Optical storage and charging integrated station energy storage system

Abstract

The invention relates to the technical field of electric digital data processing and energy storage system safety control, in particular to an optical storage and charging integrated station energy storage system. Comprising a data acquisition module which is used for acquiring a state feature vector of an energy storage system, planned scheduling power of an upper-layer energy management system, an external power grid interference index, and a real-time safety state and an economic value index of a battery management system in real time; the risk assessment module is used for generating a predictive risk score; the decision threshold module is used for calculating a dynamic intervention threshold; the arbitration decision module is used for generating a negative sign; the instruction correction module is used for determining a derating factor in response to the override mark and generating a correction scheduling instruction; or when the negative sign is not generated, outputting the planned scheduling power as a corrected scheduling instruction. According to the invention, early warning and active prevention of the potential thermal runaway risk are realized, and the intrinsic safety level and the operation reliability of the system are obviously improved.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing and energy storage system safety control technology, specifically to an integrated photovoltaic-storage-charging site energy storage system. Background Technology

[0002] With the development of integrated photovoltaic, energy storage and charging stations, the role of energy storage systems in energy dispatch is becoming increasingly important; their operation is subject to the dual constraints of the economic dispatch instructions of the upper-level energy management system and the physical safety protection of the battery management system.

[0003] Currently, the system's security mainly relies on the BMS; the BMS provides real-time physical protection by monitoring physical sensor data such as real-time voltage and real-time temperature; at the same time, the EMS tends to issue high-power dispatch commands based on market economic value indicators in order to pursue short-term economic benefits.

[0004] However, traditional methods relying on physical sensors have significant data lag. Physical sensors cannot reflect the electrochemical instability caused by accumulated micro-damage inside the battery in a timely manner, which may lead to the system entering a hidden danger state where the physical characterization is normal, but the internal state is already in a high-risk steady state. If the system encounters a strong economic dispatch command issued by EMS or extreme interference from the external power grid at this time, the lagging BMS physical protection mechanism will not be able to respond in time, putting the battery at risk of breaking through the thermal runaway failure boundary.

[0005] Therefore, how to resolve the decision-making conflict between the economic requirements of EMS and the lagging physical protection of BMS, and how to proactively identify and quantify the predictive risks caused by internal micro-damage before physical alarms occur, are technical problems that urgently need to be solved in this field. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention discloses an integrated photovoltaic-storage-charging site energy storage system. Specifically, the technical solution of this invention includes:

[0007] The data acquisition module is used to acquire in real time the state feature vector of the energy storage system, the planned scheduling power of the upper-level energy management system, the external power grid interference index, the real-time safety status of the battery management system, and the economic value index.

[0008] The risk assessment module is used to calculate the electrochemical instability index based on the state feature vector through a preset digital twin model; and to calculate the dynamic stress factor based on the planned dispatch power and grid interference index; and then to generate a predictive risk score by combining the electrochemical instability index and the dynamic stress factor.

[0009] a decision threshold module configured to determine a static security baseline based on a preset thermal runaway failure boundary and a basic safety margin, and calculate a dynamic intervention threshold in combination with an economic value index;

[0010] an arbitration decision module configured to generate a veto flag when the predictive risk score is greater than the dynamic intervention threshold and the real-time safety state is normal;

[0011] an instruction correction module configured to determine a derating factor in response to the veto flag, and correct the planned dispatch power based on the derating factor to generate a corrected dispatch instruction; or output the planned dispatch power as the corrected dispatch instruction when the veto flag is not generated.

[0012] Preferably, the state feature vector comprises real-time voltage, real-time temperature, cumulative cycle number, and preset manufacturing defect level.

[0013] Preferably, the risk assessment module is configured to calculate an electrochemical instability index via a preset digital twin model, comprising:

[0014] inputting the state feature vector into the preset digital twin model;

[0015] calculating the electrochemical instability index representing internal vulnerability by the digital twin model.

[0016] Preferably, the risk assessment module is configured to calculate a dynamic stress factor, comprising:

[0017] determining a power stress based on a ratio of the planned dispatch power to a preset rated power;

[0018] determining a grid disturbance stress based on a grid disturbance index;

[0019] weighting and summing the power stress and the grid disturbance stress to obtain the dynamic stress factor.

