unclear identification of disturbance root causes, poor coordination among multiple energy storage units, and lack of adaptive optimization capability in the compensation process. Based on these shortcomings, the technical problem this invention aims to solve is: how to provide a rapid energy storage compensation method and system that can proactively, accurately, collaboratively, and adaptively suppress voltage fluctuations in the rail contact network and achieve continuous system optimization. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a method and system for rapid compensation of energy storage based on the identification of voltage fluctuation characteristics of railway contact network, so as to solve the problems of untimely, inaccurate, inefficient and prone to secondary fluctuations in contact network voltage compensation caused by response lag, unclear identification of disturbance root causes, poor coordination of multiple energy storage units and lack of adaptive optimization capability.

[0006] To achieve the above objectives, the present invention adopts the following technical solution;

[0007] A rapid energy storage compensation method based on the identification of voltage fluctuation characteristics of railway overhead contact lines includes the following steps:

[0008] Simultaneously acquire contact network voltage and current signals, process the voltage signals to extract a composite feature vector that simultaneously contains disturbance root cause attributes and transient rate of change;

[0009] Based on composite eigenvectors, a model that integrates dynamic phasor analysis and feedforward prediction is used to output in real time a transient energy deficit sequence that covers the ultra-short-term future caused by different perturbation sources.

[0010] Based on the transient energy deficit sequence, and combined with the principle of nearby compensation and the system voltage recovery state, a collaborative control strategy integrating overshoot compensation and active damping is generated to drive the energy storage units distributed along the line.

[0011] The coordinated control strategy is implemented in advance, and compensating energy is injected into the overhead contact line. Dual correction is performed based on the compensated voltage residual.

[0012] As a further aspect of the present invention, the processing of the voltage signal to extract a composite feature vector that simultaneously contains the root cause attributes of the disturbance and the transient rate of change includes:

[0013] Wide-area synchronous signal acquisition was carried out at multiple key locations in the traction power supply system, including traction substations, sectioning stations, and monitoring points along the line. All measuring equipment used a clock synchronization protocol for time synchronization.

[0014] Identify different types of perturbations and construct perturbation root source feature vectors;

[0015] Design a fifth-order asymmetric differential filter to calculate the derivatives of the voltage signal and form a differential eigenvector;

[0016] The perturbation root feature vector and the differential feature vector are concatenated and standardized to output a composite feature vector.

[0017] As a further aspect of the present invention, the step of identifying different root cause perturbation types and constructing perturbation root cause feature vectors includes:

[0018] The CEEMDAN algorithm, which is an adaptive noise complete set empirical mode decomposition algorithm, is used to decouple the complex voltage signal into a series of independent intrinsic mode function (IMF) components. The decomposed IMF components are mapped to the physical disturbance sources they represent, thus forming a preliminary identification of the disturbance type.

[0019] The energy and center frequency of each IMF component are calculated. Based on the principle that the energy amplitude is significantly higher than the background noise level and the frequency falls in the typical power disturbance frequency band, one or more key IMF components representing the main power disturbance are selected. The selected key IMF components representing the main power disturbance are used as processing objects, and the energy ratio and center frequency of each key IMF component are extracted as core features.

[0020] Finally, the energy percentages and center frequencies of all key IMF components are concatenated in a predetermined order to form a one-dimensional, structured perturbation root cause feature vector.

[0021] As a further aspect of the present invention, the step of outputting in real time a transient energy deficit sequence covering the ultra-short-term future caused by different perturbation sources through a model that integrates dynamic phasor analysis and feedforward prediction includes:

[0022] Based on the dynamic phasor analysis method, the instantaneous change trajectory of the amplitude and phase of the key IMF components is tracked. Combined with the equivalent impedance relationship of the system, the instantaneous power fluctuation caused by the root cause of the disturbance is quickly estimated, and a preliminary estimate of the energy deficit is obtained.

[0023] The composite feature vector is input into a trained feedforward neural network (FNN) model to predict future transient energy deficit sequences.

[0024] A multi-scale fusion strategy is adopted to fuse the preliminary energy deficit estimate with the prediction value of the feedforward neural network (FNN) model to generate a complete transient energy deficit sequence covering the present and the ultra-short-term future.

[0025] The virtual impedance value, which reflects the current strength of the system, is calculated in real time and introduced as an additional input into the feedforward neural network (FNN) model to achieve adaptive correction of the model.

[0026] As a further aspect of the present invention, a multi-scale fusion strategy is adopted: the current and near-future moments are mainly based on instantaneous estimates, the far-future moments are mainly based on neural network predictions, and the intermediate transition periods are smoothed by linear weighting. Finally, a complete transient energy deficit sequence covering the current and ultra-short-term future is generated. The sequence is based on time, clearly defining the energy deficit value and trend at each moment.

[0027] As a further aspect of the present invention, the method for generating a coordinated control strategy that integrates overshoot and active damping by combining the principle of proximity compensation with the system voltage recovery state is as follows:

[0028] A compensation power allocation model is constructed based on electrical distance, and compensation power is allocated to each energy storage unit according to the inverse weight of distance, ensuring that the energy storage unit closest to the disturbance point undertakes the main compensation task.

[0029] The formula for calculating the power allocation of an energy storage unit is:

[0030] ;

[0031] For the first The power distribution of each energy storage unit; The total compensation power required to meet the energy deficit at this disturbance point; For the first The electrical distance from each energy storage unit to the disturbance point; It is the sum of the reciprocals of the electrical distances from all energy storage units to the disturbance point;

[0032] Define and calculate the recovery status index. When the recovery status index is greater than the preset recovery status threshold, start the compensation exit procedure.

[0033] A fusion control strategy combining overshoot compensation and active damping is adopted to generate complete execution commands for each energy storage unit.

[0034] As a further aspect of the present invention, the recovery state index comprehensively considers the current voltage value and the voltage recovery rate. When the voltage recovers to the preset recovery state threshold and the recovery rate is relatively fast, it is determined that the power grid has self-healing capability, and the compensation exit procedure is initiated to prevent excessive voltage caused by overcompensation.

[0035] Recovery status indicators The formula is expressed as:

[0036] ;

[0037] This represents the current actual voltage of the overhead contact line. This is the rated voltage of the overhead contact line; This represents the voltage recovery ratio; For voltage recovery rate, This represents the change in the contact network voltage. The time interval is the time difference between two voltage sampling operations performed by the system. , This is a weighting coefficient used to balance the influence of the two parameters, and it is fixed after calibration based on actual working conditions.

