A data processing method for multi-type adjustable resource collaborative control

Abstract

This invention relates to the field of power system technology, and more particularly to a data processing method for coordinated control of multiple types of adjustable resources. The method includes: constructing a multi-dimensional spatiotemporal feature sequence library by collecting and preprocessing operational data of multiple resource types; extracting core feature vectors reflecting transient coupling and quantifying the nonlinear correlation strength using dynamic manifold embedding technology; evaluating real-time safety margin based on feature vectors and trajectory sensitivity methods; generating a coordinated control instruction set with the goal of maximizing the margin using a multi-constraint heuristic search algorithm; and implementing closed-loop monitoring and dynamic instruction correction using predictive control theory. This invention achieves accurate characterization of nonlinear coupling mechanisms, optimizes multi-source response matching, and effectively improves the transient stability and operational safety of the system.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a data processing method for coordinated control of multiple types of adjustable resources. Background Technology

[0002] In modern energy transmission systems, hybrid AC / DC operation has become a core method for improving the efficiency of power grid resource allocation, and its operational safety and stability are directly related to the reliability of energy supply. Among these, flexible DC transmission technology and various types of energy storage systems serve as effective means to enhance the grid's regulation capabilities, and their connection location, capacity allocation, and response characteristics have a profound impact on the system's transient stability.

[0003] When dealing with the stability analysis of strongly coupled AC / DC systems, existing technologies often fail to fully characterize the highly nonlinear coupling characteristics after the integration of multiple types of resources using traditional physical mechanism models. This leads to biases in the quantitative assessment of the transient stability impact mechanism. Furthermore, for multi-source heterogeneous resource characteristic data, existing analysis methods lack effective dynamic embedding and multi-dimensional correlation analysis capabilities, making it impossible to accurately obtain the real-time safety margin of the system under extreme operating conditions.

[0004] Furthermore, current collaborative control logic largely relies on offline preset schemes, failing to achieve fine-grained matching of response characteristics for multiple types of resources, resulting in poor suppression of transient fluctuations in complex fault scenarios. Summary of the Invention

[0005] The purpose of this invention is to provide a data processing method for collaborative control of multiple types of adjustable resources, in order to solve the technical problems in the background art.

[0006] This invention provides a data processing method for collaborative control of multiple types of adjustable resources, including: S1. Real-time acquisition of operational status data of flexible DC converter stations, various types of energy storage power stations, and key nodes of AC power grids using synchronous phasor measurement units and remote terminal units, followed by noise reduction and normalization processing to form a multi-dimensional spatiotemporal feature sequence, in order to construct a multi-source heterogeneous feature database of AC / DC strongly coupled systems. S2. Using the local linear embedding algorithm and Laplace eigenmap technique, multi-source heterogeneous feature data in high-dimensional space are mapped to low-dimensional manifold space. Core feature vectors reflecting the dynamic coupling characteristics of flexible DC and multiple types of energy storage systems in transient processes are extracted. By calculating the dynamic interaction entropy between different resources, the nonlinear correlation strength inside the AC / DC strongly coupled system is quantitatively described, so as to establish a correlation analysis model based on dynamic manifold embedding. S3. Based on the extracted core feature vectors, a transient stability discrimination function is constructed. The contribution of the voltage support characteristics of flexible DC and the power compensation characteristics of the energy storage system to the power angle stability, voltage stability and frequency stability of the system is analyzed. The sensitivity of key operating parameters of the system to disturbances is calculated by the trajectory sensitivity analysis method. The real-time safety margin value of the system under the current operating conditions is obtained to quantitatively evaluate the transient stability impact mechanism and safety margin. S4. Taking the maximization of safe operation margin as the objective function, and comprehensively considering the current limit of the flexible DC converter, the internal circulation current constraint of the converter island, the charging and discharging depth limit of multiple types of energy storage, and the cycle life loss cost, a multi-constraint heuristic search algorithm is used to generate a set of coordinated control instructions, including the flexible DC power step value, energy storage output priority, and dynamic reactive power compensation ratio, so as to formulate a coordinated optimization control strategy for multiple types of adjustable resources. S5. The coordinated control instruction set is sent to the underlying controller for execution, while continuously monitoring the system frequency fluctuation and voltage drop rate. Based on predictive control theory, the effect after execution is evaluated in a rolling manner. When the actual operating trajectory deviates from the preset safety domain, the secondary adjustment logic is triggered to correct the control parameters within a preset period, so as to implement real-time closed-loop monitoring and dynamic correction of control instructions.