[0020] Preferably, the risk assessment module is configured to generate a predictive risk score, comprising:

[0021] calculating an external stress amplification coefficient based on the dynamic stress factor;

[0022] multiplying the electrochemical instability index and the external stress amplification coefficient to obtain the predictive risk score.

[0023] Preferably, the decision threshold module is configured to calculate the dynamic intervention threshold, comprising:

[0024] determining the dynamic intervention threshold based on the static security baseline, a preset economic hedging coefficient, and the economic value index.

[0025] Preferably, the arbitration decision module is configured to generate the veto flag, comprising:

[0026] determining whether the predictive risk score is greater than the dynamic intervention threshold value;

[0027] determining whether the real-time safety state is normal;

[0028] when the predictive risk score is greater than the dynamic intervention threshold value and the real-time safety state is normal, determining that the veto flag is an execution veto;

[0029] when the predictive risk score is less than or equal to the dynamic intervention threshold value or the real-time safety state is abnormal, determining that the veto flag is a non-execution veto.

[0030] Compared with the prior art, the present application has the following beneficial effects:

[0031] 1、The present application decouples and quantifies the electrochemical instability index inside the energy storage system by constructing a digital twin model; at the same time, it generates a forward-looking predictive risk score by combining the dynamic stress caused by grid disturbance and dispatching plan; this method surpasses the real-time protection of traditional BMS, realizes early warning and active prevention of potential thermal runaway risk, and significantly improves the intrinsic safety level and operation reliability of the system.

[0032] 2、The present application innovatively sets a dynamic intervention threshold value. This threshold value is not only based on the static baseline determined by the thermal runaway failure boundary and safety margin, but also introduces economic value indicators and hedging coefficients for dynamic adjustment; this enables the system to flexibly balance operating efficiency and risk margin under the premise of ensuring safety, avoiding one-size-fits-all overprotection, and realizing the optimization decision of the energy storage system between safety and economy.

[0033] 3、The present application establishes a closed-loop management mechanism for arbitration decision and instruction revision; by comparing the predictive risk score with the dynamic intervention threshold value in real time, the system can generate a veto flag in advance when the real-time safety state is normal, and actively intervene in the upper-level dispatching. This design realizes accurate veto of high-risk planned power, prevents the accumulation and evolution of potential risks, and ensures that the dispatching instruction is always executed within the safety boundary.

[0034] 4、When executing instruction revision, the present application adopts an accurate derating strategy based on risk overshoot; when the veto flag is triggered, the system does not simply suspend operation, but calculates a derating factor according to the specific magnitude of the predicted risk exceeding the dynamic threshold value, combined with the derating response coefficient. This fine-tuned power revision method realizes proportional adjustment of risk and maximizes the impact of safety intervention on normal operation of the system. BRIEF DESCRIPTION OF DRAWINGS

[0035] The present application will be further explained in conjunction with the drawings and examples:

[0036] Figure 1 This is a system structure diagram of the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0038] Example 1

[0039] like Figure 1 As shown, this embodiment of the invention discloses an integrated photovoltaic-storage-charging site energy storage system, comprising:

[0040] The data acquisition module is used to acquire in real time the state feature vector of the energy storage system, the planned scheduling power of the upper-level energy management system, the external power grid interference index, the real-time safety status of the battery management system, and the economic value index.

[0041] The risk assessment module is used to calculate the electrochemical instability index based on the state feature vector through a preset digital twin model; and to calculate the dynamic stress factor based on the planned dispatch power and grid interference index; and then to generate a predictive risk score by combining the electrochemical instability index and the dynamic stress factor.

[0042] The decision threshold module is used to determine the static safety baseline based on the preset thermal runaway failure boundary and basic safety margin, and to calculate the dynamic intervention threshold in combination with economic value indicators.

[0043] The arbitration decision module is used to generate a veto flag when the predictive risk score is greater than the dynamic intervention threshold and the real-time safety status is normal.

[0044] The instruction correction module is used to respond to the veto flag, determine the derating factor, and correct the planned scheduling power based on the derating factor to generate a corrected scheduling instruction; or, if no veto flag is generated, output the planned scheduling power as a corrected scheduling instruction.

