A method for optimizing the allocation of energy storage in distribution networks based on renewable energy consumption

Abstract

This invention provides a method for optimizing energy storage configuration in distribution networks based on renewable energy consumption, belonging to the field of energy storage optimization configuration. The method involves: a synchronization feature module for time-domain synchronization processing to generate feature vectors; a frequency mapping modeling module to construct an equivalent feature matrix including inter-axis cross terms; an admittance reconstruction module to inject this into the static topology to generate a real-state dynamic admittance model and solve for the impedance matrix; a sensitivity evaluation module to map and generate real-state voltage sensitivity and determine the impedance offset index; a configuration module to perform nonlinear hedging to generate configuration data; and a calibration module to perform recursive calibration. The system also generates source-grid attribution weights based on normalized cross-correlation coefficients, enabling accurate determination and differentiated correction of causes on the equipment side and grid side, and introduces a self-evolving evolutionary flow for closed-loop optimization. This invention effectively reduces asynchronous interference from heterogeneous data, improves the accuracy of risk attribution and control, and ensures the stability of the distribution network.

Description

Technical Field

[0001] This invention belongs to the field of energy storage optimization configuration, and relates to a method for optimizing the configuration of energy storage in distribution networks based on the consumption of new energy sources. Background Technology

[0002] With the increasing penetration of new energy sources in power distribution networks, the randomness and volatility of distributed power sources such as photovoltaics and wind power pose significant challenges to the voltage stability of the power grid. Energy storage systems, as a key means to improve the flexibility and absorption capacity of power distribution networks, typically rely on precise measurement of the voltage sensitivity of network nodes for optimized configuration and real-time control. By performing impedance modeling and sensitivity analysis on each access node, the energy storage system can be guided to perform power offsetting, thereby suppressing voltage exceedances and smoothing power fluctuations. This is a core technical path to ensure the safe and stable operation of power distribution networks with a high proportion of new energy access.

[0003] However, existing energy storage configuration technologies are mainly based on static physical topology to establish admittance models, whose core assumption is that the electrical characteristics of the power grid are rigidly determined only by physical parameters such as conductor impedance. This model leads to severe time-domain asynchrony between operational data and equipment control parameters. At the same time, it ignores the influence between active and reactive power dynamics, fails to identify risk factors, and is prone to system oscillations or insufficient regulation accuracy. Summary of the Invention

[0004] 1. The technical problem to be solved:

[0005] How to effectively reduce the interference of asynchronous sampling time scale and communication delay on the accuracy of virtual-real coupling modeling, improve the accuracy of measuring electrical distance deviation and voltage sensitivity evolution caused by the controlled characteristics of grid-connected equipment, enhance the automatic attribution and differentiated regulation capability of coupling risks between equipment control layer and grid topology layer, thereby improving the advantages of voltage stability and energy storage configuration results of distribution network in the process of new energy consumption.

[0006] 2. Technical Solution:

[0007] To address the above problems, this invention provides a method for optimizing the configuration of energy storage in distribution networks based on renewable energy consumption, comprising the following steps:

[0008] Step 1: Running status data stream and internal controlled feature parameter set.

[0009] Step 2: Synchronization Feature Construction: Extract signal correlation features from the acquired operating status data stream and, based on the signal correlation features, perform time-domain synchronization processing on the operating status data stream and the acquired internal controlled feature parameter set to generate a target composite feature vector.

[0010] Step 3: Frequency mapping modeling: Based on the target composite feature vector and the preset mapping parameters, frequency domain characteristic mapping processing is performed to construct the equivalent electrical feature matrix of grid-connected new energy equipment in the coordinate transformation dimension. The equivalent electrical feature matrix includes inter-axis cross feature terms that characterize the coupling interaction between active and reactive power dynamics.

[0011] Step 4: Admittance Reconstruction: The equivalent electrical characteristic matrix corresponding to each grid-connected new energy equipment is used as a controlled feature increment and injected into the obtained static physical topology parameters to reconstruct the real dynamic admittance model of the target distribution network, and the real node impedance matrix characterizing the electrical distance offset is obtained based on the real dynamic admittance model.

[0012] Step 5: Sensitivity Assessment: By extracting the mutual impedance tensor from the real-state node impedance matrix, the real-state voltage sensitivity characteristics of the absorption node relative to the energy storage node are generated and the impedance offset index of the real-state voltage sensitivity characteristics relative to the preset initial sensitivity benchmark is determined.

[0013] Step 6: Configuration: Obtain the reference operating parameters of the energy storage system based on the static physical topology parameters, and perform nonlinear hedging correction on the reference operating parameters according to the impedance offset index to generate multidimensional coupling configuration data of the energy storage system, and adjust the power output state of the energy storage system.

[0014] Step 7: Calibration, used to recursively calibrate the mapping parameters of the equivalent electrical characteristic matrix based on the measured voltage feedback deviation after running the multidimensional coupling configuration data. The recursive calibration of the mapping parameters proceeds to step 3, and so on.

[0015] In step 2, the specific process of signal association feature extraction is as follows:

[0016] Step 201: Extract non-stationary oscillation features from the voltage waveform using an approximate entropy algorithm, and determine the first signal entropy value characterizing the intensity of controlled impedance fluctuations. The approximate entropy algorithm refers to a nonlinear analysis method that quantifies the complexity and irregularity of time series. The non-stationary oscillation features refer to the dynamic oscillation modes in the voltage waveform whose frequency or amplitude changes with time. The first signal entropy value is a scalar index that quantifies the complexity of the voltage waveform.

[0017] Step 202: Based on the first signal entropy value, dynamically trigger a resampling request for the centralized control gain weight of the internal controlled feature parameters, and generate a first feedback closed loop flow that changes the parameter perception depth in real time according to the running state. The resampling request refers to the instruction signal that triggers high-frequency acquisition of control parameters according to the change of signal entropy value. The first feedback closed loop flow refers to the iterative loop mechanism of signal entropy value monitoring, resampling triggering, parameter updating and entropy value re-evaluation.

[0018] The specific process of the time-domain synchronization process is as follows:

[0019] Step 211: Obtain the first sampling time stamp of the grid-connected node and the second communication time stamp of the internal controlled feature parameter set; determine the dynamic time deviation between the first sampling time stamp and the second communication time stamp. The first sampling time stamp refers to the timestamp corresponding to each sampling point in the running status data stream, and the second communication time stamp refers to the timestamp carried when the control parameters are transmitted to the data processing unit through the communication protocol. The dynamic time deviation refers to the time difference between the two time stamps and its timing change characteristics.

[0020] Step 212: Align the running status data stream using a preset interpolation prediction algorithm, and reversely correct the cache trigger frequency of the running status data stream collected in the synchronization feature module according to the rate of change of the dynamic time deviation. The interpolation prediction algorithm refers to a mathematical method for predicting missing time values ​​based on existing data points.

[0021] In step 3, the construction process of the equivalent electrical characteristic matrix is ​​as follows:

[0022] Based on the target composite eigenvector, a two-way coupled admittance model considering the dynamic characteristics of the phase-locked loop is established in a synchronous rotating coordinate system.

[0023] The synchronous rotating coordinate system refers to the dq coordinate frame that rotates synchronously with the base frequency of the power grid, and the bidirectional coupled admittance model refers to the admittance matrix expression form that simultaneously includes the coupling paths from the d-axis to the q-axis and from the q-axis to the d-axis.

