Energy storage simulation battery regulation and control system based on fiber grating sensor

Abstract

The invention relates to the technical field of battery simulators, and discloses a fiber grating sensor-based energy storage simulation battery regulation and control system, which comprises a simulation module and a control module, and is characterized in that the simulation module is internally provided with a heating sheet for simulating heating, an air bag for generating strain and a fiber grating main body for sensing; the control module comprises an industrial personal computer for operating a neural network model, a fiber bragg grating demodulator, a programmable power supply for driving the heating plate and a unit for regulating and controlling the pressure of the air bag, and when the system works, a fiber bragg grating body measures temperature and strain, demodulates the temperature and strain and sends the demodulated temperature and strain to the industrial personal computer; a neural network model in the industrial personal computer predicts target temperature and strain and forms two parallel closed loops. Accurate decoupling and synchronous measurement of temperature change and strain data are achieved through double-fiber grating differential measurement, the dynamic response speed and precision of simulation are improved through a predictive neural network model, and complex electro-thermal coupling physical characteristics of a real battery are reproduced by means of a parallel double-closed-loop control framework.

Description

Technical Field

[0001] This invention relates to the field of battery simulator technology, specifically to an energy storage simulation battery regulation system based on a fiber Bragg grating sensor. Background Technology

[0002] With the rapid development of renewable energy and the widespread adoption of electric vehicles, energy storage technology, especially chemical battery energy storage technology represented by lithium-ion batteries, is playing an increasingly important role. The Battery Management System (BMS) is the core component ensuring the safe, reliable, and efficient operation of battery packs. In the research, development, testing, and verification of BMS, as well as in the development and testing of battery-interacting devices such as power electronic converters and charging piles, directly using real battery packs for experiments is not only costly and inefficient but also poses certain safety risks. This is especially true during fault injection testing or extreme condition testing. As the core component of electric vehicles, consumer electronics, and large-scale grid energy storage, the safety, reliability, and lifespan of energy storage batteries, particularly lithium-ion batteries, are crucial to the performance of the entire system. Using real battery packs for experiments during battery research, testing, and the calibration and verification of the Battery Management System (BMS) is not only costly and time-consuming but also presents uncontrollable safety risks under extreme conditions. Therefore, battery simulator technology, which can accurately simulate the physical characteristics of real batteries, has emerged and become a key link in the product development and verification process.

[0003] Existing battery simulators primarily focus on simulating the electrical characteristics of batteries. They reproduce the battery's voltage and current response curves using a programmable power supply, and are used to test the charge / discharge control and state-of-charge (SOC) estimation functions of battery management systems. However, a real battery is a complex electro-thermal-mechanical multiphysics coupling system during operation. Electrochemical reactions during charging and discharging generate significant heat, leading to increased battery temperature. Simultaneously, ion insertion and extraction cause microscopic volume changes in the electrode materials, macroscopically manifesting as physical expansion or contraction, generating internal stress and external strain. These thermal and mechanical effects are not isolated but interconnected and tightly coupled, profoundly impacting battery performance, lifespan, and safety.

[0004] Currently, the technical solutions capable of simulating these physical characteristics are still imperfect, with their main deficiency lying in the insufficient ability to reproduce the coupling effects of multiple physics fields. Existing physical simulation devices often separate the simulation of physical quantities such as temperature and deformation. For example, they generate a preset, static temperature distribution through heating elements or apply a fixed stress through mechanical devices. This approach ignores the dynamic and real-time interaction between the thermal field and the stress field under real-world conditions. Furthermore, the sensing and control methods used in existing simulation systems also have limitations. Traditional sensors such as thermocouples and strain gauges are susceptible to electromagnetic interference, have slow response speeds, and are difficult to accurately characterize the physical field distribution. This results in low fidelity of the feedback information acquired by the control system, leading to control strategies that are mostly simple, hysteresis-based feedback control. This makes it impossible to accurately and quickly track the drastic physical characteristic changes of real batteries under dynamic conditions such as rapid charging and discharging, resulting in significant deviations between the simulation results and the real situation, making it difficult to meet the needs of high-precision research and development and testing. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an energy storage simulation battery control system based on fiber Bragg grating sensors. This system solves the technical problem that existing battery simulation technologies, due to limitations in sensing and control methods, cannot accurately and dynamically reproduce the tightly coupled thermo-mechanical physical characteristics inside a real battery, resulting in significant deviations between simulation results and actual conditions.

[0006] To address the aforementioned technical problems, this invention provides an energy storage analog battery regulation system based on a fiber Bragg grating sensor.