[0045] The integrated photovoltaic-storage-charging site energy storage system of this embodiment includes a data acquisition module, a risk assessment module, a decision threshold module, an arbitration decision module, and an instruction correction module. The modules work together to form a closed-loop predictive risk hedging control architecture.

[0046] The data acquisition module collects multidimensional heterogeneous state data necessary for predictive risk hedging control and forms the sensing foundation of the entire system. In this embodiment, this module is used to acquire the state feature vector of the BESS energy storage system in real time. The planned and dispatched power of the upper-level energy management system (EMS) External power grid related power grid interference indicators , battery management system BMS based on real-time safety state of physical sensor and market-related economic value indicators ;

[0047] state feature vector At the moment, the feature set representing the physical and historical degradation state of the BESS; as the input of the digital twin model for estimating internal micro-damage; physical sensor data collected in real time and preset BESS archive data;

[0048] real-time safety state BMS based on real-time voltage , real-time temperature and other physical sensor data to determine the current safety status; as the physical basis of the arbitration decision module, it is used to define the exercise boundary of the AI veto right; real-time polling BMS to obtain;

[0049] This module not only collects traditional physical sensor data, but also integrates preset manufacturing defect levels and external market and power grid data, solving the technical problems of sensor data isolation and inability to comprehensively consider the inherent characteristics of the battery and environmental stress in the prior art, providing multi-dimensional input for subsequent predictive risk assessment;

[0050] Risk assessment module, combining the internal vulnerability of BESS and external transient stress, generating a single predictive decision indicator for quantifying the safety distance of thermal runaway ;

[0051] This module realizes the calculation of the electrochemical instability index : Based on the state feature vector collected in real time, the electrochemical instability index is calculated through the preset digital twin model ; the electrochemical instability index is a dimensionless prediction indicator representing the current internal cumulative micro-damage and vulnerability of the BESS; solving the technical problem that physical sensor data is lagging and cannot reflect the electrochemical high-risk steady state, i.e. internal micro-damage has been accumulated; output of the digital twin model ;

[0052] This module realizes the calculation of the dynamic stress factor : Based on the planned dispatch power issued by EMS and the monitored power grid disturbance index , the dynamic stress factor is calculated; the dynamic stress factor A dimensionless index quantifying the instantaneous stress degree of BESS caused by current scheduling instructions and grid disturbances; provides a basis for quantifying the composite stress of BESS caused by composite grid extreme events; multi-factor stress superposition model The calculation result of the multi-factor stress superposition model

[0053] The module realizes the generation of the predictive risk score The module combines the electrochemical instability index With the dynamic stress factor To generate the final predictive risk score ; The predictive risk score Quantifies the comprehensive index of the safety distance from the thermal runaway failure boundary Build a single decision index that integrates the electrochemical instability index And external stress Solve the decision-making conflict between predictive risk and real-time rules; risk exposure assessment model The calculation result of the risk exposure assessment model

[0054] The decision threshold module determines the dynamic triggering threshold of AI predictive veto according to the preset engineering safety baseline and real-time economic considerations;

[0055] The module is used to determine the static safety baseline Based on the preset dimensionless thermal runaway failure boundary And the basic safety margin ; The static safety baseline Is the default safety judgment threshold of the system without considering real-time economic benefits; the standard safety margin is defined as: ;

[0056] The module combines the normalized economic value index , Economic hedging coefficient Calculate the dynamic intervention threshold ; The dynamic intervention threshold Consider the real-time economic benefits and operational risk preferences of AI predictive veto; solve the decision-making authorization conflict, and realize the hedging of predictive risk through dynamic adjustment of the threshold; dynamic adjustment of the decision threshold model ;

[0057] The design of the dynamically adjusted threshold Introduces the economic value index , So that the threshold Is inversely related to economic benefits; when economic benefits Increase, This reduces risk sensitivity, thereby proactively increasing security sensitivity and making it easier to trigger hedging. This reflects the innovation of this system in strengthening risk management through strategies while pursuing economic benefits.