[0007] In some embodiments, during S1, the process of constructing the multi-source heterogeneous feature database of the AC / DC strongly coupled system involves collaborative detection by distributed fiber optic sensors and intelligent electronic devices. The distributed fiber optic sensors are deployed using Bragg grating technology inside the converter transformer windings, between the cells of the energy storage battery pack, and at the heat dissipation parts of the converter valve to extract physical parameters including temperature, strain, and partial discharge. The data acquisition process involves performing protocol conversion through an edge computing gateway to convert the fieldbus protocol into a high-speed internal transmission protocol, and using a synchronization time system to synchronize the acquisition devices in different regions, controlling the sampling time synchronization error within a preset synchronization error threshold range.

[0008] In some embodiments, in S1, the denoising process of the original data is based on an adaptive filtering algorithm of wavelet transform, selecting basis functions with orthogonality and compact support, setting the decomposition level to a preset level, removing high-frequency random noise components through threshold shrinkage, and retaining signal features that reflect transient changes; in the denoising process, an improved exponential weighted moving average logic is introduced, and by dynamically adjusting the weighting coefficients, the smoothing intensity is reduced when the signal slope is detected to exceed a preset change threshold, thus retaining the true transient frontier information; The normalization process employs a maximum-minimum scaling logic to uniformly map the value range of heterogeneous data to a preset numerical range, eliminating the influence of different physical dimensions on the model convergence speed and weight allocation. For parameters exhibiting nonlinear variation characteristics, such as the state of charge, a mapping function is used to perform nonlinear normalization, improving the sensitivity of parameter evolution in the extreme value region.

[0009] In some embodiments, during the execution of the local linear embedding algorithm in S2, a preset number of nearest neighbor points are selected, the low-dimensional embedding mapping is optimized by minimizing the reconstruction error function, and the local topological structure in the original high-dimensional data space is preserved by utilizing the linear combination relationship between the original high-dimensional vector and the reconstruction weight. The calculation of dynamic interaction entropy is based on the sliding window technique. By setting a preset time window width and a preset time step, the key coupling path that dominates transient instability in the system is identified by analyzing the conditional probability distribution between different resource time-series signals.

[0010] In some embodiments, in S3, when quantifying the impact mechanism of transient stability and the safety margin, the transient stability discrimination function integrates the transient energy function method and the extended equal area criterion to calculate the sum of transient kinetic energy and transient potential energy of the system at the moment of fault occurrence and clearing in real time. By comparing the total energy value of the current system with the critical energy value in real time, the energy margin coefficient is obtained. When the energy margin coefficient is greater than the preset safety threshold, the system is determined to be in a safe and stable state. The trajectory sensitivity analysis method uses the forward difference scheme to calculate the partial derivatives of the system state variables with respect to the initial values ​​and control parameters. The calculation time step is dynamically adjusted according to the system frequency fluctuation rate to cover the entire transient evolution cycle and identify a preset number of key control parameters that have the highest degree of influence on transient stability.

[0011] In some embodiments, in S4, when formulating a multi-type adjustable resource collaborative optimization control strategy, the improved particle swarm optimization algorithm is used to solve the objective function, a preset population size, a preset number of iterations and a preset learning factor are set, and a mutation operator is introduced to perform random perturbation on some particles when the population aggregation degree exceeds a preset aggregation threshold. The collaborative optimization model defines flexible DC, energy storage and AC power grid as the main players in the game, and seeks the equilibrium point that ensures the overall safe operation margin of the system and balances the losses of all parties. The various types of energy storage include at least electrochemical energy storage, supercapacitors, and flywheel energy storage. Different types of energy storage are dynamically prioritized in the control based on frequency response characteristics, available energy state, and response speed. Supercapacitors respond first during the first preset response time period in the early stage of a fault to suppress transient overvoltage. Flywheel energy storage provides frequency support during the second preset response time period. Electrochemical energy storage continues to regulate in the quasi-steady-state stage after the preset node.