[0024] The equivalent electrical characteristic matrix is ​​quantified using the following formula:

[0025] ,

[0026] in Let denote the equivalent electrical characteristic matrix, and s denote the Laplace operator. Indicates the filter inductance parameters. and denoted by , respectively, the proportional gain and integral gain of the current loop control, j represents the imaginary unit used to describe the phase and complex characteristics of the AC system, ω0 represents the fundamental frequency, and Θ(P) represents the nonlinear correction coefficient determined based on the real-time output active power P.

[0027] In step 4, the specific process of generating the real-state nodal impedance matrix is ​​as follows:

[0028] The equivalent electrical characteristic matrix is ​​used as a controlled feature increment and input to the corresponding node position in the benchmark admittance model to construct a dual topology space model. The dual topology space model is then subjected to manifold inversion operation by a sparse matrix solver to generate a real node impedance matrix.

[0029] The baseline admittance model refers to the traditional admittance matrix established based on static physical topology parameters. The dual topology space model refers to the comprehensive admittance matrix after integrating static topology and dynamic controlled features. The manifold inversion operation refers to the numerical calculation method for solving the inverse of the admittance matrix while maintaining matrix sparsity.

[0030] In step 5, the process of generating real-state voltage sensitivity characteristics includes: calculating the dynamic projection magnitude of the mutual impedance tensor in the real-state node impedance matrix in the direction of the real-time voltage vector, and determining the dynamic projection magnitude as the real-state voltage sensitivity characteristics.

[0031] The dynamic projection modulus refers to the projection length of the mutual impedance tensor vector in the direction of the current voltage vector. The real-state voltage sensitivity characteristic, through the real-time voltage vector direction, enables the sensitivity assessment to reflect the dynamic characteristics of the current operating point of the power grid.

[0032] After determining the impedance offset index, the sensitivity evaluation module further includes:

[0033] Extract the real-time temporal gradient of the impedance offset index within a preset time window; synchronously acquire the amplitude evolution characteristics of the inter-axis crossover feature; calculate the normalized cross-correlation coefficient between the real-time temporal gradient and the amplitude evolution characteristics, and generate source-grid attribution weight features to characterize the coupling degree between the equipment control layer and the power grid topology layer.

[0034] The real-time time-domain gradient refers to the derivative of the impedance offset exponent with time; the amplitude evolution characteristic refers to the trend of the value of the inter-axis crossover characteristic term with time; and the normalized cross-correlation coefficient refers to the normalized measure of the correlation between two time series at different time offsets.

[0035] The process of determining the dynamic attribution state based on the source-network attribution weight characteristics includes:

[0036] The source network attribution weight features are compared with a preset judgment threshold, which is the critical value that distinguishes between equipment-side causes and network structure-side causes.

[0037] If the source-network attribution weight feature is greater than the judgment threshold, then the voltage risk of the current absorption node is marked as the equipment-side induced state. The equipment-side induced state refers to the operating state in which the voltage risk is mainly caused by improper control parameters of grid-connected equipment or control coupling instability.

[0038] If the source network attribution weight feature is not greater than the judgment threshold, then the voltage risk of the current absorption node is marked as a grid-side induced state. The grid-side induced state refers to the operating state in which the voltage risk is mainly caused by excessive line impedance or weak grid structure.

[0039] In step 6, after marking the trigger state and before generating the multidimensional coupling configuration data, the following differential correction logic is performed:

[0040] If the current state is marked as a device-side induced state, the proportional gain used to achieve the nonlinear hedging correction is dynamically adjusted to enhance the energy storage system's ability to suppress internal coupling instability.

[0041] If the current state is marked as a grid-side induced state, the proportional gain of the nonlinear hedging correction remains unchanged, and the upper limit of the power change rate in the multidimensional coupling configuration data is forcibly limited to compensate for insufficient electrical damping on the grid side.

[0042] In step 7, the closed-loop recursive calibration also performs self-evolutionary flow logic:

[0043] The consistency of the mapping parameter correction amount generated by recursive calibration with the source network attribution weight feature is verified to determine the deviation between the two in the feedback adjustment direction. The mapping parameter correction amount refers to the equivalent electrical characteristic matrix parameter adjustment value calculated by the calibration module based on the voltage feedback deviation. The feedback adjustment direction deviation refers to the angle or difference measure between the mapping parameter correction direction and the direction indicated by the attribution weight feature.

[0044] Based on the deviation, the weighting weights of the source network attribution weight features related to the temporal change gradient are adjusted in reverse, and the calibrated attribution weight features are synchronously fed back to admittance reconstruction and sensitivity assessment. The weighting weights refer to the proportion of the temporal change gradient in the attribution weight feature calculation formula.

[0045] 3. Beneficial effects:

[0046] In this invention, by injecting the equivalent electrical feature matrix containing inter-axis crossover feature terms as a controlled feature increment into the static physical topology parameters, a real dynamic admittance model and a real node impedance matrix are reconstructed and generated. This effectively solves the problem of the disconnect between the virtual and real models caused by ignoring the controlled characteristics of power electronic equipment in the prior art, and improves the real-time performance and accuracy of the evaluation of the real voltage sensitivity characteristics of the admittance node.

[0047] In this invention, by calculating the normalized cross-correlation coefficient between the real-time time-domain change gradient of the impedance offset index within a preset time window and the amplitude evolution characteristics of the inter-axis cross-feature term, a source-grid attribution weight feature is generated to characterize the coupling degree between the equipment control layer and the grid topology layer. Based on this, a differentiated correction logic is performed, which realizes the accurate identification and classification of voltage risk causes. This overcomes the technical defect of traditional control methods that cannot distinguish between internal coupling instability on the equipment side and insufficient electrical damping on the grid side, and enhances the operational robustness of the distribution network under high-proportion renewable energy consumption conditions. Attached Figure Description

[0048] Figure 1 This is a flowchart of a distribution network energy storage optimization method based on the consumption of new energy sources. Detailed Implementation

[0049] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0050] A distribution network energy storage optimization method based on renewable energy consumption, such as Figure 1 As shown, it includes the following steps:

[0051] Step 1: Running status data stream and internal controlled feature parameter set.

[0052] Step 2: Synchronization Feature Construction: Extract signal correlation features of grid-connected nodes from the acquired operating status data stream. The signal correlation features perform time-domain synchronization processing on the operating status data stream and the acquired internal controlled feature parameter set to generate a target composite feature vector.

[0053] In one embodiment, the signal association feature extraction process is the core mechanism for achieving adaptive adjustment of parameter-aware depth, and the specific process includes:

[0054] Step 201: Extract the non-stationary oscillation characteristics in the voltage waveform using the approximate entropy algorithm, and determine the first signal entropy value characterizing the intensity of the controlled impedance fluctuation;

[0055] Step 202: Based on the first signal entropy value, dynamically trigger a resampling request for the centralized control gain weight of the internal controlled feature parameters, and generate a first feedback closed loop flow that changes the parameter perception depth in real time according to the running state.

[0056] The approximate entropy algorithm refers to a nonlinear analysis method that quantifies the complexity and irregularity of time series. Specifically, it can be implemented by calculating the log-likelihood of the pattern repetition probability in the voltage waveform sequence. Its function is to identify the non-stationary oscillation component caused by equipment control in voltage fluctuations.

[0057] The non-stationary oscillation characteristic refers to the dynamic oscillation mode in the voltage waveform where the frequency or amplitude changes with time. Specifically, it can be manifested as impedance fluctuations caused by changes in controller parameters, and its function is to reflect the dynamic change characteristics of the controlled state of the equipment.

