Energy storage system scheduling method, device, equipment and system

Abstract

The invention provides an energy storage system scheduling method, device, equipment and system, and the method comprises the steps: obtaining an operation parameter of an energy storage system, and constructing a multi-dimensional disturbance structure tensor according to the operation parameter; according to the multi-dimensional disturbance structure tensor, establishing a neural synaptic stress structure simulation network model, and according to the neural synaptic stress structure simulation network model, obtaining a synaptic response mapping matrix; performing channel excitation on the current disturbance structure tensor based on the synaptic response mapping matrix to obtain an excitation response sequence; determining a strategy behavior tensor path based on the excitation response sequence and the current disturbance structure tensor; and generating a control instruction according to the strategy behavior tensor path and the strategy evaluation function, and controlling the energy storage system to execute the control instruction. The time sequence sensing capability of the dynamic characteristics of the energy storage system is improved, the problem that a traditional method is insufficient in heterogeneous parameter coupling relation expression capability is solved, differential selection of behavior paths is achieved, and the scheduling effect of the energy storage system is improved.

Description

Technical Field

[0001] This invention relates to the field of vehicle control technology, and specifically to a method, apparatus, equipment and system for scheduling energy storage systems. Background Technology

[0002] With the large-scale integration of new energy sources, the volatility of the power system has intensified. As a key means of peak shaving and valley filling, energy storage faces challenges such as multiple disturbances, high time-varying characteristics, and nonlinear response in its regulation strategies.

[0003] Current energy storage regulation mostly adopts fixed threshold rules or is based on single variable modeling, lacking the ability to identify the system's coupling relationship between complex disturbances, resulting in lag in strategy response, single behavior path, and limited regulation effect. Summary of the Invention

[0004] In view of this, the present invention aims to provide an energy storage system scheduling method, apparatus, equipment and system to solve the problems in the prior art that lack the ability to identify the system coupling relationship between complex disturbances, resulting in lag in strategy response, single behavior path and limited regulation effect.

[0005] This invention provides a method for scheduling an energy storage system, the method comprising: Obtain the operating parameters of the energy storage system, and construct a multidimensional perturbation structure tensor based on the operating parameters; Based on the multidimensional perturbation structure tensor, a neural synaptic stress response structure network model is established, and based on the neural synaptic stress response structure network model, a synaptic response mapping matrix is ​​obtained. Based on the synaptic response mapping matrix, the current perturbation structure tensor is channel-excited to obtain an excitation response sequence; Based on the excitation response sequence and the current perturbation structure tensor, determine the policy behavior tensor path; Based on the policy behavior tensor path and policy evaluation function, control commands are generated, and the energy storage system is controlled to execute the control commands.

[0006] In one possible embodiment, the step of establishing a simulated neural synaptic stress structure network model based on the multidimensional perturbation structure tensor, and obtaining a synaptic response mapping matrix based on the simulated neural synaptic stress structure network model, includes: Obtain a preset basic model structure, wherein the basic model structure includes synapses and neurons; The multidimensional perturbation structure tensor is decomposed to obtain the core tensor and factor matrix; Based on the core tensor and factor matrix, tensor features are obtained; The tensor features are converted into the basic model structure parameters using a preset function; Based on the aforementioned basic model structure and basic model structure parameters, a neural synaptic stress response structure network model is established. The multidimensional perturbation structure tensor is input into the neural synaptic stress structure network model to obtain the synaptic response mapping matrix.

[0007] In one possible embodiment, the step of channel-exciting the current perturbed structure tensor based on the synaptic response mapping matrix to obtain an excitation response sequence includes: The current perturbation structure tensor is input into the neural synaptic stress structure network model; The current perturbation structure tensor is projected onto the target neuron of the neural synaptic stress structure network model using the synaptic response mapping matrix. The target neuron is used to perform excitation calculations to obtain an excitation response sequence.

[0008] In one possible embodiment, determining the policy behavior tensor path based on the excitation response sequence and the current perturbation structure tensor includes: Based on the excitation response sequence, the main channel is identified; Calculate the acceleration trend of the main channel and the stability index of the current perturbation structure tensor; The strategy behavior tensor path is determined based on the acceleration of the main channel and the stability index of the current perturbation structure tensor.

[0009] In one possible embodiment, generating control commands based on the policy behavior tensor path and policy evaluation function, and controlling the energy storage system to execute the control commands, includes: The parameters corresponding to the policy behavior tensor path are input into the policy evaluation function to obtain the policy evaluation value; Based on the strategy evaluation value, a control strategy is obtained, and based on the control strategy, control instructions are generated; Control the energy storage system to execute the control commands.

[0010] In one possible embodiment, obtaining a control policy based on the policy evaluation value and generating control instructions based on the control policy includes: The strategy evaluation value is normalized to obtain a normalized strategy evaluation value; The execution path is determined based on the normalized strategy evaluation value; Control instructions are generated based on the execution path.

[0011] In one possible embodiment, obtaining the operating parameters of the energy storage system and constructing a multidimensional perturbation structure tensor based on the operating parameters includes: The operating parameters of the energy storage system are collected through the sensing and acquisition unit; The running parameters under the same timestamp are arranged into a multidimensional tensor according to a preset dimension; The multidimensional tensor is divided into N window tensors using a sliding window mechanism; A pre-defined graph neural network is used to fuse N window tensors to obtain a multidimensional perturbation structure tensor, where N is a positive integer greater than 1.