[0012] In some embodiments, during S5, when implementing real-time closed-loop monitoring and dynamic correction of control commands, the collaborative control command set is transmitted through a dedicated optical fiber communication network, and the transmission delay and communication packet loss rate are kept within a preset communication quality range. Predictive control theory uses an augmented state-space model for online prediction, models the delay term of the control command as a state variable, and achieves advance compensation for the lag in response to multiple types of resources by minimizing the quadratic performance index between the predicted output and the reference trajectory. During instruction execution, the system state vector is evolved using state-space equations, and the smoothness constraint of the control action is incorporated into the performance index function to prevent frequent control adjustments from causing mechanical stress damage to power electronic devices.

[0013] In some embodiments, the secondary regulation logic includes at least emergency power boost, rapid reactive power cut-off, and dynamic adjustment of DC voltage setpoint; wherein, the emergency power boost adopts a compensation strategy based on sliding mode control, outputting a strong nonlinear gain when the voltage drop rate exceeds a preset drop threshold, forcing the energy storage system to reach the maximum discharge rate, and the magnitude of the correction is nonlinearly scaled according to the absolute value of the deviation.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention overcomes the limitations of traditional physical mechanism models in dealing with high-dimensional, nonlinear systems by using dynamic manifold embedding technology, and can capture the interactive effects between flexible DC and multiple types of energy storage; 2. This invention is based on a fusion evaluation method of trajectory sensitivity and energy function, which realizes real-time quantification of the system's safety margin, controls the evaluation error within a preset minimum error range, and provides stable criteria with reference value for power grid operators; 3. This invention significantly reduces the range of transient voltage fluctuations in the system by formulating a refined collaborative control strategy to match resources with different response characteristics at the millisecond level. 4. This invention uses a multi-objective optimization algorithm to ensure stability while balancing control costs and equipment lifespan, significantly extending the service life of the energy storage system. At the same time, it effectively avoids the risk of blockage of flexible DC under fault conditions. 5. Based on predictive control and fast correction logic, this invention can respond to environmental changes in real time during complex transient processes, significantly shortening the closed-loop adjustment cycle of control commands. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the data processing method for collaborative control of multiple types of adjustable resources according to the present invention; Figure 2 This is a schematic diagram illustrating the principle of extracting core feature vectors according to an embodiment of the present invention. Detailed Implementation

[0017] The following will be based on embodiments of the present invention. Figures 1-2 The technical solutions in the embodiments of the present invention will be clearly and completely described together. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Application Overview Flexible DC transmission technology and various types of energy storage systems are effective means to improve the grid's regulation capabilities. Their connection location, capacity allocation, and response characteristics have a profound impact on the system's transient stability. For the complex operating environment of AC / DC strongly coupled systems, this invention utilizes advanced data processing technology to analyze the dynamic response mechanism of flexible DC and energy storage resources under fault disturbances, and based on this, achieves multi-resource collaborative optimization control. This is a key step in improving the system's ability to withstand disturbances during transient processes. Therefore, the method of this invention is described in detail below.

[0019] Example 1 First, in S1, the synchronous phasor measurement unit and remote terminal unit are used to collect the operating status data of the flexible DC converter station, various types of energy storage power stations and key nodes of the AC power grid in real time, and perform noise reduction and normalization processing to form a multi-dimensional spatiotemporal feature sequence, so as to construct a multi-source heterogeneous feature database of AC-DC strongly coupled system.