[0058] Arbitration decision-making module, based on AI-predicted future risks The current normal operation is consistent with the actual BMS test. Arbitration will be conducted between the parties to determine whether AI has the right to veto the EMS instructions;

[0059] This module is used for predictive risk scoring. Greater than the dynamic intervention threshold Furthermore, the real-time safety status of the battery management system. When normal, a veto flag is generated. ; veto symbol It is a binary flag indicating whether this architecture has obtained the predictive veto right to EMS instructions; it serves as the core output of this module for resolving data trust conflicts; and it represents the judgment result of the preset authorization veto logic.

[0060] When the predictive risk score is less than or equal to the dynamic intervention threshold, or when the real-time safety status is abnormal, a flag indicating that the veto has not been executed is generated.

[0061] The core lies in This is the condition of normal operation; it clarifies that the veto power of the AI ​​architecture only applies to high-risk steady states where the physical BMS has not yet detected danger; once the physical sensor has triggered an alarm, the system will rely on the highest priority physical protection of the BMS, and the AI ​​will automatically relinquish the highest control; this logic resolves the conflict of permissions between physical measurement and AI prediction, and ensures a safe and smooth switch between predictive intervention and physical protection.

[0062] The instruction correction module, based on the output of the arbitration decision module, adjusts the original planned scheduling power of the EMS. The system can be modified or output directly to generate the final risk-hedging scheduling instruction for execution. ;

[0063] When no veto flag is generated, this module follows EMS instructions and schedules the power accordingly. As a corrected scheduling instruction Output directly;

[0064] In response to the veto flag, this module determines the deduction factor. ; reduction factor A dimensionless multiplier used to actively reduce short-term performance in exchange for BESS health recovery; converting risk overshoot into a residual proportional gain to power; a variation of the proportional element in proportional-integral-derivative control. ;

[0065] This module is based on the depreciation factor. Correcting the scheduled power To generate corrected scheduling instructions ,Right now ;

[0066] This invention's integrated photovoltaic-storage-charging site energy storage system proactively intervenes and rejects dispatch commands from the upper-level energy management system (EMS) prior to battery physical alarms; it achieves this by incorporating electrochemical prediction... With dynamic stress Construct a predictive risk score And compare the score with a dynamically adjusted threshold. Arbitration is conducted; this predictive veto mechanism based on this model solves the core decision-making dilemma between AI prediction and physical measurement, successfully preventing the battery from exceeding the thermal runaway failure boundary when encountering extreme events under high-risk steady-state conditions. This extends the lifespan of BESS and enhances the inherent security of the system.

[0067] Example 2

[0068] The state feature vector includes: real-time voltage, real-time temperature, cumulative cycle count, and preset manufacturing defect level;

[0069] This embodiment further specifies the state feature vector based on embodiment 1. Composition; State feature vector It explicitly includes four key parameters: real-time voltage Real-time temperature Cumulative number of loops and preset manufacturing defect levels ;

[0070] , This is real-time physical sensor data, representing the current macroscopic operating state of the battery; Historical degradation data represents the visible lifespan loss of the battery. The preset archive data represents inherent manufacturing defects in the battery that cannot be detected by real-time sensors; for example, its value range can be normalized to 0 (no defect) to 1 (critical defect).

[0071] By , Real-time macro data and , Historical micro / inherent data together constitute the state feature vector This system can comprehensively characterize the inherent properties of batteries; this provides a basis for digital twin models. Provided analysis , with , a nonlinear correlation between the basis, so as to more accurately estimate the cumulative micro-damage that the internal sensor cannot perceive, and improve the prediction accuracy of the electrochemical instability index .