[0012] In a second aspect, the present invention provides an energy storage system scheduling device, the device comprising: The acquisition module is used to acquire the operating parameters of the energy storage system and construct a multidimensional perturbation structure tensor based on the operating parameters. A module is established to build a simulated neural synaptic stress structure network model based on the multidimensional perturbation structure tensor, and to obtain a synaptic response mapping matrix based on the simulated neural synaptic stress structure network model. The channel excitation module is used to excite the current perturbation structure tensor through the channel based on the synaptic response mapping matrix to obtain an excitation response sequence. The determination module is used to determine the policy behavior tensor path based on the excitation response sequence and the current perturbation structure tensor; The generation module is used to generate control commands based on the policy behavior tensor path and policy evaluation function, and to control the energy storage system to execute the control commands.

[0013] Thirdly, this application provides an electronic device, the device comprising: a memory and a processor; the memory is used to store related program code; the processor is used to call the program code to execute the energy storage system scheduling method described in any of the implementations of the first aspect.

[0014] Fourthly, this application provides an energy storage system in which the electronic device described in the third invention is provided.

[0015] Fifthly, this application provides a computer-readable storage medium for storing a computer program for executing the energy storage system scheduling method described in any implementation of the first aspect.

[0016] Sixthly, this application provides a computer program product, which includes a computer program / instruction, and when the computer program / instruction is executed by a processor, it implements the energy storage system scheduling method described in any of the implementations of the first aspect above.

[0017] In the above implementation of the present invention, the operating parameters of the energy storage system are obtained, and a multidimensional perturbation structure tensor is constructed based on the operating parameters; a neural synaptic stress-inducing network model is established based on the multidimensional perturbation structure tensor, and a synaptic response mapping matrix is ​​obtained based on the neural synaptic stress-inducing network model; channel excitation is performed on the current perturbation structure tensor based on the synaptic response mapping matrix to obtain an excitation response sequence; a strategy behavior tensor path is determined based on the excitation response sequence and the current perturbation structure tensor; a control command is generated based on the strategy behavior tensor path and the strategy evaluation function, and the energy storage system is controlled to execute the control command. Through the method provided by the present invention, a neural synaptic stress-inducing network model can be established based on a multidimensional perturbation structure tensor, thereby obtaining a synaptic response mapping matrix; channel excitation is performed on the current perturbation structure tensor based on the synaptic response mapping matrix to obtain an excitation response sequence; a strategy behavior tensor path is determined based on the excitation response sequence and the current perturbation structure tensor, thereby generating a control command based on the strategy behavior tensor path and the strategy evaluation function. This paper proposes a method to achieve multi-parameter fusion modeling of the operating state of an energy storage system using a multi-dimensional perturbation structure tensor. This significantly improves the time-series perception capability of the dynamic characteristics of the energy storage system and overcomes the problem of insufficient ability to express the coupling relationship of heterogeneous parameters in traditional methods. By exciting channels based on the synaptic response mapping matrix, the selection elasticity of tensor paths under perturbation forcing conditions is enhanced, breaking through the technical bottlenecks of existing control strategies' update lag and response rigidity. By introducing a strategy evaluation function and determining control commands based on the strategy evaluation function and the strategy behavior tensor path, differentiated selection of behavior paths under the strategy evaluation function is achieved, thereby improving the scheduling effect of the energy storage system. Attached Figure Description

[0018] Figure 1 A flowchart of an energy storage system scheduling method provided in an embodiment of the present invention.

[0019] Figure 2 The flowchart provided in this embodiment of the invention describes how to establish a simulated neural synaptic stress structure network model based on the multidimensional perturbation structure tensor, and how to obtain a synaptic response mapping matrix based on the simulated neural synaptic stress structure network model.

[0020] Figure 3 A schematic diagram of an energy storage system scheduling device provided in an embodiment of the present invention.

[0021] Figure 4 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] An embodiment of the present invention provides a method for scheduling an energy storage system. The method involves acquiring the operating parameters of the energy storage system and constructing a multidimensional perturbation structure tensor based on these parameters; establishing a neural synaptic stress-inducing network model based on the multidimensional perturbation structure tensor; obtaining a synaptic response mapping matrix based on the neural synaptic stress-inducing network model; performing channel excitation on the current perturbation structure tensor based on the synaptic response mapping matrix to obtain an excitation response sequence; determining a policy behavior tensor path based on the excitation response sequence and the current perturbation structure tensor; generating control commands based on the policy behavior tensor path and a policy evaluation function; and controlling the energy storage system to execute the control commands. The method provided by this invention allows for the establishment of a neural synaptic stress-inducing network model based on a multidimensional perturbation structure tensor, thereby obtaining a synaptic response mapping matrix; performing channel excitation on the current perturbation structure tensor based on the synaptic response mapping matrix to obtain an excitation response sequence; determining a policy behavior tensor path based on the excitation response sequence and the current perturbation structure tensor; and generating control commands based on the policy behavior tensor path and a policy evaluation function. This paper proposes a method to achieve multi-parameter fusion modeling of the operating state of an energy storage system using a multi-dimensional perturbation structure tensor. This significantly improves the time-series perception capability of the dynamic characteristics of the energy storage system and overcomes the problem of insufficient ability to express the coupling relationship of heterogeneous parameters in traditional methods. By exciting channels based on the synaptic response mapping matrix, the selection elasticity of tensor paths under perturbation forcing conditions is enhanced, breaking through the technical bottlenecks of existing control strategies' update lag and response rigidity. By introducing a strategy evaluation function and determining control commands based on the strategy evaluation function and the strategy behavior tensor path, differentiated selection of behavior paths under the strategy evaluation function is achieved, thereby improving the scheduling effect of the energy storage system.