[0020] Specifically, high-precision synchronous phasor measurement units (PMUs) and remote terminal units (RTUs) enable real-time and synchronous acquisition of operational status data from flexible DC converter stations, various types of energy storage power stations, and key nodes in the AC power grid. At the hardware deployment level for data acquisition, distributed fiber optic sensors and intelligent electronic devices (IEDs) work collaboratively. The fiber optic sensors, utilizing Bragg grating technology, are deployed on the windings of the converter transformer, the cell gaps of the energy storage battery pack, and key heat dissipation components of the converter valve to acquire physical parameters such as temperature, strain, and partial discharge. The sampling accuracy is set to at least 16 bits to ensure the comprehensiveness and accuracy of the data source. The data acquisition frequency is set to 12.8 kHz to capture high-frequency components during transient processes, and the sampling time synchronization error is controlled within 1 microsecond using a BeiDou or GPS timing system to ensure timescale consistency.

[0021] The types of data acquired include electrical physical quantities and state quantities, specifically including the voltage vectors (amplitude and phase angle) of each bus, the current vectors of the lines, the active power and reactive power output of the converter station, the system frequency deviation, the modulation ratio of the flexible DC converter, the total harmonic distortion rate on the AC side, and the state of charge (SOC), individual voltage consistency coefficient, and internal resistance state of the energy storage system.

[0022] After acquiring the raw data, denoising was performed using an adaptive filtering algorithm based on wavelet transform. Specifically, Daubechies wavelets, which possess orthogonality and compact support, were selected as the basis functions, and the decomposition level was set to 5 levels. High-frequency random noise components were removed using soft or hard thresholding methods, while retaining signal characteristics reflecting transient changes (such as wavefronts at the moment of a fault). The signal-to-noise ratio (SNR) of the denoised signal was improved by more than 25 dB.

[0023] The denoised data was then normalized. The normalization process used the max-min scaling method to map the value range of all heterogeneous data to the numerical range of [0, 1] or [-1, 1].

[0024] Next, in S2, the local linear embedding algorithm and Laplace eigenmap technique are used to map the multi-source heterogeneous feature data in the high-dimensional space to the low-dimensional manifold space. The core feature vectors that reflect the dynamic coupling characteristics of flexible DC and multiple types of energy storage systems in the transient process are extracted. By calculating the dynamic interaction entropy between different resources, the nonlinear correlation strength inside the AC-DC strongly coupled system is quantitatively described, so as to establish a correlation analysis model based on dynamic manifold embedding.

[0025] Specifically, the Locally Linear Embedding (LLE) algorithm and the Laplacian Eigenmap (LE) technique are used to map multi-source heterogeneous feature data in high-dimensional space to a low-dimensional manifold space. In the LLE algorithm, the number of nearest neighbors, k, is selected as a preset number between 12 and 20. The low-dimensional embedding mapping is optimized by minimizing the reconstruction error function, which is mathematically expressed as follows: in, x i The original high-dimensional vector; w ij To reconstruct the weights, this process preserves the local topology in the original high-dimensional data space, ensuring that the nonlinear response characteristics of flexible DC and energy storage resources are not distorted during dimensionality reduction, thereby extracting core feature vectors that reflect transient dynamic coupling characteristics.

[0026] Based on feature extraction, the strength of nonlinear correlations is quantified by calculating the dynamic interaction entropy between different resources. The dynamic interaction entropy is calculated using a sliding window technique, with a time window width of 20 milliseconds (one power frequency cycle) and a step size of 5 milliseconds. By analyzing the conditional probability distribution between the time-series signals of two different resources (such as flexible DC power command and energy storage discharge current), the key coupling paths that dominate transient instability in the system are identified. When the interaction entropy of a certain path exceeds a preset threshold, it is defined as a strongly coupled link and serves as an important reference for subsequent control command allocation.

[0027] Next, in S3, based on the extracted core feature vectors, a transient stability discrimination function is constructed. The contribution of the voltage support characteristics of flexible DC and the power compensation characteristics of the energy storage system to the system's power angle stability, voltage stability, and frequency stability is analyzed. The sensitivity of key system operating parameters to disturbances is calculated by the trajectory sensitivity analysis method, and the real-time safety margin value of the system under the current operating conditions is obtained to quantitatively evaluate the transient stability impact mechanism and safety margin.