[0072] Embodiment 3

[0073] The risk assessment module is configured to calculate the electrochemical instability index via a preset digital twin model, comprising:

[0074] inputting the state feature vector into the preset digital twin model;

[0075] calculating the electrochemical instability index representing the internal vulnerability by the digital twin model;

[0076] This embodiment is based on embodiment 2, and specifically describes the calculation process of the electrochemical instability index in the risk assessment module; the process aims to generate a predictive risk indicator beyond physical sensors through digital data processing ;

[0077] inputting the real-time collected state feature vector into the preset digital twin model ; the digital twin model is a pre-trained model based on electronic digital data processing; specifically, the model can be implemented by a three-layer feedforward neural network, which includes an input layer matching the state feature vector , containing four features, a hidden layer containing 16 nodes, and an output layer outputting a scalar ; the model is trained offline by a backpropagation algorithm, using mean square error as a loss function to minimize the difference between the model output and the actual micro-damage indicator measured by destructive physical detection in offline experiments; the model is trained based on a large amount of offline experimental data simulating implicit manufacturing defects and accelerated micro-degradation;

[0078] the electrochemical instability index representing the current internal vulnerability of the BESS is calculated by the digital twin model through analyzing the nonlinear correlation between the parameters in ;

[0079] ;

[0080] for electrochemical instability index at time t, normalized scalar, value range , dimensionless; is a pre-trained digital twin model, whose internal parameters such as neural network weights are fixed through offline training; is state feature vector at time t, whose parameters are derived from real-time acquisition;

[0081] The technical motivation of this formula is to solve the problem that physical sensors such as voltage and temperature data are lagging and cannot reflect the electrochemical high-risk steady state; by highly correlating the dimensionless output of with the micro-damage results of destructive physical detection DPA in the laboratory, the quantitative prediction of internal cumulative micro-damage is realized;

[0082] This embodiment effectively converts multi-dimensional and heterogeneous input into a single, quantifiable, and physically destructive prediction index through the digital data processing capability of the digital twin model ; this solves the information gap between the current normal and the future collapse, and provides a leading core prediction index for subsequent risk quantification that is ahead of physical alarms.

[0083] Embodiment 4

[0084] The risk assessment module is used to calculate the dynamic stress factor, including:

[0085] Determine the power stress based on the ratio of the planned dispatch power to the preset rated power;

[0086] Determine the grid disturbance stress based on the grid disturbance index;

[0087] Weighted sum of power stress and grid disturbance stress to get dynamic stress factor;

[0088] This embodiment describes the calculation process of the dynamic stress factor in the risk assessment module based on embodiment 1; this process quantifies the degree of composite instantaneous stress of the external environment on the BESS through a multi-factor stress superposition model;

[0089] Determine the power stress based on the ratio of the planned dispatch power to the preset rated power , that is ;

[0090] Determine the grid disturbance stress based on the normalized grid disturbance index obtained through external grid state monitoring, that is ;

[0091] The dynamic stress factor is obtained by weighting and summing the power stress and the grid interference stress. ;

[0092] ;

[0093] This is a dynamic stress factor, dimensionless (1), and its structure is ensured by the formula. ; The power stress weight is a preset engineering safety factor calibrated based on the stress sensitivity test data of BESS battery cells, with a dimension of 1. Its calibration method, for example, involves maintaining grid interference in the experimental environment. Applying to BESS cells Groups of different constant power Load, measurement Steady-state stress response of BESS under constant power load conditions Using the least squares method Perform fitting to solve ; EMS planned power, in watts (W), sourced from real-time data collection; This represents the steady-state stress response of BESS, used as the response variable to be fitted during the calibration process;

[0094] The rated power of BESS is in watts (W) and is derived from the default equipment file. The grid interference weight is also calibrated based on the stress sensitivity experimental data of BESS cells, and has a dimension of 1; its calibration method is similar, maintaining... and apply Groups of different Interference, measurement of the first Steady-state stress response of BESS under grid disturbance conditions Using the least squares method Perform fitting to solve ; A normalized power grid interference index, for example, could be the absolute value of the power grid frequency deviation. With standard frequency ratio It is either a weighted combination of multiple interfering factors, with a dimension of 1, and is derived from real-time data collection.

[0095] The technical motivation behind this formula is to quantify the combined stresses caused by complex power grid extreme events on BESS; through the analysis of... Take the absolute value to ensure that any significant disturbance to the power grid in any direction, such as excessively high or low frequency, is considered a positive stress, so that... Accurately reflect the composite stress;

[0096] The embodiment realizes the power scheduling and the power grid environment The two main external stresses are quantified and superimposed; this overcomes the defect of relying on a single index to evaluate external stress, so that The instantaneous risk exposure of BESS in pursuing benefits and facing extreme events can be accurately reflected under the dual pressure;

[0097] It should be noted that the weighted summation model adopted in the embodiment is an approximation adopted for the purpose of simplifying calculation, which effectively quantifies the two main stresses; in other embodiments, a nonlinear model such as a product model can also be used to better represent the coupling amplification effect between the two stresses.