[0058] The first signal entropy value is a scalar index that quantifies the complexity of the voltage waveform. Specifically, it can be represented by a numerical value output by an approximate entropy algorithm. Its function is to determine the degree of influence of the equipment control parameters on the power grid state at the current moment. When the entropy value increases, it indicates that the irregularity of the voltage waveform is enhanced, which usually means that the equipment control state is undergoing drastic changes; when the entropy value is small, it indicates that the voltage waveform is relatively regular and the equipment control state is relatively stable.

[0059] The resampling request refers to the instruction signal that triggers high-frequency acquisition of control parameters based on changes in signal entropy. Specifically, it can be achieved by setting an entropy threshold and increasing the sampling frequency when the threshold is exceeded. Its function is to improve the accuracy of parameter perception when drastic fluctuations in the control state of the device are detected.

[0060] The first feedback closed-loop flow refers to an iterative loop mechanism of signal entropy monitoring, resampling triggering, parameter updating, and entropy re-evaluation. Specifically, it can achieve adaptive matching between parameter sampling frequency and the dynamic characteristics of the power grid through a closed-loop control algorithm. Its function is to enable the system to maintain dynamic tracking capability against changes in the equipment control layer. The core advantage of this mechanism is that it can automatically increase the sampling density when the equipment control parameters change drastically to capture the transient characteristics of the control state, while reducing the sampling frequency when the equipment is running smoothly to save communication resources and computational overhead.

[0061] Through the above technical solution, this invention achieves sensitive capture of changes in the controlled state of grid-connected equipment. Traditional methods use a fixed sampling frequency, which may miss critical transient information when equipment control parameters change rapidly, and waste sampling resources when the equipment is operating smoothly. This solution identifies non-stationary oscillation characteristics in voltage waveforms through an approximate entropy algorithm, enabling timely detection of impedance fluctuations caused by changes in equipment control parameters, and adaptively adjusts the parameter sensing depth through a dynamic resampling mechanism. This adaptive sampling strategy based on signal entropy not only improves the accuracy of the target composite feature vector in representing the dynamic characteristics of the equipment, but also lays a data foundation for the accurate construction of the subsequent equivalent electrical feature matrix, ensuring that the frequency domain mapping process can capture the complete dynamic evolution trajectory of the controlled state of the equipment.

[0062] In one embodiment, the time-domain synchronization process is a key technical step in solving the problem of time-domain asynchrony between runtime data and device control parameters. The specific process includes:

[0063] Step 211: Obtain the first sampling timescale of the grid-connected node and the second communication timescale of the internal controlled feature parameter set; determine the dynamic time deviation between the first sampling timescale and the second communication timescale;

[0064] Step 212: Align the running status data stream using a preset interpolation prediction algorithm, and reverse-correct the cache trigger frequency of the running status data stream collected in the synchronization feature module according to the rate of change of the dynamic time deviation.

[0065] The first sampling time stamp refers to the timestamp mark corresponding to each sampling point in the operating status data stream. Specifically, it can be obtained through the built-in GPS clock synchronization mechanism of the PMU device. Its function is to identify the time reference for power grid status observation.

[0066] The second communication timestamp refers to the timestamp carried when the control parameters are transmitted to the data processing unit through the communication protocol. Specifically, it can be reflected by the difference between the device controller clock and the data acquisition system clock. Its function is to reflect the actual time when the control parameters are acquired.

[0067] The dynamic time deviation refers to the time difference between two time scales and its temporal change characteristics. Specifically, it can be obtained by subtracting the time scales and calculating their temporal change rate. Its function is to quantify the degree of data asynchrony caused by communication delay and sampling out-of-sync.

[0068] The interpolation prediction algorithm refers to a mathematical method that predicts missing time values ​​based on existing data points. Specifically, it can be implemented using cubic spline interpolation or Kalman filtering. Its function is to align the running data stream with the control parameter set to a unified time reference. The buffer trigger frequency refers to the periodic frequency at which the data acquisition module reads data from the buffer for processing. This can be achieved by adjusting the clock division ratio of the sampling trigger. Its function is to dynamically adapt to changes in communication delay to reduce time deviations in future moments. The reverse correction mechanism monitors the changing trend of dynamic time deviations and adjusts the data acquisition frequency forward, thereby suppressing the cumulative effect of time-domain asynchrony at the source. This mechanism is equivalent to establishing a negative feedback control loop, enabling the data acquisition system to adaptively track dynamic changes in communication delay.

[0069] Through the above technical solution, this invention achieves precise time-domain synchronization of operational data and equipment control parameters. Compared to traditional methods that rely solely on static timestamp alignment, this solution dynamically monitors the rate of change of time deviation and inversely corrects the sampling frequency, proactively adapting to fluctuations in communication latency and fundamentally eliminating data asynchrony. This time-domain synchronization mechanism provides spatiotemporally consistent foundational data for subsequent frequency-domain modeling and admittance reconstruction, solving the problem of dynamic characteristic distortion caused by data time-domain mismatch in traditional static models, and ensuring that the equivalent electrical characteristic matrix accurately reflects the true controlled state of the equipment at a specific moment.

[0070] Step 3: Frequency mapping modeling: Based on the target composite feature vector and the preset mapping parameters, frequency domain characteristic mapping processing is performed to construct the equivalent electrical feature matrix of grid-connected new energy equipment in the coordinate transformation dimension. The equivalent electrical feature matrix includes inter-axis cross feature terms that characterize the coupling interaction between active and reactive power dynamics.

[0071] In one embodiment, the construction of the equivalent electrical characteristic matrix is ​​the core step in mapping equipment control parameters to impedance increments that can be injected into the power grid model. The specific process is as follows:

[0072] Based on the target composite feature vector, a two-way coupled admittance model considering the dynamic characteristics of the phase-locked loop is established in a synchronous rotating coordinate system.

[0073] Specifically, the inter-axis crossover feature term in the equivalent electrical feature matrix characterizes the cross-axial interaction between active and reactive power, and the inter-axis crossover feature term is used to perform feedforward hedging correction on the multi-dimensional coupled configuration data generated by the configuration module.

[0074] The synchronous rotating coordinate system refers to the dq coordinate frame that rotates synchronously with the grid's base frequency. Specifically, the time-varying sinusoidal quantities in the three-phase abc coordinate system can be converted into DC quantities through the Park transformation. Its function is to simplify dynamic analysis and achieve natural separation of active and reactive components. The phase-locked loop (PLL) dynamic characteristics refer to the dynamic response characteristics exhibited by the PLL during grid phase tracking. Specifically, it can be characterized by parameters such as proportional gain and integral gain in the PLL transfer function. Its function is to introduce the cross-coupling effect of voltage phase fluctuations on the d-axis and q-axis quantities.

[0075] The bidirectional coupled admittance model refers to the admittance matrix expression that simultaneously includes the coupling paths from the d-axis to the q-axis and from the q-axis to the d-axis. Specifically, it can be obtained by establishing a state-space model through the small-signal linearization method and converting it into a frequency domain transfer function. Its role is to fully describe the bidirectional interaction relationship of active and reactive power dynamics.

[0076] This bidirectional coupling mechanism stems from the dynamic deviation that the phase-locked loop generates when tracking the grid phase. This deviation is coupled to the d-axis and q-axis components through the angle error term of the Park transform, thereby causing changes in active power to affect reactive power response, and vice versa.