[0024] Please see Figure 1 In one exemplary embodiment, an energy storage system scheduling method is provided, applied to an energy storage system including multiple energy storage units, such as lithium battery packs and supercapacitors, as well as associated power converters, sensors, and communication modules. In an embodiment of the present invention, a 2MWh lithium iron phosphate battery pack is used as the energy storage system. The method may include the following steps: S101: Obtain the operating parameters of the energy storage system and construct a multidimensional perturbation structure tensor based on the operating parameters.

[0025] In practice, the operating parameters of the energy storage system can be collected through sensing and acquisition units. These operating parameters include dynamic parameters such as the grid frequency change rate, battery current fluctuation, battery temperature gradient, and environmental weather disturbance rate. After obtaining the operating parameters, a multidimensional disturbance structure tensor can be constructed based on them.

[0026] Specifically, the process of obtaining the operating parameters of the energy storage system and constructing a multidimensional perturbation structure tensor based on the operating parameters may include: The operating parameters of the energy storage system are collected through the sensing and acquisition unit; The running parameters under the same timestamp are arranged into a multidimensional tensor according to a preset dimension; The multidimensional tensor is divided into N window tensors using a sliding window mechanism; A pre-defined graph neural network is used to fuse N window tensors to obtain a multidimensional perturbation structure tensor, where N is a positive integer greater than 1.

[0027] Specifically, the operating parameters of the energy storage system at the same timestamp are arranged into a multidimensional tensor according to a preset dimension; a sliding window mechanism is used to divide the multidimensional tensor into several window tensors; a preset graph neural network is used to fuse the several window tensors to obtain a multidimensional perturbation structure tensor.

[0028] It should be noted that the energy storage system includes multiple energy storage units, such as lithium battery packs and supercapacitors, as well as associated power converters, sensors, and communication modules. In the embodiments of this invention, a 2MWh lithium iron phosphate battery pack is used as the energy storage system. Through a high-frequency sensing acquisition unit, the grid frequency change rate, battery current fluctuation value, battery temperature gradient value, and environmental meteorological disturbance rate (second-order difference of wind speed / light intensity change) are collected every second.

[0029] The operating parameters of energy storage systems at the same time stamp are arranged into a multidimensional tensor according to a preset dimension, where the dimension of the multidimensional vector can be... , This represents the length of the sampling time window.

[0030] By constructing the operating parameters of an energy storage system at the same timestamp into a multidimensional tensor according to preset dimensions, the system's state information at different spatial nodes and parameter dimensions can be comprehensively captured, demonstrating the collaborative correlation of multidimensional data. Subsequently, a sliding window mechanism is introduced to divide the multidimensional tensor into several time-continuous window tensors, which not only preserves temporal information but also enhances the dynamic adaptability of feature extraction. Next, a preset graph neural network is used to perform tensor fusion processing on the multiple window tensors, fully exploring the spatial structural relationships and nonlinear features between parameters, improving the global consistency and semantic relevance of feature representation. Compared with traditional linear processing methods, this fusion method has greater expressive power and modeling flexibility, effectively enhancing the robustness and generalization performance of feature tensors in subsequent modeling, compression, and control.

[0031] Furthermore, after constructing the multidimensional perturbation structure tensor, the decision to enter the transfer criterion mechanism—that is, to proceed to the construction of the neural synaptic stress-inducing network model and subsequent steps—can be determined based on the multidimensional perturbation structure tensor. This allows for dynamic judgment on whether the reconstruction behavior of the tensor channel structure is triggered.

[0032] Specifically, the triggering conditions for the tensor structure migration mechanism include at least one of the following: (1) The rate of change of the tensor principal axis exceeds the set threshold; (2) The magnitude of the continuous-time gradient change of the tensor global information entropy is greater than a set threshold; (3) Tensor in Time and The difference in the nuclear norm of the third-order tensor between time points exceeds a set proportion.

[0033] If any of the above conditions are met, the process of constructing a neural synaptic stress response network model and subsequent steps is triggered, thereby avoiding unnecessary computation and communication overhead.

[0034] The calculation process for the rate of change of the principal axis of the tensor includes: performing PCA principal component analysis on the current tensor and the tensor of the previous period respectively, and extracting the first principal component vector. If the angle between it and the direction of the principal component of the previous period satisfies: ; This indicates a sudden change in the direction of the tensor principal axis, necessitating preparation for constructing a neural synaptic stress response network model. To set a threshold.

[0035] The process of determining the change in the gradient of the tensor global information entropy includes: Construct the information entropy curve using the marginal probability distribution of the tensor expanded along the time axis. If two consecutive periods satisfy: ; If the system's tensor state exhibits a perturbation-type structural mutation, it is marked as structurally unstable and requires preparation for constructing a neural synaptic stress-induced structural network model.

[0036] in, This is the threshold for the acceleration of information entropy.

[0037] The process of change in the nuclear norm of the third-order tensor includes: Computation of the third-order HOSVD kernel tensor and The rate of change of the Frobenius norm: The specific formulas include: ; like This indicates an abnormal structural fluctuation, requiring preparation for constructing a neural synaptic stress response network model. The set kernel tensor sensitivity.

[0038] S102, Based on the multidimensional perturbation structure tensor, establish a neural synaptic stress structure network model, and based on the neural synaptic stress structure network model, obtain the synaptic response mapping matrix.

[0039] After obtaining the multidimensional perturbation structure tensor, a neural synaptic stress response structure network model can be established based on the multidimensional perturbation structure tensor. The synaptic response mapping matrix can be obtained by inputting the multidimensional perturbation structure tensor into the neural synaptic stress response structure network model.