[0098] Embodiment 5

[0099] The risk assessment module is configured to generate a predictive risk score, comprising:

[0100] Based on the dynamic stress factor, calculate the external stress amplification coefficient;

[0101] The electrochemical instability index is multiplied by the external stress amplification coefficient to obtain the predictive risk score;

[0102] The embodiment describes the generation process of the final predictive risk score of the risk assessment module based on embodiment 4; this process realizes the fusion of internal vulnerability and external stress;

[0103] Based on the dynamic stress factor , calculate the external stress amplification coefficient, that is ;

[0104] The electrochemical instability index is multiplied by the external stress amplification coefficient to obtain the final predictive risk score ;

[0105] ;

[0106] The predictive risk score has a dimension of 1; The instability index is as described above, and has a dimension of 1; The dynamic stress factor is also as described above, and has a dimension of 1;

[0107] The derivation logic of the formula is: risk equals internal vulnerability Multiplication of external stress amplification coefficient ; in the form of , it can ensure that the risk equals the basic instability when there is no external stress ; As an amplification coefficient, it makes the change of show a multiplicative effect;

[0108] By taking the internal predictive index as the base and the external real-time stress as the multiplication amplification coefficient, the embodiment realizes the nonlinear product fusion of two independent risk dimensions; this multiplicative effect can accurately reflect that when the system is in a high-risk steady state, i.e. high and encounters an extreme event, i.e. high, the risk will be multiplied and amplified, rapidly approaching the thermal runaway failure boundary ; this provides a risk quantification index with more unpredictable synergistic effects than a simple weighted model.

[0109] Embodiment 6

[0110] A decision threshold module for calculating a dynamic intervention threshold, comprising:

[0111] Based on the static safety baseline, the preset economic hedging coefficient, and the economic value index, the dynamic intervention threshold is determined;

[0112] This embodiment describes the calculation process of the dynamic intervention threshold in the decision threshold module based on embodiment 1; this process aims to realize predictive risk hedging;

[0113] According to the preset global constant, normalize the thermal runaway failure boundary and the basic safety margin with a dimension of 1 to determine the static safety baseline ; wherein and are global safety constants preset according to the safety test data of BESS; for example, is determined by conducting multiple destructive thermal runaway experiments on BESS samples, combining the model and risk assessment model of the present application, and statistically determining the critical predictive risk score value that actually causes thermal runaway, and taking the lower limit of the 99% confidence interval; the basic safety margin is a fixed margin set according to the operation safety standard or industry specification, for example, taking ;

[0114] ;

[0115] Based on the above static security baseline Preset economic hedging coefficient and real-time acquired normalized economic value indicators Determine the dynamic intervention threshold ;

[0116] ;

[0117] for The dynamic intervention threshold at time t, with a dimension of 1, is used for... Compare; For static safety baseline; The economic hedging coefficient is a core strategy parameter determined by the operator's risk appetite, with a dimension of 1. A normalized economic value indicator, for example, by using the current real-time electricity price. With average daily electricity price Comparison, such as The data obtained is derived from real-time data from the power grid market and has a dimension of 1. It is a preset minimum security threshold, for example This is to ensure that even under extremely high economic benefits, the safety threshold will not drop to an unreasonable level;

[0118] This formula is passed Item, realizing the Negative dynamic adjustment; when economic benefits When height increases, The value will decrease, making it lower than ;

[0119] This embodiment introduces economic value. The associated dynamic adjustment mechanism enables adaptive risk decision-making thresholds; when the site is in a high-profit power grid market environment, that is... high, It will proactively lower the sensitivity of the system, making it more sensitive to security and easier to trigger hedging. This ensures that the security defenses are dynamically strengthened when economic orders impose high loads on BESS without sacrificing the basic security margin.