[0077] Feedforward offsetting correction refers to a feedforward control strategy that compensates for the coupling effect between active and reactive power in advance when the energy storage control command is generated. Specifically, it can be achieved by applying the inter-axis crossover term as a compensation gain to the control loop. Its function is to counteract the undesirable coupling between d-axis and q-axis quantities caused by phase-locked loop dynamics, thereby improving the offsetting accuracy of energy storage. The physical essence of this mechanism is that when the energy storage system injects active power, the presence of the inter-axis crossover term will inevitably cause fluctuations in reactive power. Through feedforward compensation, the reactive component contributed by the inter-axis crossover term is subtracted from the active power command in advance, thereby achieving decoupled control of active and reactive power.

[0078] In one embodiment, the equivalent electrical characteristic matrix is ​​quantified using the following formula:

[0079] ;

[0080] in, Let denote the equivalent electrical characteristic matrix, and s denote the Laplace operator. Indicates the filter inductance parameters. and denoted by , respectively, the proportional gain and integral gain of the current loop control, j represents the imaginary unit used to describe the phase and complex characteristics of the AC system, ω0 represents the fundamental frequency, and Θ(P) represents the nonlinear correction coefficient determined based on the real-time output active power P.

[0081] The physical meaning of the above formula is that the 's' term in the numerator represents the differential characteristic, reflecting the variation of inductive impedance with frequency, and its frequency domain response is: The second-order transfer function in the denominator characterizes the dynamic response of the current loop controller, where, The term corresponds to the inductive inertial element. The damping effect introduced by the corresponding proportional control, The term corresponds to the stiffness provided by integral control. The evaluation at the fundamental frequency means extracting the admittance characteristics at that point, so that the equivalent matrix reflects the impedance behavior of the equipment under power frequency operating conditions. At this point, the numerator is... The denominator is Dividing the two yields the complex admittance value, with its real part corresponding to the conductance component and its imaginary part corresponding to the susceptance component. The nonlinear correction coefficient Θ(P) introduces the modulation effect of output power on the admittance characteristics, reflecting the nonlinear effect of the device's dynamic characteristics changing with the operating point under large-signal conditions. Specifically, it can be designed as follows: The form is , where α is a nonlinear coefficient. For rated power, this design ensures that the admittance characteristics of the device are close to the small-signal linearization result when operating under light load (smaller P), and close to the rated power when operating under heavy load (smaller P). The admittance characteristics will shift due to nonlinear effects.

[0082] Through the above technical solution, this invention achieves accurate quantification of the virtual impedance characteristics of grid-connected equipment. Traditional static impedance models only consider hardware parameters such as filters, neglecting the influence of controller dynamic characteristics on impedance. This solution establishes a bidirectional coupled admittance model considering phase-locked loop dynamics in a synchronous rotating coordinate system and introduces inter-axis cross-characteristic terms to characterize the active and reactive power coupling effect, accurately capturing the virtual impedance characteristics of the equipment under controlled conditions. Simultaneously, by introducing a nonlinear correction coefficient based on real-time power, the equivalent electrical characteristic matrix can adapt to different operating conditions, reflecting the impedance characteristic changes of the equipment from light load to heavy load. This accurate frequency domain modeling provides accurate controlled characteristic increments for subsequent admittance reconstruction, ensuring that the reconstructed real-state admittance model can truly reflect the grid electrical characteristics, including the equipment's virtual impedance.

[0083] Step 4: Admittance Reconstruction: The equivalent electrical characteristic matrix corresponding to each grid-connected new energy equipment is used as a controlled feature increment and injected into the obtained static physical topology parameters to reconstruct the real dynamic admittance model of the target distribution network, and the real node impedance matrix characterizing the electrical distance offset is obtained based on the real dynamic admittance model.

[0084] In one embodiment, the generation of the real-state node impedance matrix is ​​a key step in injecting controlled feature increments into the static topology and solving for the real-state impedance characteristics. The specific process includes:

[0085] The equivalent electrical characteristic matrix is ​​used as a controlled feature increment and input to the corresponding node position in the benchmark admittance model to construct a dual topology space model. The dual topology space model is then subjected to manifold inversion operation by a sparse matrix solver to generate a real node impedance matrix.

[0086] The reference admittance model refers to a traditional admittance matrix established based on static physical topology parameters. Specifically, it can be constructed using physical parameters such as conductor impedance and transformer impedance according to the nodal admittance method, and its matrix elements... Represents the admittance relationship between node i and node j, with diagonal elements. The self-admittance of node i (equal to the sum of the admittances of all branches connected to this node), and the off-diagonal elements. The mutual admittance between nodes i and j (equal to the negative of the branch admittance connecting the two nodes) serves to provide a basic electrical characteristic characterization of the power grid topology.

[0087] The dual topology space model refers to the comprehensive admittance matrix after integrating static topology and dynamic controlled features. Specifically, it can be represented by the equivalent electrical feature matrix. The results are obtained by superimposing the corresponding node numbers onto the corresponding positions in the reference admittance matrix. , where Δ This refers to the amount of controlled characteristic increments injected at the corresponding node location for each grid-connected device. Its function is to simultaneously reflect the grid characteristics affected by both the grid structure and equipment control.

[0088] The manifold inversion operation refers to a numerical computation method for efficiently solving the inverse of the admittance matrix while preserving matrix sparsity. Specifically, it can be implemented using sparse matrix techniques such as LU decomposition or Cholesky decomposition. Its function is to transform the admittance space into the impedance space to obtain the impedance relationships between nodes. .

[0089] Step 5: Sensitivity Assessment: By extracting the mutual impedance tensor from the real-state node impedance matrix, the real-state voltage sensitivity characteristics of the absorption node relative to the energy storage node are generated and the impedance offset index of the real-state voltage sensitivity characteristics relative to the preset initial sensitivity benchmark is determined.

[0090] In one embodiment, the process of generating real-state voltage sensitivity characteristics includes: calculating the dynamic projection magnitude of the mutual impedance tensor in the real-state node impedance matrix in the direction of the real-time voltage vector, and determining the dynamic projection magnitude as the real-state voltage sensitivity characteristics.

[0091] The dynamic projection magnitude refers to the projection length of the mutual impedance tensor vector onto the current voltage vector direction, which can be specifically calculated using vector dot product. And obtain the modulo value, where, Let be the real-time voltage vector of node j, and its function is to extract the effective sensitivity components related to the current operating state.

[0092] The real-state voltage sensitivity feature, by considering the real-time voltage vector direction, enables the sensitivity assessment to reflect the dynamic characteristics of the current operating point of the power grid. Its physical meaning is: when energy storage node i injects unit power, the voltage change amplitude of absorption node j in the current voltage direction is more real-time and accurate than the traditional static sensitivity analysis based on unit power disturbance.

[0093] In one embodiment, the sensitivity assessment, after determining the impedance offset index, further includes:

[0094] Extract the real-time temporal gradient of the impedance offset index within a preset time window; synchronously acquire the amplitude evolution characteristics of the inter-axis crossover feature; calculate the normalized cross-correlation coefficient between the real-time temporal gradient and the amplitude evolution characteristics, and generate source-grid attribution weight features to characterize the coupling degree between the equipment control layer and the power grid topology layer.

[0095] The real-time time-domain gradient refers to the derivative of the impedance offset exponent with time, which can be specifically obtained by analyzing time-series data. Perform difference operations The purpose of this method is to quantify the dynamic evolution rate of the power grid's sensitivity characteristics.