[0028] Specifically, the system calculates in real time the sum of transient kinetic energy (related to rotor angular velocity deviation) and transient potential energy (related to system power angular displacement and bus voltage) at the moment of fault occurrence and clearing. The energy margin coefficient is calculated by comparing the current total energy value of the system with the critical energy value in real time. When the energy margin coefficient is greater than a preset safety threshold of 0.15, the system is determined to be in a safe and stable state.

[0029] Simultaneously, trajectory sensitivity analysis is used to calculate the sensitivity of key system operating parameters to disturbances. Trajectory sensitivity analysis employs a forward difference scheme to calculate the partial derivatives of system state variables (such as power angle, frequency, and voltage) with respect to initial values ​​and control parameters (such as flexible DC gain and energy storage response coefficient). The calculation time step is set to 1 millisecond, covering the entire transient evolution cycle (typically 2 to 5 seconds after a fault occurs). This method identifies the top 5 key control parameters with the greatest impact on transient stability, and based on these, the real-time safety margin of the system under the current operating conditions is obtained.

[0030] Then, in S4, with the objective function of maximizing the safety margin, the current limit of the flexible DC converter, the internal circulation current constraint of the converter island, the charging and discharging depth limit of multiple types of energy storage, and the cycle life loss cost are comprehensively considered. A multi-constraint heuristic search algorithm is used to generate a set of coordinated control instructions, including the flexible DC power step value, the energy storage output priority, and the dynamic reactive power compensation ratio, so as to formulate a coordinated optimization control strategy for multiple types of adjustable resources.

[0031] Specifically, the following factors are considered when setting constraints: the current limit of the flexible DC converter (usually set at 1.2 times the rated current), the internal circulating current constraint of the converter island, the charge and discharge depth limit of various types of energy storage, such as the SOC of electrochemical energy storage needing to be maintained between 20% and 90%, and the economical operation constraint based on cycle life loss cost.

[0032] An improved Particle Swarm Optimization (PSO) algorithm was used to generate a cooperative control instruction set. The PSO algorithm was set to a population size of 100, 200 iterations, and learning factors c1 and c2 both set to 2.0. A mutation operator was introduced to randomly perturb some particles when the population aggregation became too high, in order to escape local optima. The generated control instruction set included flexible DC power step values, output priorities for different types of energy storage, and dynamic reactive power compensation ratios.

[0033] To address the differences in energy storage resources, various types of energy storage include at least electrochemical energy storage, supercapacitors, and flywheel energy storage. Dynamic prioritization is employed in the control process: supercapacitors, due to their high power density, prioritize response within the initial 0-100 milliseconds of a fault to suppress transient overvoltages; flywheel energy storage provides frequency support within 100-500 milliseconds; and electrochemical energy storage then takes over regulation in the quasi-steady-state phase after 500 milliseconds to restore energy balance.

[0034] Finally, in S5, the coordinated control instruction set is sent to the underlying controller for execution, while continuous monitoring of system frequency fluctuations and voltage drop rates is maintained. Based on predictive control theory, the effect after execution is evaluated in a rolling manner. When the actual operating trajectory deviates from the preset safety domain, the secondary adjustment logic is triggered to correct the control parameters within a preset period, so as to implement real-time closed-loop monitoring and dynamic correction of control instructions.

[0035] Specifically, the generated collaborative control command set is sent to the underlying controller via a dedicated fiber optic communication network. The communication network adopts a dedicated 10G EPON architecture for power systems, with transmission latency controlled within 1 millisecond and a packet loss rate of less than 0.01%, ensuring the real-time nature of command issuance.

[0036] During instruction execution, the system continuously monitors the frequency fluctuation rate (df / dt) and voltage sag rate (dv / dt). The post-execution effect is evaluated using predictive control (MPC) theory. MPC employs a state-space model for online prediction. in, x(k) This is the system state vector; u(k) This is a coordinated control command. By minimizing the quadratic performance index between the predicted output and the reference trajectory, it enables advance compensation for response lags of various resource types, such as the chemical reaction delay in electrochemical energy storage.