[0040] Specifically, refer to Figure 2 The process of establishing a simulated neural synaptic stress structure network model based on the multidimensional perturbation structure tensor, and obtaining the synaptic response mapping matrix based on the simulated neural synaptic stress structure network model, may include: Obtain a preset basic model structure, wherein the basic model structure includes synapses and neurons; The multidimensional perturbation structure tensor is decomposed to obtain the core tensor and factor matrix; Based on the core tensor and factor matrix, tensor features are obtained; The tensor features are converted into the basic model structure parameters using a preset function; Based on the aforementioned basic model structure and basic model structure parameters, a neural synaptic stress response structure network model is established. The multidimensional perturbation structure tensor is input into the neural synaptic stress structure network model to obtain the synaptic response mapping matrix.

[0041] In the specific implementation process, a pre-defined basic model structure can be obtained first. This basic model structure includes multiple synapses and neurons connected to the synapses, including preneurons and postneurons. Then, the multidimensional perturbation structure tensor is subjected to Bucker decomposition to obtain the core tensor. Factor matrix Factor matrix Factor matrix sum factor matrix Among them, the core tensor Factor matrix represents the interaction relationships between dimensions. Represents perturbation characteristics, factor matrix Representative intensity features, factor matrix Representing response characteristics, factor matrix It represents the characteristics of time.

[0042] For each synapse in the basic model structure, tensor features are extracted based on the core tensor and factor matrix. These tensor features include perturbation feature vectors, response feature vectors, sensitivity vectors, and time and intensity features. The specific extraction process includes: for each synapse in the basic model structure, based on the preneuron index i, from... Extracting perturbation feature vectors Based on the post-neuron index j, from Extracting response feature vectors Based on the extracted perturbation feature vector and response feature vector, the sensitivity vector is calculated. The specific calculation formula is as follows: Sensitivity Vector: .from and Extract intensity and time features.

[0043] The perturbation feature vector, response feature vector, sensitivity vector, intensity, and time features are mapped to parameters of a predefined basic model structure using a mapping function. This mapping function can be anything from `map_theta0()` to `map_sigma_delta()`, etc. No specific restrictions are imposed.

[0044] By customizing the parameters of the basic model structure, the parameters of the basic model structure are adjusted, and the sensitivity vector is bound to the corresponding synapse, thereby establishing a neural synaptic stress response structure network model.

[0045] The multidimensional perturbation structure tensor is converted into a vector and input into a simulated neural synaptic stress structure network model. Each synaptic path in the simulated neural synaptic stress structure network model dynamically adjusts its excitation threshold based on the instantaneous perturbation amplitude, temperature gradient, and current fluctuation amplitude in the multidimensional perturbation structure vector, forming a synaptic response mapping matrix. This matrix is ​​then output through the simulated neural synaptic stress structure network model. The corresponding excitation response amplitude is output for each synaptic path. Disturbance peak Stimulus slope The path contribution can be obtained by processing the signal through a sigmoid activation function and then entering an in-log nonlinear combination function. This allows for the setting of a neural synaptic stress-inducing network model. There are synaptic paths, corresponding to 6 types of perturbation coupling structures.

[0046] S103, based on the synaptic response mapping matrix, channel excitation is performed on the current perturbation structure tensor to obtain the excitation response sequence.

[0047] In the specific implementation process, the current perturbation structure tensor is channel-excited through the synaptic response mapping matrix, where the current perturbation structure tensor is the perturbation structure tensor at the current moment.

[0048] The process of channel-exciting the current perturbation structure tensor based on the synaptic response mapping matrix to obtain the excitation response sequence specifically includes: The current perturbation structure tensor is input into the neural synaptic stress structure network model; The current perturbation structure tensor is projected onto the target neuron of the neural synaptic stress structure network model using the synaptic response mapping matrix. The target neuron is used to perform excitation calculations to obtain an excitation response sequence.

[0049] Specifically, the current perturbation structure tensor is input into the simulated neural synaptic stress structure network model. Through the computational unit of the simulated neural synaptic stress structure network model, the input space and channel information of the current perturbation structure tensor are aggregated onto the membrane potential of the target neuron of the simulated neural synaptic stress structure network model according to the weights and rules defined by the synaptic response mapping matrix. This simulates the presynaptic pulse passing through synapses of varying intensities and converging into the total current on the dendrites of the postsynaptic neuron.

[0050] The activation calculation is performed through the target neuron. Specifically, the target neuron spatiotemporally integrates the currents from all synaptic pathways, i.e., all connections defined by the mapping matrix, to calculate the total input current. This total input current is then input into the membrane potential dynamics equation of the target neuron to update the membrane potential. If the updated membrane potential exceeds a set threshold, the target neuron generates an output pulse (1) at that moment; otherwise, it remains at rest (0). For each significant connection, or all connections, defined in the synaptic response mapping matrix, the time point at which a pulse appeared from a certain channel / location of the input tensor is recorded, along with the postsynaptic current time waveform ◦ generated after the pulse passes through the weights and filtering of that specific synaptic pathway, and the decay of the current time waveform over time. The recorded information is then combined to obtain the activation response sequence for each synaptic pathway. The activation response sequence represents the activity level of the synapse at that moment.

[0051] By introducing a pseudo-synaptic stress mechanism, the strategy excitation threshold is dynamically adjusted based on the parameters of the current disturbance structure tensor, thereby establishing a synaptic response mapping matrix. This effectively enhances the selectivity of the control path under disturbance forcing conditions and breaks through the technical bottlenecks of existing control strategies' update lag and response rigidity.

[0052] S104, Based on the excitation response sequence and the current perturbation structure tensor, determine the strategy behavior tensor path.

[0053] Specifically, the main channel is identified by stimulating the response sequence, and the policy behavior tensor path is determined based on the acceleration trend of the main channel and the stability index of the current perturbation structure tensor.