[0038] As a further aspect of the present invention, the fusion control strategy combining overshoot compensation and active damping includes:

[0039] Overshoot compensation: Set an initial compensation power greater than the predicted value of the transient energy deficit sequence, and use the excess power to form a strong drive to push the voltage to quickly recover to a reasonable range;

[0040] The formula for calculating the initial overshoot compensation power is:

[0041] ;

[0042] This is the initial overshoot compensation power; The overshoot coefficient is calibrated according to the inertial characteristics of different power grid sections to ensure that it can quickly overcome inertia without causing excessive shock. Predict the deficit value at the current moment in the transient energy deficit sequence;

[0043] To avoid voltage overshoot due to continuous overshoot, the initial overshoot decays exponentially, as shown by the following formula:

[0044] Overshoot compensation power The calculation formula is:

[0045] ;

[0046] For overshoot compensation power; To compensate for the time; The decay time constant;

[0047] Active damping: By real-time detection of the system voltage frequency characteristics, oscillation signal parameters are identified. For each oscillation signal parameter, a damping current component with opposite phase and matching amplitude is immediately generated. The calculation formula is:

[0048] ;

[0049] This is the damping current component; The amplitude of the damping current; It is the oscillation angular frequency; This is the initial phase of the oscillation; This is the phase offset, ensuring that the damping current and the oscillation voltage are strictly out of phase.

[0050] While controlling each energy storage unit to output compensation power, the corresponding damping current component is injected simultaneously to actively offset oscillation energy, suppress the expansion of oscillation amplitude, and ensure a smooth and fluctuation-free voltage recovery process.

[0051] Finally, the proportion of compensation power allocated to each energy storage unit is combined with the initial power of overshoot compensation, the decay rate, and the current injection parameters of active damping to generate a complete execution command for each energy storage unit. At the same time, the recovery state threshold is associated to clarify the adjustment rhythm and exit timing of the compensation power.

[0052] As a further aspect of the present invention, the step of implementing the advanced coordinated control strategy and injecting compensating energy into the contact network, and performing dual correction based on the compensated voltage residual includes:

[0053] By utilizing the future time data in the transient energy deficit sequence, and the future time deficit data inherent in the sequence itself, compensation commands are issued in advance to offset the inherent delays in signal transmission, calculation and processing, and equipment operation, so as to achieve compensation in place before the disturbance occurs.

[0054] Rapid exit is achieved through a dual confirmation mechanism of hardware circuits and software algorithms. When the contact network voltage reaches the preset exit voltage threshold, the compensation power output is immediately cut off.

[0055] A dual closed-loop mechanism combining instantaneous and long-term corrections is employed to correct the system from two dimensions: instantaneous compensation accuracy and long-term prediction accuracy. This includes:

[0056] Instantaneous correction: If the accumulated energy of the voltage residual exceeds the threshold after compensation is completed, a supplementary compensation pulse with a very short duration is immediately triggered.

[0057] Long-term correction: The error between the predicted energy deficit and the actual energy deficit is continuously compared, and the parameters of the feedforward neural network (FNN) model are adjusted online using a recursive least squares method with a forgetting factor.

[0058] The second aspect of this application provides a system for a rapid energy storage compensation method based on the identification of voltage fluctuation characteristics of railway overhead contact lines, comprising:

[0059] The signal acquisition and feature extraction module is used to simultaneously acquire contact network voltage and current signals and extract composite feature vectors.

[0060] The energy deficit prediction module is used to output a transient energy deficit sequence based on the composite feature vector by using a model that integrates dynamic phasor analysis and feedforward prediction.

[0061] The collaborative control strategy generation module is used to generate a collaborative control strategy that integrates overshoot and active damping based on the transient energy deficit sequence, the nearest compensation principle and the system voltage recovery state.

[0062] The compensation execution and correction module is used to execute the cooperative control strategy in advance, inject compensation energy, and perform dual closed-loop correction based on the voltage residual.

[0063] In addition, multiple energy storage units distributed along the line, driven by the aforementioned coordinated control strategy, inject compensating energy into the overhead contact line.

[0064] In summary, due to the adoption of the above technical solution, the beneficial technical effects of the invention are as follows:

[0065] First, this invention fundamentally improves the speed and foresight of compensation. Through wide-area synchronous high-speed sampling and composite feature vector extraction, the system can accurately capture microsecond-level voltage transient details and their physical origins. More importantly, by utilizing a predictive model that integrates dynamic phasor analysis and feedforward neural networks, the system can output a transient energy deficit sequence covering the ultra-short-term future, thereby achieving proactive compensation before disturbances occur, effectively overcoming the response lag problem caused by the inherent delays in detection, calculation, and equipment operation in traditional methods.

[0066] Secondly, this invention significantly enhances the accuracy, coordination, and stability of compensation. By introducing a proximity-based compensation principle based on electrical distance, the system can intelligently allocate the main compensation tasks to the energy storage unit closest to the disturbance point, greatly reducing energy transmission losses and delays and improving overall compensation efficiency. Simultaneously, the integrated overshoot compensation and active damping coordinated control strategy not only rapidly overcomes grid inertia to accelerate voltage recovery through initial overshoot power but also actively suppresses potential oscillations by injecting damping current in real time, ensuring a fast and stable compensation process. Combined with intelligent judgment of the system voltage recovery state, timely withdrawal of compensation power is achieved, effectively preventing overcompensation.

[0067] Finally, this invention endows the system with continuous self-optimization capabilities. Through a dual closed-loop correction mechanism combining instantaneous correction based on voltage residuals and long-term correction based on online adjustment of predictive model parameters, the system can not only refine itself after a single compensation to eliminate steady-state deviations, but also continuously revise its predictive model using operational data, enabling it to adapt to changes in grid structure, load characteristics, and other factors. This transforms the entire system from a static device into an intelligent system with learning and evolutionary capabilities, maintaining excellent compensation performance over the long term.

[0068] In summary, this invention, through the organic combination of feature recognition, advanced prediction, collaborative control, and closed-loop correction, forms a complete solution that is rapid in response, precise in action, highly efficient in collaboration, and possesses self-evolutionary capabilities, ultimately achieving efficient, rapid, and stable suppression of voltage fluctuations in the track contact network. Attached Figure Description

[0069] Figure 1 This is a flowchart of a rapid energy storage compensation method based on the identification of voltage fluctuation characteristics of the rail contact network.

[0070] Figure 2 The flowchart is for S100, a rapid energy storage compensation method based on the identification of voltage fluctuation characteristics of the rail contact network.