[0120] Example 7

[0121] The arbitration decision module, used to generate veto flags, includes:

[0122] Determine whether the predictive risk score is greater than the dynamic intervention threshold;

[0123] determining whether the real-time safety state is normal;

[0124] when the predictive risk score is greater than the dynamic intervention threshold and the real-time safety state is normal, determining that the override flag is an executed override;

[0125] when the predictive risk score is less than or equal to the dynamic intervention threshold or the real-time safety state is abnormal, determining that the override flag is an unexecuted override;

[0126] The embodiment based on embodiment 6 elaborates the authorization override logic rule of the arbitration decision module in detail;

[0127] determining whether the predictive risk score is greater than the dynamic intervention threshold ;

[0128] determining whether the real-time safety state of the battery management system is normal, that is indicating that the physical BMS has not triggered an alarm;

[0129] when the condition is met and that is, the risk prediction is over-standard and the physical BMS has not alarmed, determining that the override flag is an executed AI override, that is ;

[0130] when the predictive risk score is less than or equal to the dynamic intervention threshold or the real-time safety state is abnormal, that is or , determining that the override flag is an unexecuted override, that is ;

[0131] The rule principle is that this condition; it clearly shows that the override right of the architecture is only for the AI prediction of future collapse risk under the current normal condition of physical measurement;

[0132] The preset authorization override logic of the embodiment is the core of solving the data trust conflict; through logical and association and , the architecture realizes the predictive override of the EMS economic instruction, while avoiding the permission conflict between AI and physical BMS in the alarm state; this makes the system be able to safely and reliably implement proactive risk intervention before the BMS physical alarm occurs.

[0133] Embodiment 8

[0134] The instruction correction module is configured to determine a derating factor, and includes:

[0135] when the override flag is an executed override;

[0136] a risk overshoot amount of the predictive risk score exceeding the dynamic intervention threshold;

[0137] determine a derating amplitude based on the risk overshoot amount and a preset derating response coefficient;

[0138] subtract the derating amplitude from the reference value to obtain a derating factor;

[0139] the instruction correction module is configured to generate a corrected dispatch instruction, including:

[0140] when the veto flag is an execution veto, multiply the planned dispatch power by the derating factor to generate the corrected dispatch instruction;

[0141] The embodiment details how the instruction correction module determines the derating factor and generates the corrected dispatch instruction when the veto flag is an execution veto based on embodiment 7. ;

[0142] when , calculate a risk overshoot amount of the predictive risk score exceeding the dynamic intervention threshold, i.e. ;

[0143] determine a derating amplitude based on the risk overshoot amount and a preset derating response coefficient, i.e. ;

[0144] subtract the derating amplitude from the reference value 1, and use the function to ensure that is not negative, to obtain the derating factor ;

[0145] ;

[0146] is the derating factor, with a dimension of 1, and a value of ; is the derating response coefficient, which is a preset safety policy parameter, with a dimension of 1;

[0147] The proportional P part of the formula is designed to implement an active derating strategy: the more the risk exceeds, the greater the derating amplitude. The linear proportional relationship used in this embodiment is to achieve a fast and stable response; in other embodiments, the derating amplitude can also be a nonlinear function of the risk overshoot amount, for example , to achieve a more aggressive derating response to a larger risk overshoot, further improving the physical fidelity, wherein ​​​​​​​​​A preset nonlinear exponent greater than 1;

[0148] EMS planned scheduling power With deflator Multiply to generate a corrected scheduling instruction. ;

[0149] ;

[0150] By adopting proportional control based on risk overshoot To determine the derating range, this embodiment achieves both precision and responsiveness in instruction correction; high This value signifies that the system is highly responsive, capable of rapidly reducing command power and proactively lowering short-term economic indicators, thereby achieving the fastest possible recovery of the BESS's internal health with minimal intervention. This closed-loop control corrects the EMS's efficiency-maximizing commands based on physical measurements, ensuring safe execution under predicted risks.