[0054] Specifically, the process of determining the policy behavior tensor path based on the excitation response sequence and the current perturbation structure tensor may include: Based on the excitation response sequence, the main channel is identified; Calculate the acceleration trend of the main channel and the stability index of the current perturbation structure tensor; The strategy behavior tensor path is determined based on the acceleration of the main channel and the stability index of the current perturbation structure tensor.

[0055] In the specific implementation process, the main channel for information transmission is identified from the excitation-response sequence. The process of identifying the main channel may include: statistically analyzing the excitation frequency and intensity of each synaptic path corresponding to the excitation-response sequence, calculating the information entropy based on the excitation frequency and intensity, identifying the channel with the most concentrated information based on the information entropy, and determining the main channel by combining time correlation analysis.

[0056] The process involves calculating the acceleration trend of the main channel and the stability index of the current perturbation structure tensor. Based on these factors, the strategy behavior tensor path is determined. The calculation of the main channel's acceleration trend includes extracting historical time series data, preprocessing the extracted historical time series (including outlier detection and handling), and obtaining a preprocessed historical time series. Using the preprocessed historical time series, velocity and acceleration are calculated, an initial acceleration trend is extracted, and the acceleration trend is corrected using velocity data to obtain the final acceleration trend. The calculation of the stability index includes calculating the variance of each element of the perturbation tensor within a time window and aggregating the variances to obtain the stability index. Based on the correlation between the stability index, acceleration trend, and strategy, a specific strategy is determined. The corresponding strategy library is accessed to obtain the corresponding path. For example, if the stability index and acceleration trend indicate a discharge strategy, the discharge strategy library is accessed to determine the specific parameters, which are then converted into a strategy behavior tensor path.

[0057] S104, Based on the policy behavior tensor path and policy evaluation function, generate control commands and control the energy storage system to execute the control commands.

[0058] Specifically, the strategy evaluation function is first constructed. This function is a strategy evaluation function with aging evolution rate and perturbation structure strength as the primary optimization objectives. The policy evaluation function The expression can be specifically represented by the following formula: ; It should be noted that the range of this function is [0, +]. when A value close to 0 indicates that the perturbation evolution is stable, the SOH decreases very slowly, the synaptic response is stable, and the behavioral strategy selection is reasonable.

[0059] when When the value exceeds a certain threshold, for example, when it exceeds 10, it indicates that the SOH decreases more sharply, the disturbance fluctuations are more intense, the synaptic channels are more easily activated and disordered, the strategy behavior may be abnormally unstable, and immediate regulation is required.

[0060] in, This is the starting point for strategy evaluation. The two points together form the integration interval, representing the termination point of the strategy evaluation. This is the maximum fault tolerance coefficient of the energy storage system to sudden disturbances, representing the system's robustness in the face of disturbances; The exponential decay coefficient of SOH change with control intensity reflects the impact of health status on behavioral conservatism. The slope of the battery's SOH health over time, i.e., the current aging rate of the system; Let be the response trajectory of the perturbation principal component, representing the trend of the perturbation structure changing over time, and be the time evolution function of the perturbation path of the tensor principal axis. For the first The excitation response amplitude of a synaptic pathway, i.e., the synaptic response output, represents the intensity of the discharge response of a pathway in a neural structure.

[0061] For the first The amplitude of the perturbation peak of the policy behavior tensor path indicates the severity of the perturbation along that path. This is the synaptic excitation rate adjustment coefficient, which regulates the response amplification speed; For the first The instantaneous current fluctuation value of the strategy behavior tensor path is the stimulus intensity at the synaptic receiver. For the first The interference resonance coefficient of the path strategy behavior tensor is used to describe the interference synchronization coupling effect. The frequency nonlinear transformation adjustment parameter controls the sinusoidal modulation period within the response channel; For the first The amplitude of the temperature perturbation along the policy behavior tensor path represents the degree of influence of temperature on the policy response; This is a parameter for adjusting the sensitivity of tensor structure migration, used to control the flexibility of strategy switching. This serves as a stability index for the current perturbation structure tensor, reflecting whether the current perturbation behavior structure exhibits a tendency towards higher-order deformation. Specifically, to adjust the structural stability of the control strategy, this stability index is used in the embodiments. Defined as the standard deviation of the change in the tensor nuclear norm of the principal component tensor over the past 10 seconds. A value exceeding 0.2 indicates that the control strategy structure faces a risk of mutation, which will be addressed through... Adjust the final response strength to prevent strategy collapse.

[0062] A dynamically coupled integral optimization function is formed through a strategy evaluation function to assess the synaptic behavior strength, SOH risk, and tensor structure stability under the influence of disturbances over time, thereby controlling the system to... Minimizing the value is the objective of the strategy path selection and behavior switching, avoiding predictive errors and deep discharges that could damage the system.

[0063] The parameters mentioned above are set as follows: Disturbance index coefficient Take 0.35, synaptic activation factor The resonance modulation factor has a range of [0.8, 1.2]. Tensor transfer sensitivity in the range [0.5, 1.5]. Take 1.0, synaptic response amplitude The initial value range is [0.1, 1.5], and the current disturbance... The peak-to-peak values ​​of the dynamic current signal with a BMS sampling interval of 2Hz are taken, and all parameters are obtained through actual measurement and calibration.

[0064] Specifically, the process of generating control commands based on the policy behavior tensor path and policy evaluation function, and controlling the energy storage system to execute the control commands, includes: The parameters corresponding to the policy behavior tensor path are input into the policy evaluation function to obtain the policy evaluation value; Based on the strategy evaluation value, a control strategy is obtained, and based on the control strategy, control instructions are generated; Control the energy storage system to execute the control commands.