[0037] If the actual operating trajectory deviates from the preset safety range (e.g., the voltage recovery trajectory falls below the preset lower limit), the secondary adjustment logic is immediately triggered. The rapid correction logic includes: emergency power boost, increasing the active power output of the flexible DC powertrain in milliseconds; rapid reactive power cut-off to prevent voltage overshoot caused by overcompensation; and dynamic adjustment of the DC voltage setpoint. The magnitude of the correction is non-linearly scaled based on the absolute value of the deviation.

[0038] Example 2 Based on Example 1, this embodiment further refines the construction details of the multi-source heterogeneous feature database in AC / DC strongly coupled systems and its data processing flow under extreme operating conditions.

[0039] In S1, to facilitate the construction of a multi-source heterogeneous feature database, intelligent gateways with edge computing capabilities are deployed at the converter station and energy storage power station. These gateways are responsible not only for protocol conversion (such as converting the IEC 61850 protocol to a high-speed internal protocol) but also for local preprocessing of the acquired high-frequency sampled values ​​(SV). For the acquired voltage and current vectors, the gateway internally performs Fourier transforms to extract the fundamental component and 2nd to 50th harmonic characteristics. Simultaneously, feedback from distributed fiber optic sensors is used to calculate the equivalent thermal parameters inside the converter transformer in real time.

[0040] In the denoising stage, to address impulse noise in the transient process, the system introduces an improved exponentially weighted moving average (EWMA) algorithm based on wavelet transform. This algorithm can dynamically adjust the weighting coefficients, and when the signal slope exceeds a preset threshold (determined as a transient abrupt change rather than noise), it reduces the smoothing intensity to preserve the true transient frontier. During normalization, for parameters with nonlinear changes such as SOC, the Sigmoid function is used for mapping, giving it higher sensitivity in the extreme value region.

[0041] In S2, this embodiment refines the calculation process of dynamic cross-entropy, treating the control quantities of the flexible DC system (such as the inner loop current command) and the output quantities of the energy storage system as two random variable sequences. By constructing a symbolic dynamic sequence, the mutual information between the two is calculated. The application of the sliding window technique is not limited to the time dimension but also extended to the frequency dimension. Through cross-entropy analysis of different frequency bands, the decoupling characteristics of low-frequency oscillations and transient instability are identified.

[0042] In S3, the transient stability discrimination function further integrates the real-time calculation of the system short-circuit capacity (SCC). By analyzing the difference vector between voltage and current before and after the disturbance, the equivalent short-circuit impedance of the AC bus is estimated online. When the SCC is lower than a preset critical value (such as the strong coupling discrimination threshold of the node), the system automatically raises the warning level of the safety margin and increases the calculation frequency of trajectory sensitivity analysis. The calculation step size of the forward differential scheme is dynamically adjusted according to the system frequency fluctuation rate: when the frequency fluctuation is severe, the step size is reduced from 1 millisecond to 0.1 milliseconds to capture extremely fast electromagnetic transient processes.

[0043] In S4, the multi-constraint heuristic search algorithm sets differentiated fitness functions for different types of energy storage resources. For electrochemical energy storage, the fitness function incorporates the lithium plating risk model and thermal runaway early warning constraints; for supercapacitors, it mainly focuses on the rapid fluctuation range of the terminal voltage. In the improved particle swarm optimization algorithm, the inertia weight adopts a nonlinear decreasing strategy, maintaining a large detection range in the early stage of the search and achieving accurate convergence in the later stage of the search.

[0044] The flexible DC power step value in the coordinated control command set is refined into multiple stages. For example, within 10 milliseconds after a fault is cleared, the first-level emergency active power support command is issued; at 50 milliseconds, the reactive power compensation ratio is adjusted according to the voltage recovery. The output priority of various types of energy storage is dynamically allocated in real time based on the available energy and response speed of each energy storage unit.