[0007] The system includes a simulation module and a control module.

[0008] The simulation module is used to construct the physical body of the simulated battery. It includes an aluminum battery casing filled with a heat transfer medium. Inside the casing, there are heating elements to simulate battery heat generation and air bladders to generate controllable physical deformation. Simultaneously, a fiber Bragg grating body for monitoring temperature and strain is mounted on the simulation module, and a protective cover is provided for its physical protection.

[0009] The control module is used to perform sensor interpretation and closed-loop control. It includes an industrial computer, a fiber Bragg grating demodulator, a programmable power supply, and an airbag pressure control unit. The airbag pressure control unit includes a miniature air pump.

[0010] Regarding system structure connections:

[0011] The fiber grating body is optically connected to the fiber grating demodulator.

[0012] The fiber optic grating demodulator communicates with the industrial control computer via a network cable.

[0013] The programmable power supply is electrically connected to the heating element via wires, providing the latter with driving current.

[0014] The industrial control computer establishes control communication with the programmable power supply and the micro air pump, respectively.

[0015] The miniature air pump is connected to the air bladder via a rubber tube that passes through a pre-drilled hole in the battery casing.

[0016] The fiber Bragg grating body consists of a sensing fiber Bragg grating and a reference fiber Bragg grating. The two fiber Bragg gratings are positioned to place them in the same temperature field, but the reference fiber Bragg grating is isolated from the structure under test to prevent it from being affected by strain. The center wavelength shift (Δλ) of the fiber Bragg grating (FBG) is... B The relationship between the change in temperature (ΔT) and the change in strain (Δ∈) can be expressed by the following formula:

[0017] Δλ B =K T ·ΔT+K ∈ ·Δ∈;

[0018] Among them, K T K is the temperature sensitivity coefficient. ∈ This is the strain sensitivity coefficient.

[0019] For this system:

[0020] The wavelength shift of the sensing fiber optic grating is: Δλ sens =K T,sens ·ΔT+K ∈,sens ·Δ∈;

[0021] The wavelength drift of the reference fiber grating is: Δλ ref =K T,ref ·ΔT;

[0022] In the formula:

[0023] Δλ sens This represents the center wavelength shift of the sensing fiber optic grating.

[0024] Δλ ref The center wavelength shift of the reference fiber grating;

[0025] K T,sens The temperature sensitivity coefficient of the sensing fiber Bragg grating;

[0026] K ∈,sens The strain sensitivity coefficient of the sensing fiber optic grating;

[0027] KT,ref The temperature sensitivity coefficient of the reference fiber grating;

[0028] ΔT represents the temperature change at the locations of the two fiber gratings;

[0029] Δ∈ represents the strain change experienced by the sensing fiber optic grating.

[0030] The fiber grating demodulator monitors the center wavelength shift Δλ of the two fiber gratings in real time. sens and Δλ ref By solving the above set of equations, the independent temperature change ΔT and strain change Δ∈ can be separated and calculated. The demodulator then sends the decoupled temperature and strain data to the industrial control computer via the Ethernet interface.

[0031] Bi-objective parallel control based on neural network model

[0032] After receiving real-time, independent temperature and strain data, the industrial control computer runs a built-in hybrid neural network model. This model is preferably a physical information recurrent neural network based on gated recurrent units or long short-term memory networks. The model is trained offline, and the training dataset contains measured temperature and strain data of real batteries under various charge and discharge conditions, along with the corresponding electrical and thermal parameters.

[0033] During the operational phase, the model takes real-time sensed temperature and strain data, along with user-preset simulation parameters, as input, and calculates and outputs two target commands in parallel:

[0034] Target heating command: used to regulate the temperature field.

[0035] Target strain command: used to control physical deformation.

[0036] Closed-loop control of temperature field and physical deformation

[0037] This system contains two parallel, independent closed-loop control loops:

[0038] Temperature control loop: The industrial control computer compares the target temperature output by the model with the actual measured temperature fed back by the fiber optic grating body to obtain the temperature deviation. Based on this deviation, the target current value to be applied to the heating element is calculated through a temperature control algorithm. This target current value is sent as a command to the programmable power supply. According to the command, the programmable power supply accurately outputs the corresponding current to the heating element through wires, thereby regulating the temperature field of the simulation module. To achieve the simulation of a non-uniform temperature field, the heating element can be set as multiple zone heating elements, each controlled separately by multiple independent channels of the programmable power supply.