[0065] In practice, the parameters corresponding to the policy behavior tensor path, such as the tensor perturbation factor, synaptic response output, and stability index, are substituted into the policy evaluation function to obtain the policy evaluation value. Tensor perturbation factors include , , , , , as well as Based on the strategy evaluation value The system selects a corresponding control strategy, generates control commands based on the corresponding control strategy, and controls the energy storage system to execute the control commands.

[0066] Specifically, when The system selects a conventional charging strategy at that time; when At that time, choose a low-amplitude smooth charge / discharge behavior; when When the system enters a strategy freeze state, it switches to the backup path tensor and temporarily freezes the synaptic weight adaptive update process to ensure the safe operation of the system.

[0067] Alternatively, in obtaining the strategy evaluation value Subsequently, before selecting a control strategy, the obtained strategy evaluation value can be evaluated. conduct Function normalization processing, and based on the normalized result The value range is mapped to the corresponding behavior path execution scheme.

[0068] Specifically, the process of obtaining a control strategy based on the strategy evaluation value and generating control instructions based on the control strategy includes: The strategy evaluation value is normalized to obtain a normalized strategy evaluation value; The execution path is determined based on the normalized strategy evaluation value; Control instructions are generated based on the execution path.

[0069] Evaluation value of the obtained strategy conduct The function is normalized to obtain the normalized policy evaluation value, and the corresponding execution path scheme is determined based on the normalized policy evaluation value.

[0070] Among them, the normalized The correspondence between values ​​and behavior path execution schemes is as follows: When the value is below 0.3, the freeze behavior path is executed, which includes pausing the current tensor path switching and maintaining the strategy state of the previous cycle; When the value is between 0.3 and 0.7, the standard strategy path is executed, which includes selecting the optimal combination of behaviors based on the current response level of the tensor channel; When the value is higher than 0.7, a jump reconstruction path is performed, including rebuilding the tensor structure and enabling heterogeneous path weights to generate new policy candidate paths; In this embodiment, before the control strategy selection behavior is executed, the output result of the strategy evaluation function is normalized to compress the scoring interval and enhance the sensitivity of the behavior mapping within the numerical interval. The introduction of a piecewise normalized interval mapping method enables control path switching to have a clear threshold response interval, achieving differentiated selection of behavior paths under the strategy evaluation function, improving the scheduling effect of the energy storage system, and thus enhancing the ability to suppress the aging risk of the energy storage system.

[0071] The energy storage system scheduling method of the present invention can be executed by an edge controller, and every 10 minutes all tensor states, SOH evolution data, and synaptic weight matrices are uploaded to a cloud server for synaptic structure updates and policy network evolution optimization for the next cycle, so as to form a closed loop starting from tensors, neural responses, structural evaluation, and finally policy control, forming a long-term, stable structural memory that is not easily overwritten by new information.

[0072] Furthermore, after controlling the energy storage system to execute the control command, the strategy execution result is obtained and matched with the Ψ function value. If a mismatch occurs between the strategy result and the Ψ function value in multiple consecutive operating cycles, it is determined that the current behavior path structure is inconsistent with the disturbance response trend, and a path structure self-evolution process is triggered. Specifically, when the absolute deviation between the two exceeds 0.25 in three consecutive control cycles, and the SOC fluctuation amplitude of the energy storage system exceeds ±3% during the same period, a path freezing step is executed.

[0073] The path structure self-evolution process includes: Freeze the current policy behavior tensor path and enter policy replacement mode; Based on the rate of change of the main channel perturbation intensity of the perturbation structure tensor, the policy response delay of the previous cycle, and the trend of the SOH slope, a candidate policy path combination structure is constructed. After freezing, by querying the historical perturbation tensors and corresponding control behavior sequences of the last 100 times, the top 3 perturbation channels with the highest rate of change of derivative are extracted. Based on the extracted perturbation channels, a candidate policy path combination set is constructed, and each candidate policy path contains no less than 5 control policy behavior nodes.

[0074] From the candidate policy path combination structure, the behavioral path most relevant to the current Ψ function mapping trend is selected as the execution path for the next cycle, and this path replacement event is recorded on the edge control side for subsequent structure reconstruction and optimization. Specifically, the cosine similarity is calculated using the main channel vector of the current perturbation tensor and the vectors of each behavioral label in the candidate policy path to obtain a similarity score. The candidate policy paths are then scored based on the Ψ value response characteristics, and the path with the highest score is selected as the replacement path and the control behavior is executed.

[0075] After execution, record the change in disturbance intensity and the change trend of energy storage SOH corresponding to the path, and use the path structure as a new behavior node to update the tensor behavior structure diagram for subsequent control reference.

[0076] Meanwhile, the ratio of local variance to mean is used to identify the three most important channels in the current tensor perturbation, and the activation priority of subsequent behaviors is dynamically adjusted based on their scores.

[0077] In another embodiment, the number of times the direction of the first derivative of the energy storage SOH changes within 5 consecutive control cycles is statistically analyzed. If the number of direction switching exceeds 3 times, it can be determined that it has entered a degradation trend and trigger path freezing.

[0078] For the replaced path, the perturbation label and control results of the replaced path are compared with... The function response is uploaded to the simulation verification module, and the error is compared by calling the simulation model under the historical similar environment strategy. If the simulation error decreases by more than 10% compared with the historical strategy, the effectiveness of the alternative path is confirmed.