[0045] In S5, predictive control theory employs a more complex augmented state-space model, directly modeling the delay term of control commands as state variables. This means the system can anticipate command lags of 20 to 50 milliseconds due to communication network congestion and compensate for them in advance in the current control output. In the fast-correction logic, emergency power boosting uses a compensation strategy based on sliding mode control (SMC). When the voltage drop rate exceeds 0.5 pu / s, the sliding mode controller outputs a strongly nonlinear gain, forcing the energy storage system to instantly reach its maximum discharge rate.

[0046] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0047] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A data processing method of multi-type adjustable resource collaborative control, characterized in that, include: S1. Real-time acquisition of operational status data of flexible DC converter stations, various types of energy storage power stations, and key nodes of AC power grids using synchronous phasor measurement units and remote terminal units, followed by noise reduction and normalization processing to form a multi-dimensional spatiotemporal feature sequence, in order to construct a multi-source heterogeneous feature database of AC / DC strongly coupled systems. S2. Using the local linear embedding algorithm and Laplace eigenmap technique, multi-source heterogeneous feature data in high-dimensional space are mapped to low-dimensional manifold space. Core feature vectors reflecting the dynamic coupling characteristics of flexible DC and multiple types of energy storage systems in transient processes are extracted. By calculating the dynamic interaction entropy between different resources, the nonlinear correlation strength inside the AC / DC strongly coupled system is quantitatively described, so as to establish a correlation analysis model based on dynamic manifold embedding. S3. Based on the extracted core feature vectors, a transient stability discrimination function is constructed. The contribution of the voltage support characteristics of flexible DC and the power compensation characteristics of the energy storage system to the power angle stability, voltage stability and frequency stability of the system is analyzed. The sensitivity of key operating parameters of the system to disturbances is calculated by the trajectory sensitivity analysis method. The real-time safety margin value of the system under the current operating conditions is obtained to quantitatively evaluate the transient stability impact mechanism and safety margin. S4. Taking the maximization of safe operation margin as the objective function, and comprehensively considering the current limit of the flexible DC converter, the internal circulation current constraint of the converter island, the charging and discharging depth limit of multiple types of energy storage, and the cycle life loss cost, a multi-constraint heuristic search algorithm is used to generate a set of coordinated control instructions, including the flexible DC power step value, energy storage output priority, and dynamic reactive power compensation ratio, so as to formulate a coordinated optimization control strategy for multiple types of adjustable resources. S5. The coordinated control instruction set is sent to the underlying controller for execution, while continuously monitoring the system frequency fluctuation and voltage drop rate. Based on predictive control theory, the effect after execution is evaluated in a rolling manner. When the actual operating trajectory deviates from the preset safety domain, the secondary adjustment logic is triggered to correct the control parameters within a preset period, so as to implement real-time closed-loop monitoring and dynamic correction of control instructions.

2. The method of claim 1, wherein, In S1, during the construction of the multi-source heterogeneous feature database of the AC / DC strongly coupled system, distributed optical fiber sensors and intelligent electronic devices are used for collaborative detection. The distributed optical fiber sensors are deployed inside the converter transformer windings, between the cells of the energy storage battery pack, and in the heat dissipation parts of the converter valve using Bragg grating technology to extract physical parameters including temperature, strain, and partial discharge. The data acquisition process performs protocol conversion through an edge computing gateway, converting the fieldbus protocol into a high-speed internal transmission protocol, and uses a synchronization time system to synchronize the acquisition devices in different regions, controlling the sampling time synchronization error within a preset synchronization error threshold range.

3. The method according to claim 1, characterized in that, In S1, the denoising process of the original data is based on an adaptive filtering algorithm using wavelet transform. It selects basis functions with orthogonality and compact support, sets the decomposition level to a preset number, and removes high-frequency random noise components using a threshold shrinkage method while retaining signal characteristics reflecting transient changes. An improved exponentially weighted moving average logic is introduced during the denoising process. By dynamically adjusting the weighting coefficients, the smoothing intensity is reduced when the detected signal slope exceeds a preset change threshold, thus preserving the true transient frontier information. The normalization process uses the maximum and minimum scaling logic to uniformly map the value range of heterogeneous data to a preset numerical range, eliminating the influence of different physical dimensions on the model convergence speed and weight allocation. For parameters such as the state of charge that exhibit nonlinear changing characteristics, a mapping function is used to perform nonlinear normalization, thereby improving the sensitivity of parameter evolution in the extreme region.