[0096] The amplitude evolution feature refers to the trend of the value of the inter-axis cross feature term changing over time, which can be specifically represented by a sliding window. Interaxial cross term amplitude Standard deviation or trend line slope Characterization, its role is to reflect the time-varying characteristics of the active and reactive coupling strength of the equipment control layer.

[0097] The normalized cross-correlation coefficient refers to a normalized measure of the correlation between two time series at different time offsets, which can be specifically calculated by the Pearson correlation coefficient:

[0098] ;

[0099] It is obtained by normalizing to the [0,1] interval, and its function is to quantify the causal relationship between impedance offset and device control coupling.

[0100] The physical meaning of the source-network attribution weight characteristic is as follows: When the characteristic value is large (e.g., ρ>0.7), it indicates that the change in impedance offset exponent is highly correlated with the evolution of the amplitude of the inter-axis cross term, suggesting that voltage risk mainly originates from the instability of active and reactive power coupling at the equipment control layer. In this case, the fundamental reason for the sensitivity shift is the drastic change in the virtual impedance characteristics of the equipment. When the characteristic value is small (e.g., ρ<0.3), it indicates that the correlation between the two is weak, suggesting that voltage risk mainly originates from the weak impedance of the grid structure rather than equipment control problems. In this case, the sensitivity shift reflects more the dominant role of static topology parameters on voltage characteristics. This attribution analysis capability enables the system to distinguish between source-side causes (equipment control instability) and grid-side causes (weak grid structure), providing a basis for decision-making for differentiated control strategies.

[0101] Through the above technical solution, this invention achieves the reconstruction of a real-state admittance model that integrates static topology and dynamic controlled features. Traditional admittance models only include the physical impedance of fixed equipment such as conductors and transformers, neglecting the virtual impedance effect of controlled equipment such as grid-connected inverters. This solution injects the equivalent electrical characteristic matrix as a controlled feature increment into the baseline admittance model, enabling the reconstructed admittance model to simultaneously reflect the dual influence of grid structure impedance and equipment virtual impedance. Furthermore, the real-state node impedance matrix is ​​obtained through manifold inversion, and the dynamic projection magnitude of the mutual impedance tensor is calculated as a real-state voltage sensitivity feature, providing an accurate sensitivity criterion for the precise correction of subsequent energy storage configuration parameters. This sensitivity assessment based on the real-state model can accurately reflect the impact of changes in equipment control parameters on grid voltage characteristics, solving the deficiency of traditional static sensitivity analysis in capturing dynamic characteristics.

[0102] Step 6: Configuration: Obtain the reference operating parameters of the energy storage system based on the static physical topology parameters, and perform nonlinear hedging correction on the reference operating parameters according to the impedance offset index to generate multidimensional coupling configuration data of the energy storage system, and adjust the power output state of the energy storage system.

[0103] In one embodiment, step 6, the process of determining the dynamic attribution state based on the source network attribution weight characteristics, includes:

[0104] The source-network attribution weight feature is compared with a preset judgment threshold. If the source-network attribution weight feature is greater than the judgment threshold, the voltage risk of the current absorption node is marked as a device-side cause state. If the source-network attribution weight feature is not greater than the judgment threshold, the voltage risk of the current absorption node is marked as a grid-side cause state.

[0105] The judgment threshold refers to the critical value that distinguishes between equipment-side causes and grid-side causes. Specifically, the attribution weight characteristic distribution of the two types of causes can be determined through historical data statistical analysis, and the intersection point of the two is taken as the judgment threshold (usually set to 0.5). Its function is to provide the decision boundary for attribution status determination.

[0106] The equipment-side induced state refers to the operating state where voltage risk is mainly caused by improper control parameters or control coupling instability of grid-connected equipment. Specifically, it manifests as a high correlation between the amplitude fluctuation of the inter-axis cross term and impedance offset. A typical scenario is that the phase-locked loop gain is set too high, leading to... If the shaft coupling is too strong, adjusting the equipment control parameters (such as reducing the PLL gain) is more effective than strengthening the space frame.

[0107] The grid-side induced state refers to the operating state where voltage risk is mainly caused by excessive line impedance or weak grid structure. Specifically, it is characterized by low coupling correlation between impedance offset changes and equipment control. A typical scenario is that the connection of a remote weak grid leads to excessive node impedance, which requires improving energy storage output or strengthening the grid to improve voltage quality.

[0108] After marking the inducing state and before generating multidimensional coupling configuration data, the configuration module performs the following differential correction logic:

[0109] If the current state is marked as a device-side induced state, the proportional gain used to achieve the nonlinear hedging correction is dynamically adjusted to enhance the energy storage system's ability to suppress internal coupling instability.

[0110] If the current state is marked as a grid-side induced state, the proportional gain of the nonlinear hedging correction remains unchanged, and the upper limit of the power change rate in the multidimensional coupling configuration data is forcibly limited to compensate for insufficient electrical damping on the grid side.

[0111] The physical significance of the proportional gain adjustment mechanism lies in the following: when the cause is determined to be equipment-side, it indicates that the voltage fluctuation is mainly caused by equipment control coupling. In this case, the energy storage system needs to respond more quickly and forcefully to suppress this coupling instability. Therefore, the proportional gain is increased. From the benchmark value Increase to , where β is the adjustment factor (e.g., β=0.3 means a gain increase of 30%). This adjustment enables the energy storage system to track voltage fluctuations with a higher response speed and quickly inject counteracting power to offset voltage disturbances caused by device coupling.

[0112] The physical significance of the power change rate limiting mechanism lies in the fact that when the cause is determined to be on the grid side, it indicates that the grid impedance is too high, resulting in insufficient electrical damping. In this case, rapid changes in energy storage power may trigger grid resonance and cause voltage oscillations. Therefore, it is necessary to limit the power change rate. To ensure stable power output from energy storage, among which, Based on the impedance characteristics and damping coefficient of the grid structure, this limitation avoids resonance between high-frequency power changes in energy storage and weak points in the grid structure, ensuring the stability of voltage regulation.

[0113] Through the above technical solution, this invention realizes a differentiated energy storage configuration strategy based on source-grid attribution analysis. Traditional methods employ a uniform fixed regulation logic, which cannot distinguish the root cause of voltage risk, leading to a misalignment between the regulation strategy and the actual cause. This solution, by calculating the cross-correlation coefficient between the time-domain gradient of the impedance offset exponent and the amplitude evolution characteristics of the inter-axis cross term, can accurately determine whether the voltage risk originates from equipment control layer instability or grid structure weakness. Based on the attribution results, a differentiated correction logic is applied: for equipment-side causes, the rapid response capability of energy storage is enhanced to suppress control coupling; for grid-side causes, the rate of change of energy storage is limited to compensate for insufficient grid damping. This attribution-driven adaptive regulation strategy effectively solves the problem of regulation failure or over-regulation caused by the inability of traditional fixed regulation logic to identify the root cause of risk, significantly improving the accuracy and adaptability of the energy storage system's hedging correction, and ensuring that energy storage configuration parameters can specifically address different types of voltage risks.

[0114] Step 7: Calibration, used to recursively calibrate the mapping parameters of the equivalent electrical characteristic matrix based on the measured voltage feedback deviation after running the multidimensional coupling configuration data. The recursive calibration of the mapping parameters proceeds to step 3, and so on.