[0079] Based on the method provided in the above embodiments, the following steps are taken: First, the operating parameters of the energy storage system are obtained, and a multidimensional perturbation structure tensor is constructed based on these parameters. Then, a neural synaptic stress-inducing network model is established based on the multidimensional perturbation structure tensor. Finally, a synaptic response mapping matrix is ​​obtained based on the neural synaptic stress-inducing network model. Based on the synaptic response mapping matrix, channel excitation is performed on the current perturbation structure tensor to obtain an excitation response sequence. Based on the excitation response sequence and the current perturbation structure tensor, a strategy behavior tensor path is determined. Based on the strategy behavior tensor path and the strategy evaluation function, a control command is generated, and the energy storage system is controlled to execute the control command. The method provided by this invention allows for the establishment of a neural synaptic stress-inducing network model based on a multidimensional perturbation structure tensor, thereby obtaining a synaptic response mapping matrix. Based on the synaptic response mapping matrix, channel excitation is performed on the current perturbation structure tensor to obtain an excitation response sequence. Based on the excitation response sequence and the current perturbation structure tensor, a strategy behavior tensor path is determined, and control commands are generated based on the strategy behavior tensor path and the strategy evaluation function. This paper proposes a method to achieve multi-parameter fusion modeling of the operating state of an energy storage system using a multi-dimensional perturbation structure tensor. This significantly improves the time-series perception capability of the dynamic characteristics of the energy storage system and overcomes the problem of insufficient ability to express the coupling relationship of heterogeneous parameters in traditional methods. By exciting channels based on the synaptic response mapping matrix, the selection elasticity of tensor paths under perturbation forcing conditions is enhanced, breaking through the technical bottlenecks of existing control strategies' update lag and response rigidity. By introducing a strategy evaluation function and determining control commands based on the strategy evaluation function and the strategy behavior tensor path, differentiated selection of behavior paths under the strategy evaluation function is achieved, thereby improving the scheduling effect of the energy storage system.

[0080] Based on the above method embodiments, this invention also provides an energy storage system scheduling device. See also... Figure 3 The diagram shown is a schematic of an energy storage system scheduling device provided in an embodiment of the present invention.

[0081] The device 300 includes: The acquisition module 101 is used to acquire the operating parameters of the energy storage system and construct a multidimensional perturbation structure tensor based on the operating parameters. Module 102 is used to establish a simulated neural synaptic stress structure network model based on the multidimensional perturbation structure tensor, and to obtain a synaptic response mapping matrix based on the simulated neural synaptic stress structure network model. Channel excitation module 103 is used to excite the current perturbation structure tensor through channels based on the synaptic response mapping matrix to obtain an excitation response sequence; The determination module 104 is used to determine the strategy behavior tensor path based on the excitation response sequence and the current perturbation structure tensor; The generation module 105 is used to generate control commands based on the policy behavior tensor path and policy evaluation function, and to control the energy storage system to execute the control commands.

[0082] In one possible implementation, the step of establishing a simulated neural synaptic stress structure network model based on the multidimensional perturbation structure tensor, and obtaining a synaptic response mapping matrix based on the simulated neural synaptic stress structure network model, includes: Obtain a preset basic model structure, wherein the basic model structure includes synapses and neurons; The multidimensional perturbation structure tensor is decomposed to obtain the core tensor and factor matrix; Based on the core tensor and factor matrix, tensor features are obtained; The tensor features are converted into the basic model structure parameters using a preset function; Based on the aforementioned basic model structure and basic model structure parameters, a neural synaptic stress response structure network model is established. The multidimensional perturbation structure tensor is input into the neural synaptic stress structure network model to obtain the synaptic response mapping matrix.

[0083] In one possible implementation, the step of channel-exciting the current perturbed structure tensor based on the synaptic response mapping matrix to obtain an excitation response sequence includes: The current perturbation structure tensor is input into the neural synaptic stress structure network model; The current perturbation structure tensor is projected onto the target neuron of the neural synaptic stress structure network model using the synaptic response mapping matrix. The target neuron is used to perform excitation calculations to obtain an excitation response sequence.

[0084] In one possible implementation, determining the policy behavior tensor path based on the excitation response sequence and the current perturbation structure tensor includes: Based on the excitation response sequence, the main channel is identified; Calculate the acceleration trend of the main channel and the stability index of the current perturbation structure tensor; The strategy behavior tensor path is determined based on the acceleration of the main channel and the stability index of the current perturbation structure tensor.

[0085] In one possible implementation, generating control commands based on the policy behavior tensor path and the policy evaluation function, and controlling the energy storage system to execute the control commands, includes: The parameters corresponding to the policy behavior tensor path are input into the policy evaluation function to obtain the policy evaluation value; Based on the strategy evaluation value, a control strategy is obtained, and based on the control strategy, control instructions are generated; Control the energy storage system to execute the control commands.

[0086] In one possible implementation, obtaining a control policy based on the policy evaluation value and generating control instructions based on the control policy includes: The strategy evaluation value is normalized to obtain a normalized strategy evaluation value; The execution path is determined based on the normalized strategy evaluation value; Control instructions are generated based on the execution path.

[0087] In one possible implementation, obtaining the operating parameters of the energy storage system and constructing a multidimensional perturbation structure tensor based on the operating parameters includes: The operating parameters of the energy storage system are collected through the sensing and acquisition unit; The running parameters under the same timestamp are arranged into a multidimensional tensor according to a preset dimension; The multidimensional tensor is divided into N window tensors using a sliding window mechanism; A pre-defined graph neural network is used to fuse N window tensors to obtain a multidimensional perturbation structure tensor, where N is a positive integer greater than 1.

[0088] See Figure 4 , Figure 4 This is a schematic diagram of an electronic device provided in an embodiment of this application.

[0089] The device 400 includes a memory 401 and a processor 402; the memory 401 is used to store relevant program code; the processor 402 is used to call the program code to execute the energy storage system scheduling method described in the above method embodiment.