[0151] Example 9

[0152] The instruction correction module, used to generate corrected scheduling instructions, also includes:

[0153] When the arbitration decision module determines that the veto flag is not executed, the planned scheduling power will be directly determined as the corrected scheduling instruction;

[0154] This embodiment, based on embodiment 1, clarifies when the arbitration decision module determines the rejection flag. For the veto not being executed At that time, the instruction modifies the execution logic of the module;

[0155] when and If the AI ​​does not reject the request and the physical BMS is in normal condition, it indicates that the system is in a safe operating zone. In this case, the power will be scheduled according to the original instructions from the EMS. The direct output is the final execution instruction, which modifies the scheduling instruction. ;

[0156] when This means that the physical BMS has triggered an alarm, regardless of... and When the system automatically relinquishes control, the scheduling instructions should be corrected. The highest priority physical protection logic of the BMS should be followed, for example, ... Forced to 0 or the BMS preset safe power value.

[0157] The embodiment ensures that the system does not intervene in the original economic dispatching instruction of the EMS when the predicted risk is controllable and the physical state is normal; meanwhile, the AI prediction mechanism automatically transfers the highest control right when the physical BMS has triggered an alarm, ensures the highest priority of physical protection, and solves the permission conflict.

[0158] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application rather than limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the present application.

Claims

1. An integrated photovoltaic-storage-charging site energy storage system, characterized in that, include: The data acquisition module is used to acquire in real time the state feature vector of the energy storage system, the planned scheduling power of the upper-level energy management system, the external power grid interference index, the real-time safety status of the battery management system, and the economic value index. The risk assessment module is used to calculate the electrochemical instability index based on the state feature vector and through a preset digital twin model. Based on planned dispatch power and grid interference indicators, a dynamic stress factor is calculated; then, by combining the electrochemical instability index and the dynamic stress factor, a predictive risk score is generated. The decision threshold module is used to determine the static safety baseline based on the preset thermal runaway failure boundary and basic safety margin, and to calculate the dynamic intervention threshold in combination with economic value indicators. The arbitration decision module is used to generate a veto flag when the predictive risk score is greater than the dynamic intervention threshold and the real-time safety status is normal. The instruction correction module is used to determine the derating factor in response to the veto flag, and correct the planned scheduling power based on the derating factor to generate a corrected scheduling instruction; Alternatively, if no veto flag is generated, the planned scheduling power can be output as a corrective scheduling instruction.

2. The integrated photovoltaic-storage-charging site energy storage system according to claim 1, characterized in that, The state feature vector includes: real-time voltage, real-time temperature, cumulative number of cycles, and preset manufacturing defect level.

3. The integrated photovoltaic-storage-charging site energy storage system according to claim 1, characterized in that, The risk assessment module is used to calculate the electrochemical instability index via a preset digital twin model, including: Input the state feature vector into the preset digital twin model; The electrochemical instability index, which characterizes internal fragility, is calculated using a digital twin model.

4. The integrated photovoltaic-storage-charging site energy storage system according to claim 1, characterized in that, The risk assessment module is used to calculate the dynamic stress factor, including: The power stress is determined based on the ratio of the planned power to the preset rated power. Determine the power grid interference stress based on power grid interference indicators; The dynamic stress factor is obtained by weighted summation of power stress and grid interference stress.

5. The integrated photovoltaic-storage-charging site energy storage system according to claim 1, characterized in that, The risk assessment module is used to generate predictive risk scores, including: Calculate the external stress amplification factor based on the dynamic stress factor; The predictive risk score is obtained by multiplying the electrochemical instability index by the external stress amplification factor.

6. The integrated photovoltaic-storage-charging site energy storage system according to claim 1, characterized in that, The decision threshold module is used to calculate the dynamic intervention threshold, including: Based on the static safety baseline, the preset economic hedging coefficient, and the economic value indicator, the dynamic intervention threshold is determined.

7. The integrated photovoltaic-storage-charging site energy storage system according to claim 1, characterized in that, The arbitration decision module is used to generate a rejection flag, including: Determine whether the predictive risk score is greater than the dynamic intervention threshold; Determine whether the real-time security status is normal; When the predictive risk score is greater than the dynamic intervention threshold and the real-time safety status is normal, the rejection flag is to execute the rejection. When the predictive risk score is less than or equal to the dynamic intervention threshold, or when the real-time safety status is abnormal, the veto sign is determined as "no veto executed".

Images