[0071] Figure 3 The flowchart is for the S200 method of rapid energy storage compensation based on the identification of voltage fluctuation characteristics of the rail contact network.

[0072] Figure 4 The flowchart shows the S300 method for rapid energy storage compensation based on the identification of voltage fluctuation characteristics of the rail contact network.

[0073] Figure 5 This is a flowchart of the S400 method for rapid energy storage compensation based on the identification of voltage fluctuation characteristics of the railway overhead contact system. Detailed Implementation

[0074] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0075] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0076] The core objective of this invention is:

[0077] Construct an intelligent energy storage compensation system with advanced sensing, precise identification, coordinated action, and self-evolution capabilities. This system can identify the physical nature of overhead contact line voltage disturbances at their source and predict future energy demands in advance. Based on this, it drives energy storage units distributed along the line to perform rapid, precise, and coordinated compensation, ultimately achieving rapid and stable recovery of the overhead contact line voltage. At the same time, it ensures that the entire system can continuously optimize during long-term operation and adapt to complex and ever-changing operating conditions.

[0078] The core idea of ​​this method is:

[0079] 1. Feature-driven approach: This is the foundation and starting point of the entire method. Traditional methods often only focus on changes in voltage amplitude, while this invention delves into the physical essence and instantaneous dynamics behind the fluctuations.

[0080] Specifically, this is achieved by using CEEMDAN signal decomposition technology to decouple voltage fluctuations into intrinsic mode functions (IMF) components of different frequency bands, identifying the root causes of disturbances such as high-frequency switching noise, mid-frequency power surges, and low-frequency oscillations. Simultaneously, a fifth-order asymmetric differential filter is used to extract the derivatives of voltage changes, characterizing the transient rate of change of the fluctuations. Finally, these two methods are fused into a composite feature vector. This allows the system to go beyond simply observing voltage drops, understanding the causes and dynamic processes that lead to these drops, providing unprecedented depth of information for subsequent precise decision-making.

[0081] 2. Proactive prediction: The key to achieving compensation in advance and overcoming the inherent delay of the system is to extend compensation control from the present to the future.

[0082] Specifically, this is achieved by utilizing a fusion model to process composite feature vectors. On one hand, based on the physical model of dynamic phasor analysis, rapid instantaneous power estimation of key perturbation components is performed, ensuring a fast response at the current moment. On the other hand, a data-driven feedforward neural network (FNN) is used to learn the complex mapping relationship between perturbation features and future energy demand in historical data, directly predicting the transient energy deficit sequence in the next few milliseconds. After the two are fused, the system not only knows how much energy is currently lacking, but also predicts how much energy will be lacking at each moment in the future, providing a precise timetable for proactive actions.

[0083] 3. Collaborative Control: The core of ensuring efficient and stable implementation of compensation actions. It solves the challenges of multi-unit collaborative operations and dynamic process optimization.

[0084] Specifically, the controller generates a fusion strategy based on the predicted sequence and the system state.

[0085] Spatial coordination: Based on the principle of proximity compensation, the compensation power of each energy storage unit is allocated according to the reciprocal of the electrical distance, so that the unit closest to the disturbance point undertakes the main task and minimizes transmission loss and delay.

[0086] Time coordination: A fusion strategy of overshoot compensation and active damping is adopted. Overshoot compensation quickly overcomes grid inertia by applying an initial excess power and decaying exponentially, solving the speed problem; active damping generates reverse current in real time to actively suppress oscillations that may be caused during the compensation process, solving the stability problem.

[0087] Intelligent shutdown: By intelligently assessing the grid's self-healing capability through recovery status indicators, compensation is promptly discontinued to prevent overcompensation.

[0088] 4. Closed-loop evolution: The core competitiveness that ensures long-term superior system performance and adaptability to change. It upgrades the system from a device with a fixed function to an intelligent agent capable of learning.

[0089] Specifically, it includes two closed loops: instantaneous and long-term.

[0090] Instantaneous closed loop: If a small voltage residual exists after compensation, a microsecond-level compensation pulse is immediately triggered for fine-tuning to ensure the accuracy of a single compensation.

[0091] Long-term closed-loop: The error between predicted and actual values ​​is continuously compared, and the parameters of the FNN prediction model are fine-tuned online using recursive least squares with a forgetting factor (FF-RLS). This allows the model to gradually forget outdated data, learn new operating conditions, and thus continuously improve prediction accuracy, adapting to long-term evolutions such as changes in line impedance and the commissioning of new locomotives.

[0092] In summary, the core idea of ​​this solution runs through the entire technical solution: based on deep feature cognition, guided by advanced prediction, and using collaborative control as a means, it ultimately achieves the system's self-optimization and evolution through a closed loop, forming a complete intelligent solution that combines speed, accuracy, stability, and adaptability.

[0093] like Figure 1 As shown, it illustrates an exemplary method for rapid energy storage compensation based on the identification of voltage fluctuation characteristics of rail contact networks, specifically including the following steps:

[0094] A rapid energy storage compensation method based on the identification of voltage fluctuation characteristics of railway overhead contact lines includes:

[0095] S100. Process the voltage signal to extract a composite feature vector that simultaneously contains the root cause attributes of the disturbance and the transient rate of change;

[0096] S200. By integrating dynamic phasor analysis and feedforward prediction models, it outputs in real time a transient energy deficit sequence covering the ultra-short-term future caused by different perturbation sources.

[0097] S300. Combining the principle of proximity compensation with the system voltage recovery state, a coordinated control strategy integrating overshoot and active damping is generated;

[0098] S400. The coordinated control strategy is executed in advance and compensation energy is injected into the overhead contact line; then, double correction is performed based on the compensated voltage residual.

[0099] In the rapid energy storage compensation method based on the identification of voltage fluctuation characteristics of the rail contact network, S100 integrates the physical model of dynamic phasor analysis with the data model of feedforward neural network, adopts a multi-scale fusion strategy to generate a transient energy deficit sequence covering the ultra-short-term future, and combines virtual impedance to realize model adaptive correction, thereby achieving accurate advanced energy prediction.

[0100] Please refer to Figure 2 The diagram illustrates a flowchart of S100, an exemplary method for rapid energy storage compensation based on voltage fluctuation characteristics identification of railway overhead contact lines. This step aims to: construct a feature information database capable of deeply characterizing voltage fluctuations. Raw voltage and current signals are acquired through wide-area synchronous signal acquisition; the CEEMDAN algorithm is used for signal decomposition and disturbance source identification; transient change features are extracted using a fifth-order asymmetric differential filter; and finally, the disturbance source feature vector and the differential feature vector are fused and standardized to form a composite feature vector that combines physical essence identification and dynamic characteristic characterization, providing a comprehensive and effective input basis for subsequent energy prediction.