[0115] Recursive calibration refers to an adaptive process that iteratively updates the mapping parameters based on the feedback deviation. Specifically, it can be implemented through gradient descent or Kalman filtering algorithms. Its role is to enable the equivalent electrical feature matrix to continuously track the actual dynamic characteristics of the equipment, realize the self-evolution of the model, and thus adapt to the complex and variable characteristics of the power grid operation.

[0116] In one embodiment, the self-evolutionary mechanism is a key innovation for achieving the co-evolution of model parameter calibration and attribution analysis logic. The closed-loop recursive calibration also performs self-evolutionary flow logic, as follows:

[0117] The consistency of the mapping parameter correction amount generated by recursive calibration with the source network attribution weight feature is verified to determine the deviation between the two in the feedback adjustment direction.

[0118] Based on the deviation, the weighting of the source network attribution weight feature with respect to the time-domain change gradient is adjusted in reverse, and the calibrated attribution weight feature is synchronously fed back to the admittance reconstruction module and the sensitivity evaluation module.

[0119] The mapping parameter correction refers to the adjustment value of the equivalent electrical characteristic matrix parameters calculated by the calibration module based on the voltage feedback deviation, which may specifically include the control gain correction. Filter parameter correction amount Its function is to make the model parameters approximate the actual dynamic characteristics of the device.

[0120] The consistency verification refers to the process of verifying the logical consistency between the direction of the mapping parameter correction and the source-network attribution determination result. Specifically, it can be done by judging whether the parameter correction is aimed at reducing device-side coupling (such as reducing...). To reduce interaxial cross terms or compensate for the grid-side impedance (e.g., by increasing...) This is achieved by offsetting the network impedance, and its function is to detect whether the calibration process is consistent with the attribution analysis conclusions.

[0121] The feedback adjustment direction deviation refers to the angle or difference measure between the mapping parameter correction direction and the attribution weight feature indication direction, which can be specifically defined by defining a direction vector. ;

[0122] Calculate the cosine of the angle between the two. When cosθ is close to 1, it indicates that the directions are consistent; when it is close to -1, it indicates that the directions are contradictory. Its function is to quantify the degree of contradiction between the calibration results and the attribution judgment.

[0123] Weighted weight adjustment refers to the process of modifying the proportion of the time-domain change gradient in the attribution weight feature calculation formula, which can be achieved by adjusting the proportion based on the deviation.

[0124] The original calculation formula ;

[0125] Modified to ;

[0126] in, κ represents the adjustment gain, which is used to enable the attribution analysis mechanism to adaptively calibrate in order to improve the accuracy of the judgment.

[0127] The synchronous feedback mechanism refers to the information flow path that transmits the calibrated attribution weight features back to the admittance reconstruction and sensitivity assessment module. Specifically, it can be achieved by updating the shared parameters between modules. Its function is to enable the entire system to iteratively optimize based on the latest calibration results, forming a self-evolving closed loop.

[0128] The underlying logic of the self-evolutionary flow mechanism lies in the following: when the mapping parameter correction points to increasing the equipment control gain (indicating that the actual control characteristics of the equipment are stronger than the model's expectations, ΔK_p>0), but the source network attribution weight characteristics determine it as a network-side cause (indicating that the attribution analysis underestimates the impact of equipment control, ρ<0.5), a directional deviation cosθ<0 is generated. Based on this deviation, through... Inversely increasing the weighting α of the time-domain gradient change makes the attribution mechanism more effective. It is more sensitive, making it easier to capture impedance shifts caused by device control in the next iteration. The attribution judgment will be more inclined to device-side causes (increased ρ), thus becoming consistent with the calibration results.

[0129] This two-way feedback mechanism enables the co-evolution of model parameter calibration and attribution analysis mechanisms, ensuring that the system as a whole maintains internal logical consistency and avoiding the chaos of control strategies caused by contradictions between model parameters and attribution judgments.

[0130] Through the above technical solution, this application realizes a self-evolving collaborative mechanism for model parameter calibration and source-grid attribution analysis. In traditional methods, model parameter calibration and attribution analysis logic are independent of each other, which may lead to contradictions where calibration results point to equipment-side problems but attribution is determined to be grid-side causes, resulting in the failure of control strategies. This solution detects the contradiction between the direction of mapping parameter correction and the attribution determination results through consistency verification, and adjusts the attribution weight feature calculation logic in reverse based on the deviation, so that the calibration process not only updates model parameters, but also simultaneously optimizes the attribution analysis mechanism itself. This self-evolving evolutionary flow design endows the system with continuous learning and self-optimization capabilities, enabling it to continuously improve model accuracy and attribution accuracy during long-term operation, thereby adapting to the continuous changes in grid operating conditions. It solves the technical bottleneck of traditional fixed-parameter models lacking self-evolving capabilities, ensuring that the energy storage configuration system can maintain high accuracy and high adaptability in complex and ever-changing actual operating conditions.

[0131] Example

[0132] In a distribution network with a high proportion of renewable energy, photovoltaic (PV) capacity accounts for 45%, wind power capacity accounts for 30%, and energy storage system capacity accounts for 20% of the total grid load. The distribution network uses a 10kV voltage level, with a total main line length of approximately 50 kilometers. The PV connection point is located at the end of the main line, the wind power connection point is located in the middle, and the energy storage system is located at the substation outlet.

[0133] In the synchronous configuration, PMU devices deployed at each grid-connected node acquire operating status data streams such as voltage and current at a sampling frequency of 100Hz. Simultaneously, the system reads the internal controlled characteristic parameter set from the photovoltaic inverter and wind turbine controller via the IEC 61850 communication protocol, including the current loop proportional gain. Integral gain Filter inductor Phase-locked loop gain Parameters such as... At a certain moment, the PMU sampling timescale was detected as T1=10:30:25.123, the controller communication timescale as T2=10:30:25.156, and the dynamic time deviation Δt=33ms. The system started the interpolation prediction algorithm, using cubic spline interpolation to align the running data stream, and monitored the rate of change of time deviation over the past 10 seconds. Based on this, the buffer trigger frequency was adjusted in reverse from 100Hz to 95Hz to compensate for the increasing trend of communication latency.

[0134] After time-domain synchronization is completed, an approximate entropy analysis is performed on the voltage waveform of the grid-connected node to calculate the first signal entropy value. This value exceeded the preset threshold of 0.6, indicating that the voltage waveform exhibited significant non-stationary oscillation characteristics. The system immediately triggered a resampling request, increasing the control parameter sampling frequency from 1Hz to 10Hz, and intensively acquiring 50 sets of control parameter data over the next 5 seconds, capturing... The complete process of fluctuation from 50 to 65 and then back to 55 forms the first feedback closed loop.

[0135] Frequency mapping modeling is based on the target composite feature vector, and an equivalent electrical feature matrix is ​​constructed in a synchronous rotating coordinate system.

[0136] Substitute the current parameters , , , , (Rated power of equipment) (Currently operating at 80% load), nonlinear correction factor The equivalent admittance at the fundamental frequency is calculated as follows:

[0137] .

[0138] The inter-axis crossing characteristic term was obtained through phase-locked loop dynamic analysis:

[0139] ;

[0140] This term characterizes the coupling effect of d-axis voltage variation on q-axis current.

[0141] Admittance reconstruction involves injecting the equivalent electrical characteristic matrix of each grid-connected device into the baseline admittance model according to its node location. The baseline admittance model is constructed based on the line parameters (resistance per unit length r = 0.27 Ω / km, reactance x = 0.35 Ω / km) and topology, resulting in a 30×30 dimension node admittance matrix. Photovoltaic inverters Inject at the 25th node, the wind turbine Inject the 15th node to construct the dual topological space model:

[0142] .