4. The method according to claim 1, characterized in that, In S2, during the execution of the local linear embedding algorithm, a preset number of nearest neighbor points are selected, and the low-dimensional embedding mapping is optimized by minimizing the reconstruction error function. The local topological structure in the original high-dimensional data space is preserved by utilizing the linear combination relationship between the original high-dimensional vector and the reconstruction weight. The calculation of dynamic interaction entropy is based on the sliding window technique. By setting a preset time window width and a preset time step, the key coupling path that dominates transient instability in the system is identified by analyzing the conditional probability distribution between different resource time-series signals.

5. The method according to claim 1, characterized in that, In S3, when quantifying the impact mechanism of transient stability and the safety margin, the transient stability discrimination function integrates the transient energy function method and the extended equal area criterion to calculate the sum of transient kinetic energy and transient potential energy of the system at the moment of fault occurrence and clearing in real time. By comparing the total energy value of the current system with the critical energy value in real time, the energy margin coefficient is obtained. When the energy margin coefficient is greater than the preset safety threshold, the system is determined to be in a safe and stable state. The trajectory sensitivity analysis method uses the forward difference scheme to calculate the partial derivatives of the system state variables with respect to the initial values ​​and control parameters. The calculation time step is dynamically adjusted according to the system frequency fluctuation rate to cover the entire transient evolution cycle and identify a preset number of key control parameters that have the highest degree of influence on transient stability.

6. The method according to claim 1, characterized in that, In S4, when formulating a multi-type adjustable resource collaborative optimization control strategy, the improved particle swarm optimization algorithm is used to solve the objective function. The preset population size, preset number of iterations and preset learning factor are set, and a mutation operator is introduced. When the population aggregation degree exceeds the preset aggregation threshold, random perturbation is performed on some particles. The collaborative optimization model defines flexible DC, energy storage and AC power grid as the main players in the game, and seeks the equilibrium point that ensures the overall safe operation margin of the system and balances the losses of all parties. The various types of energy storage include at least electrochemical energy storage, supercapacitors, and flywheel energy storage. Different types of energy storage are dynamically prioritized in the control based on frequency response characteristics, available energy state, and response speed. Supercapacitors respond first during the first preset response time period in the early stage of a fault to suppress transient overvoltage. Flywheel energy storage provides frequency support during the second preset response time period. Electrochemical energy storage continues to regulate in the quasi-steady-state stage after the preset node.

7. The method according to claim 1, characterized in that, In S5, when implementing real-time closed-loop monitoring and dynamic correction of control commands, the collaborative control command set is transmitted through a dedicated optical fiber communication network, and the control transmission delay and communication packet loss rate are within the preset communication quality range. Predictive control theory uses an augmented state-space model for online prediction, models the delay term of the control command as a state variable, and achieves advance compensation for the lag in response to multiple types of resources by minimizing the quadratic performance index between the predicted output and the reference trajectory. During instruction execution, the system state vector is evolved using state-space equations, and the smoothness constraint of the control action is incorporated into the performance index function to prevent frequent control adjustments from causing mechanical stress damage to power electronic devices.

8. The method according to claim 1, characterized in that, The secondary adjustment logic includes at least emergency power boost, rapid reactive power cut-off, and dynamic adjustment of DC voltage setpoint; wherein, the emergency power boost adopts a compensation strategy based on sliding mode control, outputting a strong nonlinear gain when the voltage drop rate exceeds the preset drop threshold, forcing the energy storage system to reach the maximum discharge rate, and the magnitude of the correction is nonlinearly scaled according to the absolute value of the deviation.

Images