[0101] The specific content includes:

[0102] S110. Wide-area synchronous signal acquisition: In order to accurately track the propagation process of disturbances in the power grid, it is necessary to measure at multiple key locations in the traction power supply system simultaneously. Key locations include traction substations that supply power to the entire section, section substations that are responsible for isolating different power supply arms, and monitoring points set at certain intervals along the railway line.

[0103] All measuring devices must be synchronized using a clock synchronization protocol to ensure that data from different locations are perfectly aligned in time, with synchronization errors controlled within 1 microsecond.

[0104] At each measurement point, the instantaneous values ​​of voltage and current need to be recorded at a speed of no less than 1 MHz. The instantaneous values ​​of voltage and current are recorded synchronously to ensure that the microsecond-level voltage change details caused by the rapid switching action of the locomotive's power electronic equipment can be captured. Because the switching action of the power electronic equipment on the locomotive is extremely fast, it may cause microsecond-level voltage change. Only high-speed sampling can capture the complete details of the change.

[0105] S120. Signal decomposition, disturbance source identification and key component screening: The CEEMDAN algorithm of adaptive noise complete set empirical mode decomposition is adopted to decouple the complex voltage signal into a series of independent intrinsic mode function (IMF) components. The IMF components obtained by decomposition are mapped to the physical disturbance source they represent, forming a preliminary identification of the disturbance type.

[0106] Subsequently, the energy and center frequency of each IMF component are calculated. Based on the principle that the energy amplitude is significantly higher than the background noise level and the frequency falls within the typical power disturbance frequency band, one or more key IMF components representing the main power disturbance are selected. The selected key IMF components representing the main power disturbance are used as processing objects, and the energy ratio and center frequency of each key IMF component are extracted as core features.

[0107] Subsequently, the energy percentages and center frequencies of all key IMF components are concatenated in a predetermined order to form a one-dimensional, structured perturbation root cause feature vector.

[0108] Formula for calculating the energy of IMF components:

[0109] ;

[0110] For the first Energy of one IMF component; This represents the number of sampling points; For IMF components in the th The instantaneous amplitude of each sampling point;

[0111] The formula for center frequency is:

[0112] ;

[0113] For the first The center frequency of each IMF component; Number of frequency points; For the first Frequency values ​​at each frequency point; For this IMF component at frequency Fourier transform coefficients at the location;

[0114] S130. Transient feature extraction: Design a fifth-order asymmetric differential filter. The filter calculates the derivatives of the voltage signal and forms a differential feature vector.

[0115] S140. Feature fusion combines the perturbation root cause feature vector and the differential feature vector to form a comprehensive composite feature vector. This composite feature vector is then standardized, and the final output standardized composite feature vector combines the ability to identify the physical root cause of the perturbation with the ability to represent instantaneous dynamic characteristics, providing comprehensive and effective input features for energy deficit prediction.

[0116] The composite feature vector is like a highly condensed perturbation fingerprint, containing all the key information about the current power grid state. Before being fed into the next prediction model, the key information is standardized, that is, adjusted according to a uniform scale to eliminate the influence of different features due to different units and magnitudes, so that the prediction model can process each feature more fairly and effectively.

[0117] In one possible implementation, the mapping relationship between the IMF component and the physical disturbance source it represents is as follows:

[0118] High-frequency IMF components: mainly correspond to the pulse noise and switching harmonics caused by the high-frequency switching action of the converter switching devices in electric locomotives; their characteristics are high frequency, short duration, and small energy per cycle.

[0119] The intermediate frequency IMF component corresponds to the high-power inrush current during locomotive startup, the transient process during phase transition, or the sudden switching of large-capacity loads. It is the most important and critical type of disturbance that causes significant voltage fluctuations and transient energy deficits in the power grid. Its characteristics are concentrated energy and large rate of change.

[0120] Low-frequency IMF components and residual components: These reflect low-frequency power oscillations during multi-locomotive coordinated operation, or slow effects transmitted from distant power grid fluctuations through system impedance.

[0121] In one possible implementation, the differential eigenvector includes:

[0122] The first derivative describes the rate of voltage change and reflects the steepness of the fluctuation. For example, the absolute value of the first derivative is larger during the initial shock.

[0123] The second derivative characterizes the acceleration of voltage change and reflects the trend of fluctuation. For example, the second derivative is negative during voltage drop, and the larger the absolute value, the more violent the drop trend.

[0124] Third-order and higher derivatives describe more subtle jitter and curvature characteristics in the process of change, and can effectively distinguish between abrupt disturbances and oscillatory disturbances. For example, the higher-order derivatives of oscillatory disturbances will show alternating positive and negative changes.

[0125] High-order differential filters maintain high accuracy in the effective voltage fluctuation frequency band, ensuring that the extracted differential features reflect the real disturbances rather than spurious features introduced by measurement errors.

[0126] In the rapid compensation method for energy storage based on the identification of voltage fluctuation characteristics of the track contact network, the S200 acquires the contact network voltage and current signals through wide-area synchronous signal acquisition, identifies the root cause of disturbance through CEEMDAN decomposition, extracts transient features through differential filter, and finally fuses them to form a composite feature vector containing physical essence and dynamic characteristics.

[0127] Please refer to Figure 3 The diagram illustrates a flowchart of S200, an exemplary method for rapid energy storage compensation based on the identification of voltage fluctuation characteristics in the railway overhead contact system. This step aims to achieve accurate prediction of ultra-short-term energy deficits based on composite feature vectors. By integrating rapid physical estimation through dynamic phasor analysis with data-driven prediction from feedforward neural networks, a multi-scale strategy is employed to generate a transient energy deficit sequence covering future moments. Furthermore, by combining virtual impedance with real-time correction of model parameters, an adaptive forward-looking energy prediction system is established, providing precise decision-making basis for advanced compensation control.

[0128] The specific content includes:

[0129] S210. Instantaneous power analysis based on dynamic phasors: The dynamic phasor analysis method is adopted. By representing time-varying voltage and current signals as complex phasors whose amplitude and phase change slowly with time, the instantaneous change trajectory of the amplitude and phase of the IMF component is tracked. Based on the dynamic phasor change of the IMF component and combined with the equivalent impedance relationship of the system, the instantaneous power fluctuation caused by the root cause of the disturbance is quickly estimated, and a preliminary estimate of the energy deficit with extremely fast response speed is obtained.