[0143] The real-state nodal impedance matrix is ​​obtained by inverting the manifold using sparse LU decomposition.

[0144] .

[0145] Extract the mutual impedance between the energy storage node (node ​​1) and the photovoltaic absorption node (node ​​25):

[0146] In real-time voltage vector Calculate dynamic projection magnitude in direction .

[0147] The initial sensitivity baseline is calculated in the sensitivity assessment. The static impedance matrix is ​​calculated based on the static topology (excluding controlled feature increments). their mutual impedance The corresponding static sensitivity is 2.11Ω. The impedance offset index δ=(2.48-2.11) / 2.11=0.175, which means that the real sensitivity is 17.5% different from the static sensitivity.

[0148] Extract the impedance offset exponential sequence within the past 60-second time window. Calculate its time-domain gradient. Simultaneously extract the amplitude sequence of inter-axis cross feature terms. Calculate its standard deviation The normalized cross-correlation coefficient ρ between the two is calculated to be 0.78, which exceeds the judgment threshold of 0.5. Therefore, the voltage risk of the current absorption node is marked as the equipment-side cause state.

[0149] The configuration is adjusted based on the attribution status. Since the cause is determined to be equipment-side, the proportional gain of the energy storage system is adjusted from the baseline value. Adjust to While maintaining power change rate limit Unchanged. Combined with impedance offset index. For the baseline action parameters generated based on static topology Nonlinear hedging correction is performed, and the corrected power command is... Multidimensional coupled configuration data is generated and sent to the energy storage system.

[0150] After the energy storage system was configured according to the data, the measured photovoltaic grid connection node voltage recovered from 9.8kV to 10.1kV, which deviates from the 10.15kV predicted based on the real-state model. Based on this feedback deviation, the calibration module calculates the correction amount for the mapping parameters using the gradient descent algorithm. (Note that the actual control gain is slightly lower than the model's expectation.) Consistency checks revealed that... The attribution weight feature ρ=0.78 points to the weakening of equipment control characteristics, while the attribution weight feature ρ=0.78 points to equipment-side causes, indicating a deviation between the two. The directional deviation cosθ=0.35 is calculated, and based on this, the weight coefficient α of the time-domain gradient in the attribution weight calculation formula is increased from 0.6 to 0.68. The calibrated attribution weight feature is then synchronously fed back to the admittance reconstruction module and the sensitivity evaluation module to achieve collaborative optimization in the next iteration.

[0151] Through the above scheme, this application achieves precise matching between the energy storage configuration process and the actual dynamic characteristics of the power grid. Compared with traditional energy storage configuration methods based on static topology, this scheme can accurately capture the impact of the virtual impedance characteristics of grid-connected equipment on voltage sensitivity, identify the root causes of voltage risks through source-grid attribution analysis, and implement differentiated control strategies. In this embodiment, the energy storage system operates according to the corrected configuration data, successfully stabilizing the photovoltaic absorption node voltage within ±3% of the rated value, which is a significant improvement compared to the ±8% fluctuation range of traditional methods, and no regulation oscillation phenomenon occurs. At the same time, the self-evolving evolutionary flow mechanism continuously optimizes model parameters and attribution logic through recursive calibration, enabling the system to maintain high accuracy and high adaptability during long-term operation.

[0152] Through the above scheme, this invention achieves comprehensive adaptation between the energy storage configuration process and the actual dynamic characteristics of the power grid. The introduction of a time-domain synchronization mechanism and signal correlation feature extraction solves the problem of asynchronous operation data and equipment control parameters, providing a spatiotemporally consistent data foundation for accurate modeling. The construction of an equivalent electrical characteristic matrix considering phase-locked loop dynamics overcomes the limitation of traditional static impedance models in reflecting the virtual impedance of equipment, accurately quantifying the inter-axis coupling effect of active and reactive power. By reconstructing admittance, a real-state impedance matrix integrating the dual influences of the grid and equipment is generated, enabling voltage sensitivity assessment to reflect the true dynamic characteristics of the power grid. The introduction of a source-grid attribution analysis mechanism achieves accurate identification of causes on both the equipment and grid sides, and differentiated control strategies are implemented based on the attribution results, effectively improving the targeting of energy storage offsetting corrections. The establishment of a self-evolving evolutionary flow mechanism endows the system with continuous learning and self-optimization capabilities, ensuring high precision and adaptability in complex and ever-changing actual operating conditions. This adaptive mechanism improves the accuracy and real-time performance of power offsetting in energy storage systems, reduces the risk of voltage exceeding limits, enhances the absorption capacity of new energy sources, and provides reliable technical support for the safe and stable operation of distribution networks in scenarios with a high proportion of new energy sources.

[0153] This invention overcomes the technical bottleneck of traditional static models failing to reflect the virtual impedance of equipment under controlled conditions. By introducing inter-axis cross-feature terms to capture active and reactive power coupling effects, and based on source-grid attribution weight features, it achieves differentiated identification of causes on the equipment side and grid side. This enables energy storage configuration parameters to be adaptively adjusted to match the real-state dynamic characteristics of the power grid, thereby solving the technical defect of traditional fixed regulation logic in identifying the root causes of risks. It provides a precise and reliable energy storage optimization configuration scheme for the safe and stable operation of distribution networks in high-proportion renewable energy scenarios.

Claims

1. A power distribution network energy storage optimal configuration method based on new energy consumption, characterized in that: Includes the following steps: Step 1: Running status data stream and internal controlled feature parameter set; Step 2: Synchronization Feature Construction: Extract signal correlation features of grid-connected nodes from the acquired operating status data stream. The signal correlation features perform time-domain synchronization processing on the operating status data stream and the acquired internal controlled feature parameter set to generate a target composite feature vector. Step 3: Frequency mapping modeling: Based on the target composite feature vector and the preset mapping parameters, frequency domain characteristic mapping processing is performed to construct the equivalent electrical feature matrix of grid-connected new energy equipment in the coordinate transformation dimension. The equivalent electrical feature matrix includes inter-axis cross feature terms that characterize the coupling interaction between active and reactive power dynamics. Step 4: Admittance Reconstruction: The equivalent electrical characteristic matrix corresponding to each grid-connected new energy equipment is used as a controlled feature increment and injected into the obtained static physical topology parameters to reconstruct the real dynamic admittance model of the target distribution network, and the real node impedance matrix characterizing the electrical distance offset is obtained based on the real dynamic admittance model. Step 5: Sensitivity assessment: By extracting the mutual impedance tensor from the real-state node impedance matrix, the real-state voltage sensitivity characteristics of the absorption node relative to the energy storage node are generated and the impedance offset index of the real-state voltage sensitivity characteristics relative to the preset initial sensitivity benchmark is determined. Step 6: Configuration: Obtain the reference operating parameters of the energy storage system based on the static physical topology parameters, and perform nonlinear hedging correction on the reference operating parameters according to the impedance offset index to generate multidimensional coupling configuration data of the energy storage system and adjust the power output state of the energy storage system. Step 7: Calibration, used to recursively calibrate the mapping parameters of the equivalent electrical characteristic matrix based on the measured voltage feedback deviation after running the multidimensional coupling configuration data. The recursive calibration of the mapping parameters proceeds to step 3, and so on. 2.The method of claim 1, wherein: In step 2, the specific process of signal association feature extraction is as follows: Step 201: Extract non-stationary oscillation features from the voltage waveform using an approximate entropy algorithm, and determine the first signal entropy value characterizing the intensity of controlled impedance fluctuations. The approximate entropy algorithm refers to a nonlinear analysis method that quantifies the complexity and irregularity of time series. The non-stationary oscillation features refer to the dynamic oscillation modes in the voltage waveform whose frequency or amplitude changes with time. The first signal entropy value is a scalar index that quantifies the complexity of the voltage waveform. Step 202: Based on the first signal entropy value, dynamically trigger a resampling request for the centralized control gain weight of the internal controlled feature parameters, and generate a first feedback closed loop flow that changes the parameter perception depth in real time according to the running state. The resampling request refers to the instruction signal that triggers high-frequency acquisition of control parameters according to the change of signal entropy value. The first feedback closed loop flow refers to the iterative loop mechanism of signal entropy value monitoring, resampling triggering, parameter updating and entropy value re-evaluation. 3.The method of claim 2, wherein: The specific process of the time-domain synchronization process is as follows: Step 211: Obtain the first sampling time stamp of the grid-connected node and the second communication time stamp of the internal controlled feature parameter set; determine the dynamic time deviation between the first sampling time stamp and the second communication time stamp, wherein the first sampling time stamp refers to the timestamp corresponding to each sampling point in the running status data stream, the second communication time stamp refers to the timestamp carried when the control parameters are transmitted to the data processing unit through the communication protocol, and the dynamic time deviation refers to the time difference between the two time stamps and its timing change characteristics; Step 212: Align the running status data stream using a preset interpolation prediction algorithm, and reversely correct the cache trigger frequency of the running status data stream collected in the synchronization feature module according to the rate of change of the dynamic time deviation. The interpolation prediction algorithm refers to a mathematical method for predicting missing time values ​​based on existing data points.