[0090] Furthermore, embodiments of the present invention also provide an energy storage system, in which the aforementioned electronic device is provided, which is used to execute various steps in the aforementioned energy storage system scheduling method. Furthermore, embodiments of the present invention also provide a computer-readable storage medium for storing a computer program for executing the energy storage system scheduling method described in the above method embodiments.

[0091] This invention also provides a computer program product, which includes a computer program / instruction. When the computer program / instruction is executed by a processor, it implements the energy storage system scheduling method described in the above method embodiments.

[0092] It should be noted that the computer-readable medium described above in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0093] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of the present invention. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0094] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. In particular, for system or device embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The device embodiments described above are merely illustrative. The units or modules described as separate components may or may not be physically separate. The components shown as units or modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the units or modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0095] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functions, and operations that may be implemented according to various embodiments of the invention, including methods, apparatus, and devices. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0096] It should be understood that in this invention, "at least one (item)" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0097] It should also be noted that, in this invention, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0098] The steps of the methods or algorithms described in conjunction with the embodiments disclosed in this invention can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for scheduling an energy storage system, characterized in that, The method includes: Obtain the operating parameters of the energy storage system, and construct a multidimensional perturbation structure tensor based on the operating parameters; Based on the multidimensional perturbation structure tensor, a neural synaptic stress response structure network model is established, and based on the neural synaptic stress response structure network model, a synaptic response mapping matrix is ​​obtained. Based on the synaptic response mapping matrix, the current perturbation structure tensor is channel-excited to obtain an excitation response sequence; Based on the excitation response sequence and the current perturbation structure tensor, determine the policy behavior tensor path; Based on the policy behavior tensor path and policy evaluation function, control commands are generated, and the energy storage system is controlled to execute the control commands.

2. The method according to claim 1, characterized in that, The step of establishing a simulated neural synaptic stress structure network model based on the multidimensional perturbation structure tensor, and obtaining the synaptic response mapping matrix based on the simulated neural synaptic stress structure network model, includes: Obtain a preset basic model structure, wherein the basic model structure includes synapses and neurons; The multidimensional perturbation structure tensor is decomposed to obtain the core tensor and factor matrix; Based on the core tensor and factor matrix, tensor features are obtained; The tensor features are converted into the basic model structure parameters using a preset function; Based on the aforementioned basic model structure and basic model structure parameters, a neural synaptic stress response structure network model is established. The multidimensional perturbation structure tensor is input into the neural synaptic stress structure network model to obtain the synaptic response mapping matrix.

3. The method according to claim 1, characterized in that, The step of channel-exciting the current perturbed structure tensor based on the synaptic response mapping matrix to obtain the excitation response sequence includes: The current perturbation structure tensor is input into the neural synaptic stress structure network model; The current perturbation structure tensor is projected onto the target neuron of the neural synaptic stress structure network model using the synaptic response mapping matrix. The target neuron is used to perform excitation calculations to obtain an excitation response sequence.

4. The method according to claim 1, characterized in that, The step of determining the policy behavior tensor path based on the excitation response sequence and the current perturbation structure tensor includes: Based on the excitation response sequence, the main channel is identified; Calculate the acceleration trend of the main channel and the stability index of the current perturbation structure tensor; The strategy behavior tensor path is determined based on the acceleration of the main channel and the stability index of the current perturbation structure tensor.

5. The method according to claim 1, characterized in that, The step of generating control commands based on the policy behavior tensor path and policy evaluation function, and controlling the energy storage system to execute the control commands, includes: The parameters corresponding to the policy behavior tensor path are input into the policy evaluation function to obtain the policy evaluation value; Based on the strategy evaluation value, a control strategy is obtained, and based on the control strategy, control instructions are generated; Control the energy storage system to execute the control commands.

6. The method according to claim 5, characterized in that, The step of obtaining a control strategy based on the strategy evaluation value and generating control instructions based on the control strategy includes: The strategy evaluation value is normalized to obtain a normalized strategy evaluation value; The execution path is determined based on the normalized strategy evaluation value; Control instructions are generated based on the execution path.

7. The method according to claim 1, characterized in that, The process of acquiring the operating parameters of the energy storage system and constructing a multidimensional perturbation structure tensor based on the operating parameters includes: The operating parameters of the energy storage system are collected through the sensing and acquisition unit; The running parameters under the same timestamp are arranged into a multidimensional tensor according to a preset dimension; The multidimensional tensor is divided into N window tensors using a sliding window mechanism; A pre-defined graph neural network is used to fuse N window tensors to obtain a multidimensional perturbation structure tensor, where N is a positive integer greater than 1.

8. An energy storage system dispatching device, characterized in that, The device includes: The acquisition module is used to acquire the operating parameters of the energy storage system and construct a multidimensional perturbation structure tensor based on the operating parameters. A module is established to build a simulated neural synaptic stress structure network model based on the multidimensional perturbation structure tensor, and to obtain a synaptic response mapping matrix based on the simulated neural synaptic stress structure network model. The channel excitation module is used to excite the current perturbation structure tensor through the channel based on the synaptic response mapping matrix to obtain an excitation response sequence. The determination module is used to determine the policy behavior tensor path based on the excitation response sequence and the current perturbation structure tensor; The generation module is used to generate control commands based on the policy behavior tensor path and policy evaluation function, and to control the energy storage system to execute the control commands.

9. An electronic device, characterized in that, The device includes: a memory and a processor; the memory is used to store relevant program code; the processor is used to call the program code to execute the energy storage system scheduling method according to any one of claims 1 to 7.

10. An energy storage system, characterized in that, The energy storage system is equipped with the electronic device as described in claim 9.

Images