[0130] S220. Based on the prediction of the feedforward neural network, the composite feature vector is input into a feedforward neural network (FNN) model trained with massive historical data. The feedforward neural network (FNN) model directly predicts the transient energy deficit sequence at each moment in the next few milliseconds by learning the complex and nonlinear mapping relationship between the perturbation fingerprint and the future energy demand.

[0131] S230. Multi-scale prediction fusion: To combine the speed of physical models with the forward-looking advantages of data models, a multi-scale fusion strategy is adopted.

[0132] In one possible implementation, the multi-scale fusion strategy is as follows:

[0133] The current and near-future moments are mainly based on instantaneous estimates, while the far-future moments are mainly based on neural network predictions. The intermediate transition period uses a linear weighting method to smooth the transition weights, and finally generates a complete transient energy deficit sequence covering the present and the ultra-short-term future. The sequence is based on time, clearly defining the energy deficit value and trend at each moment.

[0134] Current moment: The instant when the disturbance occurs, i.e., the moment when data acquisition and feature extraction are synchronized;

[0135] Near future interval: The time interval from the current moment is ≤ 1 / 3 of the total forecast duration, which is the period closely connected to the current moment in the ultra-short-term forecast;

[0136] Far future interval: The time interval from the current moment is greater than 1 / 3 of the total forecast duration, which is the period far from the current moment in ultra-short-term forecasts.

[0137] The fusion strategy combines the advantages of both methods: it has the accuracy of current-moment data based on physical principles, and it also has the ability to predict future trends based on data, thereby generating a transient energy deficit sequence that is both accurate and comprehensive.

[0138] S240. Model adaptive correction: The state of the power grid is dynamic. Train movement and other situations can cause changes in the system impedance at the measurement point, which in turn affects the propagation characteristics of disturbances and the evolution of energy deficit.

[0139] To ensure that the prediction accuracy of the transient energy deficit sequence remains stable under various operating conditions, a virtual impedance value is calculated in real time as a state variable reflecting the current strength of the system. The virtual impedance value is introduced into the neural network prediction model in real time as additional input information, enabling the model to sense changes in the system state and adjust its prediction strategy accordingly, thereby maintaining high accuracy under various operating conditions and possessing good adaptive capabilities.

[0140] Please refer to Figure 4The diagram shows a flowchart of an exemplary energy storage rapid compensation method S300 based on the identification of voltage fluctuation characteristics of the rail contact network. S300 is responsible for converting the predicted energy demand into efficient, stable and coordinated action commands to drive multiple energy storage units to work in an orderly manner.

[0141] The purpose of this step is to transform the transient energy deficit sequence into precise and coordinated control commands, which ensures that each energy storage unit allocates compensation power in the optimal proportion, can exit in time after voltage recovery, suppresses the oscillation risk during the compensation process, and ultimately achieves rapid and stable recovery of the contact network voltage.

[0142] The specific content includes:

[0143] S310. The establishment of spatiotemporal coordination and local compensation has the core objective of solving the problem of how to efficiently allocate compensation power among multiple energy storage units, and avoiding the loss and delay caused by long-distance energy transmission.

[0144] A compensation power allocation model is constructed based on electrical distance. Electrical distance is represented by the impedance value of the contact network line between the energy storage unit and the disturbance point. The compensation power of each energy storage unit is allocated according to the inverse weight of distance to ensure that the energy storage unit closest to the disturbance location undertakes the main compensation task, minimizes energy transmission loss, and improves compensation efficiency.

[0145] During the process, when an energy deficit is predicted at a certain location, the system will allocate compensation power to each energy storage unit proportionally based on the electrical distance. The allocation formula is as follows:

[0146] ;

[0147] For the first The power distribution of each energy storage unit; The total compensation power required to meet the energy deficit at this disturbance point; For the first The electrical distance from each energy storage unit to the disturbance point; It is the sum of the reciprocals of the electrical distances from all energy storage units to the disturbance point;

[0148] The core logic of the formula is that the compensation power ratio of a given energy storage unit is inversely proportional to its electrical distance from the disturbance point. The smaller the electrical distance, the larger the weight ratio, and the more compensation power is allocated. Through this allocation strategy, energy storage units closer to the disturbance point receive a larger proportion of compensation power, which minimizes long-distance energy transmission in the grid, reduces line losses, shortens the transmission time of compensation energy, improves compensation speed and efficiency, and avoids compensation delays and losses caused by long-distance energy transmission.

[0149] S320. The specific implementation of intelligent judgment of recovery status is that more compensation is not better. It needs to be stopped in time after the voltage is restored to prevent overcorrection from causing the voltage to be too high and causing new fluctuations. Therefore, the core goal of this step is to solve the problem of when to stop compensation and avoid overcompensation leading to voltage overshoot or new fluctuations.

[0150] The system's recovery capability is determined by monitoring the voltage recovery rate, and recovery status indicators are defined to reflect the degree and trend of voltage recovery.

[0151] The recovery status index takes into account both the current voltage value and the voltage recovery rate. When the voltage recovers to the preset recovery status threshold and the recovery rate is relatively fast, it is determined that the power grid has self-healing capability, and the compensation exit procedure is initiated to prevent excessive voltage caused by overcompensation.

[0152] Recovery status indicators The formula is expressed as:

[0153] ;

[0154] This represents the current actual voltage of the overhead contact line. This is the rated voltage of the overhead contact line; This represents the voltage recovery ratio; For voltage recovery rate, This represents the change in the contact network voltage. The time interval is the time difference between two voltage sampling operations performed by the system. , This is a weighting coefficient used to balance the influence of the two parameters, and it is fixed after calibration based on actual working conditions.

[0155] When recovery status indicators When the voltage exceeds the preset recovery threshold, meaning it has recovered to a reasonable range close to the rated voltage and the recovery rate is relatively fast, it indicates that the power grid has a strong self-healing capability and there is no need to continue injecting compensation energy. At this time, the system immediately initiates the compensation exit procedure, controlling each energy storage unit to gradually reduce the compensation power until it stops, fundamentally preventing the problem of excessively high voltage caused by overcompensation;

[0156] The core objective of the specific implementation of the S330 control strategy synthesis is to solve the problem of how to make the compensation action both fast and stable, taking into account both voltage recovery speed and system stability, overcoming grid inertia and oscillation risks, and ultimately outputting precise execution commands for each energy storage unit.