4. The method for optimal configuration of power distribution network energy storage based on new energy consumption according to claim 3, characterized in that: In step 3, the construction process of the equivalent electrical characteristic matrix is ​​as follows: Based on the target composite feature vector, a two-way coupled admittance model considering the dynamic characteristics of the phase-locked loop is established in a synchronous rotating coordinate system; The synchronous rotating coordinate system refers to a dq coordinate frame that rotates synchronously with the power grid's base frequency. The bidirectional coupled admittance model refers to an admittance matrix expression that simultaneously includes the coupling paths from the d-axis to the q-axis and from the q-axis to the d-axis. The equivalent electrical characteristic matrix is ​​quantified using the following formula: ， wherein, represents an equivalent electrical characteristic matrix, s represents a Laplacian operator, represents a filter inductance parameter, and respectively represent a proportional gain and an integral gain of the current loop control, j represents an imaginary unit, used to describe phase and complex number characteristics in an alternating current system, ω0represents a fundamental frequency, and Θ(P) represents a non-linear correction coefficient determined based on a real-time output active power P.

5. The method for optimizing the configuration of energy storage in distribution networks based on renewable energy consumption as described in claim 4, characterized in that: In step 4, the specific process of generating the real-state nodal impedance matrix is ​​as follows: The equivalent electrical feature matrix is ​​used as a controlled feature increment and input to the corresponding node position in the benchmark admittance model to construct a dual topology space model; the dual topology space model is then subjected to manifold inversion operation by a sparse matrix solver to generate a real node impedance matrix. The baseline admittance model refers to the traditional admittance matrix established based on static physical topology parameters. The dual topology space model refers to the comprehensive admittance matrix after integrating static topology and dynamic controlled features. The manifold inversion operation refers to the numerical calculation method for solving the inverse of the admittance matrix while maintaining matrix sparsity.

6. The method for optimizing the configuration of energy storage in a distribution network based on renewable energy consumption as described in claim 5, characterized in that: In step 5, the process of generating real-state voltage sensitivity characteristics includes: calculating the dynamic projection magnitude of the mutual impedance tensor in the real-state node impedance matrix in the direction of the real-time voltage vector, and determining the dynamic projection magnitude as the real-state voltage sensitivity characteristics; The dynamic projection modulus refers to the projection length of the mutual impedance tensor vector in the direction of the current voltage vector. The real-state voltage sensitivity characteristic, through the real-time voltage vector direction, enables the sensitivity assessment to reflect the dynamic characteristics of the current operating point of the power grid.

7. The method for optimizing the configuration of energy storage in distribution networks based on renewable energy consumption as described in claim 6, characterized in that: The sensitivity assessment, after determining the impedance offset index, further includes: Extract the real-time time-domain change gradient of the impedance offset index within a preset time window; synchronously acquire the amplitude evolution characteristics of the inter-axis crossover feature; calculate the normalized cross-correlation coefficient between the real-time time-domain change gradient and the amplitude evolution characteristics, and generate source-grid attribution weight features to characterize the coupling degree between the equipment control layer and the power grid topology layer; The real-time time-domain gradient refers to the derivative of the impedance offset exponent with time; the amplitude evolution characteristic refers to the trend of the value of the inter-axis crossover characteristic term with time; and the normalized cross-correlation coefficient refers to the normalized measure of the correlation between two time series at different time offsets.

8. The method for optimizing the configuration of energy storage in a distribution network based on renewable energy consumption as described in claim 7, characterized in that: The process of determining the dynamic attribution state based on the source-network attribution weight characteristics includes: The source network attribution weight features are compared with a preset judgment threshold, which is the critical value that distinguishes between equipment-side causes and network structure-side causes. If the source-network attribution weight feature is greater than the judgment threshold, then the voltage risk of the current absorption node is marked as the equipment-side cause state. The equipment-side cause state refers to the operating state in which the voltage risk is mainly caused by improper control parameters of grid-connected equipment or control coupling instability. If the source network attribution weight feature is not greater than the judgment threshold, then the voltage risk of the current absorption node is marked as a grid-side induced state. The grid-side induced state refers to the operating state in which the voltage risk is mainly caused by excessive line impedance or weak grid structure.

9. The method for optimizing the configuration of energy storage in a distribution network based on renewable energy consumption as described in claim 8, characterized in that: In step 6, after marking the trigger state and before generating the multidimensional coupling configuration data, the following differential correction logic is performed: If the current state is marked as a device-side induced state, the proportional gain used to achieve the nonlinear hedging correction is dynamically adjusted to enhance the energy storage system's ability to suppress internal coupling instability. If the current state is marked as a grid-side induced state, the proportional gain of the nonlinear hedging correction remains unchanged, and the upper limit of the power change rate in the multidimensional coupling configuration data is forcibly limited to compensate for insufficient electrical damping on the grid side.

10. The method for optimizing the configuration of energy storage in a distribution network based on renewable energy consumption as described in claim 9, characterized in that: In step 7, the closed-loop recursive calibration also performs self-evolutionary flow logic: The consistency of the mapping parameter correction amount generated by recursive calibration with the source network attribution weight feature is verified to determine the deviation between the two in the feedback adjustment direction. The mapping parameter correction amount refers to the equivalent electrical feature matrix parameter adjustment value calculated by the calibration module based on the voltage feedback deviation. The feedback adjustment direction deviation refers to the angle or difference measure between the mapping parameter correction direction and the direction indicated by the attribution weight feature. Based on the deviation, the weighting weights of the source network attribution weight features related to the temporal change gradient are adjusted in reverse, and the calibrated attribution weight features are synchronously fed back to admittance reconstruction and sensitivity assessment. The weighting weights refer to the proportion of the temporal change gradient in the attribution weight feature calculation formula.

Images