[0157] To achieve rapid and stable voltage recovery, a control strategy combining overshoot compensation and active damping is adopted, balancing compensation response speed and system stability. This approach quickly overcomes grid inertia while effectively suppressing oscillation risks.

[0158] In one possible implementation, a control strategy combining overshoot compensation and active damping is adopted. This strategy integrates overshoot compensation and active damping, with the two functions working synergistically and complementing each other. The specific implementation is as follows:

[0159] Overshoot compensation: Due to the inherent line inductance, capacitance and load inertia of the power grid, conventional compensation power is easily constrained by inertia, resulting in slow voltage recovery. In order to quickly overcome this constraint, an initial compensation power greater than the predicted value of the transient energy deficit sequence is set. The excess power forms a strong drive to push the voltage to quickly recover to a reasonable range.

[0160] The formula for calculating the initial overshoot compensation power is:

[0161] ;

[0162] This is the initial overshoot compensation power; The overshoot coefficient is calibrated according to the inertial characteristics of different power grid sections to ensure that it can quickly overcome inertia without causing excessive shock. Predict the deficit value at the current moment in the transient energy deficit sequence;

[0163] To avoid voltage overshoot due to continuous overshoot, the initial overshoot decays exponentially, as shown by the following formula:

[0164] Overshoot compensation power The calculation formula is:

[0165] ;

[0166] For overshoot compensation power; To compensate for the time; The decay time constant;

[0167] Active damping: During the compensation process, voltage oscillations are caused by factors such as fluctuations in grid parameters, asynchronous operation of multiple energy storage units, and sudden changes in disturbance type. If these oscillations are not suppressed in time, they will lead to aggravated voltage fluctuations. By real-time detection of the frequency characteristics of the system voltage, the oscillation signal parameters are identified. For the oscillation signal parameters, a damping current component with opposite phase and appropriate amplitude is immediately generated. The calculation formula is as follows:

[0168] ;

[0169] This is the damping current component; The amplitude of the damping current; It is the oscillation angular frequency; This is the initial phase of the oscillation; This is the phase offset, ensuring that the damping current and the oscillation voltage are strictly out of phase.

[0170] While controlling each energy storage unit to output compensation power, the corresponding damping current component is injected simultaneously to actively offset oscillation energy, suppress the expansion of oscillation amplitude, and ensure a smooth and fluctuation-free voltage recovery process.

[0171] Finally, the proportion of compensation power allocated to each energy storage unit is combined with the initial power of overshoot compensation, the decay rate, and the current injection parameters of active damping to generate a complete execution command for each energy storage unit. At the same time, the recovery state threshold is associated to clarify the adjustment rhythm and exit timing of the compensation power.

[0172] Please refer to Figure 5 The diagram illustrates an exemplary flowchart of an energy storage rapid compensation method S400 based on the identification of voltage fluctuation characteristics of the rail contact network. S400 relies on the transient energy deficit sequence output by S200, and offsets the inherent delay of the system by executing ahead of time, quickly exits to avoid the risk of overcompensation, and corrects the instantaneous deviation and optimizes the long-term model through dual closed-loop correction to ensure that the compensation command is accurately implemented, while realizing the continuous evolution of system performance.

[0173] The purpose of this step is to transform the cooperative control strategy generated by S300 into actual compensation actions, to quickly inject precise energy and suppress voltage fluctuations under the premise of compensation before disturbances occur; and at the same time, to improve the accuracy and adaptability of subsequent compensation through real-time correction and model self-evolution.

[0174] The specific content includes:

[0175] S410. Proactive execution: Utilizing the future time-of-flight deficit data inherent in the transient energy deficit sequence output by S200, compensation commands are issued in advance to offset the inherent delays in signal transmission, computation, and equipment operation, thus achieving compensation in place before the disturbance occurs.

[0176] The transient energy deficit sequence output by step S200 clearly marks the energy deficit values ​​and trends for multiple consecutive future moments with time as the axis. This sequence is an accurate prediction of energy demand in the ultra-short term, providing a core basis for proactive execution. Since there are inherent delays in signal transmission between the control system and the energy storage unit, processor instruction parsing, and converter power output, if real-time commands are issued based on the current deficit, disturbances will have already occurred when compensating for energy injection, making it difficult to suppress fluctuations at their source.

[0177] Step S420 uses the logic of predicting the time - total system delay = instruction issuance time to trigger the compensation instruction in advance;

[0178] S420. Quick Exit: An independent hard exit mechanism is set up, which ensures the reliability of the exit action through a dual confirmation mechanism of hardware circuit and software algorithm;

[0179] In one possible implementation, the dual verification mechanism of hardware circuitry and software algorithm includes:

[0180] At the software level: The current voltage of the contact network is monitored in real time, and a preset exit voltage threshold is set. When the current voltage of the contact network is greater than the preset exit voltage threshold, the software immediately triggers an exit command and controls the converters of all energy storage units to stop outputting compensation power.

[0181] Hardware level: Design an independent voltage monitoring circuit in the converter. When the contact network voltage reaches the exit threshold, the compensation power output can be directly cut off without software instructions, forming a hardware interlock.

[0182] The dual confirmation mechanism ensures that, regardless of whether the software misjudges or the parameters drift, the compensation action will stop immediately as long as the voltage recovers to near the rated value, thus completely avoiding excessive voltage or new fluctuations caused by overcompensation.

[0183] S430. Dual closed-loop correction: To achieve precise control of the compensation process and continuous optimization of the system, a dual closed-loop mechanism combining instantaneous correction and long-term correction is adopted. The system is closed-loop corrected from two dimensions: instantaneous compensation accuracy and long-term prediction accuracy, to ensure that the compensation effect is both fast and accurate, and has the ability to continuously evolve.

[0184] Instantaneous compensation accuracy: After the compensation action is completed, the system will check the effect to see if there is still a slight voltage deviation; if the energy accumulated by this residual exceeds the threshold, the system will immediately trigger a supplementary compensation pulse with a very short duration to quickly smooth out the last deviation.

[0185] Long-term prediction: The system continuously compares the error between the predicted energy deficit and the actual calculated energy deficit, and introduces a recursive least squares method with a forgetting factor to perform online, micro-adjustment of the parameters of the feedforward neural network prediction model in S200.

[0186] In recursive least squares with a forgetting factor, the core role of the forgetting factor is to exponentially decay the influence of historical data during the dynamic calculation of adjustment strength. This makes the parameter update process focus more on recent data that better reflects the current state of the system, while gradually forgetting outdated and potentially inapplicable historical experience, thereby ensuring that the model can track the slow changes in system characteristics.

[0187] The core logic of recursive least squares with a forgetting factor can be summarized as follows:

[0188] New model parameters = old model parameters + adjustment magnitude × (actual value - model prediction value);

[0189] Among them, the adjustment strength is a dynamically calculated vector, which is a gain vector dynamically calculated by the algorithm. Its calculation process is dominated by the forgetting factor, which determines the magnitude and direction of the parameter update.

[0190] Example 2:

[0191] A rapid energy storage compensation system based on the voltage fluctuation characteristics identification of the railway overhead contact system is used to implement the method described in Example 1. The system includes:

[0192] The signal acquisition and feature extraction module is configured to execute step S100. This module includes widely distributed synchronous measurement units deployed at key locations such as traction substations, sectioning stations, and monitoring points along the line. A clock synchronization protocol ensures a synchronization error of less than 1 microsecond, and the module synchronously acquires instantaneous voltage and current signals at a sampling rate of no less than 1 MHz. The module has a built-in signal processor that executes the CEEMDAN algorithm to decompose the voltage signal, calculates the energy and center frequency of each key IMF component to construct the disturbance root cause feature vector, runs a fifth-order asymmetric differential filter to extract the differential feature vector, and fuses and normalizes the two to finally output a composite feature vector.

[0193] An energy deficit prediction module is configured to execute step S200. This module receives the composite feature vector and integrates a dynamic phasor analysis unit and a feedforward neural network prediction unit. The dynamic phasor analysis unit rapidly estimates instantaneous power fluctuations based on the dynamic phasors of key IMF components to obtain a preliminary estimate of the energy deficit. The feedforward neural network prediction unit uses a trained model to predict the future transient energy deficit sequence based on the composite feature vector. This module also includes a multi-scale fusion unit to fuse the predictions from the two sources and introduces a virtual impedance value for adaptive correction, ultimately outputting a high-precision transient energy deficit sequence covering the ultra-short-term future.

[0194] The collaborative control strategy generation module is configured to execute step S300. This module receives the transient energy deficit sequence and includes three core sub-modules: a power allocation sub-module, which calculates the compensation power allocation ratio of each energy storage unit based on electrical distance; a recovery state judgment sub-module, which calculates the recovery state index in real time and compares it with a preset threshold to determine the timing of compensation exit; and a control strategy synthesis sub-module, which integrates overshoot compensation algorithm and active damping algorithm to generate precise control commands for each energy storage unit that combine speed and stability.

[0195] The compensation execution and correction module is configured to execute step S400. This module includes an instruction advance execution unit, which uses future energy deficit data to issue compensation instructions in advance to offset system delays; a fast exit unit, which uses a dual hardware and software confirmation mechanism to immediately cut off power output after voltage recovery to prevent overcompensation; and a dual closed-loop correction unit, which is responsible for performing residual pulse compensation for instantaneous fine correction and using a recursive least squares method with a forgetting factor to optimize the long-term parameters of the prediction model, thereby achieving system self-evolution.

[0196] Multiple energy storage units distributed along the line are connected to the contact network through power electronic converters. These units receive and execute control commands from the compensation execution and correction module to quickly and accurately inject or stop compensation energy into the grid, thereby jointly maintaining the stability of the contact network voltage.

[0197] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A rapid energy storage compensation method based on the identification of voltage fluctuation characteristics of railway overhead contact lines, characterized in that, Includes the following steps: Simultaneously acquire contact network voltage and current signals, process the voltage signals to extract a composite feature vector that simultaneously contains disturbance root cause attributes and transient rate of change; Based on composite eigenvectors, a model that integrates dynamic phasor analysis and feedforward prediction is used to output in real time a transient energy deficit sequence that covers the ultra-short-term future caused by different perturbation sources. Based on the transient energy deficit sequence, and combined with the principle of nearby compensation and the system voltage recovery state, a collaborative control strategy integrating overshoot compensation and active damping is generated to drive the energy storage units distributed along the line. The coordinated control strategy is implemented in advance, and compensating energy is injected into the overhead contact line. Dual correction is performed based on the compensated voltage residual.

2. The method for rapid energy storage compensation based on voltage fluctuation characteristics of rail contact networks according to claim 1, characterized in that, The process of processing the voltage signal to extract a composite feature vector that simultaneously contains the root cause attributes of the disturbance and the transient rate of change includes: Wide-area synchronous signal acquisition was carried out at multiple key locations in the traction power supply system, including traction substations, sectioning stations, and monitoring points along the line. All measuring equipment used a clock synchronization protocol for time synchronization. Identify different types of perturbations and construct perturbation root source feature vectors; Design a fifth-order asymmetric differential filter to calculate the derivatives of the voltage signal and form a differential eigenvector; The root feature vector of the disturbance and the differential feature vector are concatenated and standardized to output a composite feature vector.

3. The method for rapid energy storage compensation based on voltage fluctuation characteristics of rail contact networks according to claim 2, characterized in that, The steps of identifying different root cause perturbation types and constructing perturbation root cause feature vectors include: The CEEMDAN algorithm, which is an adaptive noise complete set empirical mode decomposition algorithm, is used to decouple the complex voltage signal into a series of independent intrinsic mode function (IMF) components. The decomposed IMF components are mapped to the physical disturbance sources they represent, thus forming a preliminary identification of the disturbance type. The energy and center frequency of each IMF component are calculated. Based on the principle that the energy amplitude is significantly higher than the background noise level and the frequency falls in the typical power disturbance frequency band, one or more key IMF components representing the main power disturbance are selected. The selected key IMF components representing the main power disturbance are used as processing objects, and the energy ratio and center frequency of each key IMF component are extracted as core features. The energy percentages and center frequencies of all key IMF components are concatenated in a predetermined order to form a one-dimensional, structured perturbation root cause feature vector.

4. The method for rapid energy storage compensation based on voltage fluctuation characteristics of railway overhead contact lines according to claim 1, characterized in that, The step of outputting in real time a transient energy deficit sequence covering the ultra-short-term future caused by different perturbation sources through a model that integrates dynamic phasor analysis and feedforward prediction includes: Based on the dynamic phasor analysis method, the instantaneous change trajectory of the amplitude and phase of the key IMF components is tracked. Combined with the equivalent impedance relationship of the system, the instantaneous power fluctuation caused by the root cause of the disturbance is quickly estimated, and a preliminary estimate of the energy deficit is obtained. The composite feature vector is input into a trained feedforward neural network (FNN) model to predict future transient energy deficit sequences. A multi-scale fusion strategy is adopted to fuse the preliminary energy deficit estimate with the prediction value of the feedforward neural network (FNN) model to generate a complete transient energy deficit sequence covering the present and the ultra-short-term future. The virtual impedance value, which reflects the current strength of the system, is calculated in real time and introduced as an additional input into the feedforward neural network (FNN) model to achieve adaptive correction of the model.

5. The rapid energy storage compensation method based on the identification of voltage fluctuation characteristics of rail contact network according to claim 4, characterized in that, The multi-scale fusion strategy is adopted: the current and near-future moments are mainly based on instantaneous estimates, the far-future moments are mainly based on neural network predictions, and the intermediate transition period is smoothed by linear weighting. Finally, a complete transient energy deficit sequence covering the current and ultra-short-term future is generated. The sequence is based on time, and the energy deficit value and trend at each moment are clearly defined.

6. The method for rapid energy storage compensation based on voltage fluctuation characteristics of rail contact network according to claim 1, characterized in that, The method for generating a coordinated control strategy that combines overshoot and active damping by combining the principle of proximity compensation with the system voltage recovery state: A compensation power allocation model is constructed based on electrical distance, and compensation power is allocated to each energy storage unit according to the inverse weight of distance, ensuring that the energy storage unit closest to the disturbance point undertakes the main compensation task. The formula for calculating the power allocation of an energy storage unit is: ； For the first The power distribution of each energy storage unit; To meet the total compensation power required to satisfy the energy deficit at the disturbance point; For the first The electrical distance from each energy storage unit to the disturbance point; It is the sum of the reciprocals of the electrical distances from all energy storage units to the disturbance point; Define and calculate the recovery status index. When the recovery status index is greater than the preset recovery status threshold, start the compensation exit procedure. A fusion control strategy combining overshoot compensation and active damping is adopted to generate complete execution commands for each energy storage unit.

7. The method for rapid energy storage compensation based on voltage fluctuation characteristics of rail contact network according to claim 6, characterized in that, The recovery status index comprehensively considers the current voltage value and the voltage recovery rate. When the voltage recovers to the preset recovery status threshold and the recovery rate is relatively fast, it is determined that the power grid has self-healing capability, and the compensation exit procedure is initiated to prevent excessive voltage caused by overcompensation. Recovery status indicators The formula is expressed as: ； This represents the current actual voltage of the overhead contact line. This is the rated voltage of the overhead contact line; This represents the voltage recovery ratio; For voltage recovery rate, This represents the change in the contact network voltage. The time interval is the time difference between two voltage sampling operations performed by the system. , This is a weighting coefficient used to balance the influence of the two parameters, and it is fixed after calibration based on actual working conditions.

8. The method for rapid energy storage compensation based on voltage fluctuation characteristics of rail contact network according to claim 6, characterized in that, The fusion control strategy combining overshoot compensation and active damping includes: Overshoot compensation: Set an initial compensation power greater than the predicted value of the transient energy deficit sequence, and use the excess power to form a strong drive to push the voltage to quickly recover to a reasonable range; The formula for calculating the initial overshoot compensation power is: ； This is the initial overshoot compensation power; The overshoot coefficient is calibrated according to the inertial characteristics of different power grid sections to ensure that it can quickly overcome inertia without causing excessive shock. Predict the deficit value at the current moment in the transient energy deficit sequence; To avoid voltage overshoot due to continuous overshoot, the initial overshoot decays exponentially, as shown by the following formula: Overshoot compensation power The calculation formula is: ； For overshoot compensation power; To compensate for the time; The decay time constant; Active damping: By real-time detection of the system voltage frequency characteristics, oscillation signal parameters are identified. For each oscillation signal parameter, a damping current component with opposite phase and matching amplitude is immediately generated. The calculation formula is: ； This is the damping current component; The amplitude of the damping current; It is the oscillation angular frequency; This is the initial phase of the oscillation; This is the phase offset, ensuring that the damping current and the oscillation voltage are strictly out of phase. While controlling each energy storage unit to output compensation power, the corresponding damping current component is injected simultaneously to actively offset the oscillation energy, suppress the expansion of oscillation amplitude, and ensure a smooth and fluctuation-free voltage recovery process. The compensation power ratio allocated to each energy storage unit is combined with the initial power, decay rate, and current injection parameters of the overshoot compensation to generate a complete execution command for each energy storage unit. At the same time, the recovery state threshold is associated to clarify the adjustment rhythm and exit timing of the compensation power.

9. The method for rapid energy storage compensation based on voltage fluctuation characteristics of railway overhead contact lines according to claim 1, characterized in that, The steps of implementing the advanced coordinated control strategy, injecting compensating energy into the overhead contact line, and performing dual correction based on the compensated voltage residual include: By utilizing the future time data in the transient energy deficit sequence, and the future time deficit data inherent in the sequence itself, compensation commands are issued in advance to offset the inherent delays in signal transmission, calculation and processing, and equipment operation, so as to achieve compensation in place before the disturbance occurs. Rapid exit is achieved through a dual confirmation mechanism of hardware circuits and software algorithms. When the contact network voltage reaches the preset exit voltage threshold, the compensation power output is immediately cut off. A dual closed-loop mechanism combining instantaneous and long-term corrections is employed to correct the system from two dimensions: instantaneous compensation accuracy and long-term prediction accuracy. This includes: Instantaneous correction: If the accumulated energy of the voltage residual exceeds the threshold after compensation is completed, a supplementary compensation pulse with a very short duration is immediately triggered. Long-term correction: The error between the predicted energy deficit and the actual energy deficit is continuously compared, and the parameters of the feedforward neural network (FNN) model are adjusted online using a recursive least squares method with a forgetting factor.

10. A system for implementing the rapid energy storage compensation method based on the identification of voltage fluctuation characteristics of rail contact networks as described in any one of claims 1-9, characterized in that, include: The signal acquisition and feature extraction module is used to simultaneously acquire contact network voltage and current signals and extract composite feature vectors; The energy deficit prediction module is used to output a transient energy deficit sequence based on the composite feature vector by using a model that integrates dynamic phasor analysis and feedforward prediction. The collaborative control strategy generation module is used to generate a collaborative control strategy that integrates overshoot and active damping based on the transient energy deficit sequence, the nearest compensation principle and the system voltage recovery state. The compensation execution and correction module is used to execute the cooperative control strategy in advance, inject compensation energy, and perform dual closed-loop correction based on the voltage residual. In addition, multiple energy storage units distributed along the line, driven by the aforementioned coordinated control strategy, inject compensating energy into the overhead contact line.

Images