[0039] Strain control loop: The industrial control computer compares the target strain output by the model with the actual measured strain fed back by the fiber optic grating body to obtain the strain deviation. Based on this deviation, a strain control algorithm calculates the control command to drive the micro air pump. In addition, the industrial control computer has a pre-stored strain-pressure conversion model, which directly converts the target strain value predicted by the neural network into the target internal pressure value required by the airbag. The mathematical expression of this conversion model can be a polynomial fitting function obtained through experimental calibration:

[0040]

[0041] In the formula, P target To calculate the target pressure, ε target For the target strain predicted by the neural network, c0, c1, c2, ..., c n These are the model coefficients, obtained through data fitting, characterizing the mechanical response properties of this specific simulation module. The industrial control computer issues commands to control the micro-pump, inflating or deflating the airbag through a rubber tube, precisely adjusting its internal pressure so that the resulting deformation acts on the battery casing until the strain value measured on the fiber grating body matches the target strain value.

[0042] Through the parallel, independent, and real-time closed-loop control of the temperature field and physical deformation, the dynamic changes in external temperature and strain of the simulated battery measured by the fiber optic grating body can accurately reproduce the physical characteristics of the real battery under the corresponding charge and discharge conditions.

[0043] Further limitations on component materials:

[0044] To ensure the effectiveness of the simulation and the durability of the equipment, the filling material is expandable microspheres, specifically composite microspheres with an acrylonitrile copolymer shell and coated with isobutane or isopentane as a foaming agent. The air bladder is made of high-temperature resistant silicone rubber material, specifically methyl vinyl silicone rubber with added ferric oxide or cerium oxide heat-resistant agents.

[0045] This invention provides an energy storage analog battery regulation system based on a fiber Bragg grating sensor. It has the following advantages:

[0046] 1. This invention constructs a differential measurement structure by using a sensing fiber grating that is sensitive to both temperature and strain and a reference fiber grating that is only sensitive to temperature. By using a decoupling algorithm, the cross-sensitivity effect of temperature and strain can be effectively separated, thereby realizing synchronous, high-precision, and cross-interference-free measurement of the surface temperature and strain of the simulated battery, providing a reliable data foundation for subsequent precise closed-loop control.

[0047] 2. This invention introduces a physical information recurrent neural network model pre-trained based on real battery data. This model can actively predict the target state of the battery's temperature and strain at the next moment based on the current operating conditions and sensor data. This predictive control method overcomes the lag of traditional feedback control, thereby enabling it to more closely track and reproduce the rapidly changing nonlinear thermo-mechanical coupling behavior of real batteries, significantly improving the dynamic response speed and reproduction accuracy of the simulation.

[0048] 3. This invention designs two parallel, independent closed-loop control systems for temperature and strain. A programmable power supply precisely drives the heating element to regulate the temperature field, while a micro-pump actively controls the air bladder to regulate physical deformation. This dual-channel parallel control architecture enables independent and precise adjustment of these two key physical quantities, thereby faithfully reproducing the complex electro-thermal-mechanical effects of a real battery during charging and discharging, providing a systematic solution. Attached Figure Description

[0049] Figure 1 This is an overall block diagram of the present invention;

[0050] Figure 2 This is a schematic diagram of the hybrid neural network model architecture of the present invention;

[0051] Figure 3 This is a comparison diagram of the strain of the casing of a real battery versus a simulated battery under 8V / 2A conditions according to the present invention.

[0052] Figure 4 The strain difference diagram of the actual battery casing versus the simulated battery casing under 8V / 2A is shown in the present invention.

[0053] Figure 5 This is a temperature comparison chart of the casing of a real battery versus a simulated battery under 8V / 2A conditions according to the present invention;

[0054] Figure 6 This is a temperature difference diagram of the casing of a real battery versus a simulated battery under 8V / 2A conditions according to the present invention.

[0055] The components include: 1. Battery casing; 2. Filling material; 3. Heating element; 4. Airbag; 5. Protective cover; 6. Fiber Bragg grating body; 7. Industrial computer; 8. Fiber Bragg grating demodulator; 9. Programmable power supply; 10. Network cable; 11. Wire; 12. Miniature air pump; and 13. Rubber hose. Detailed Implementation

[0056] The technical solutions in the embodiments of the present invention will now be clearly and completely described with reference to the accompanying drawings.

[0057] See appendix Figure 1-6This invention provides an energy storage simulation battery regulation system based on a fiber Bragg grating sensor, including a control module and a simulation module.

[0058] The simulation module includes an aluminum battery casing 1, filling material 2, ceramic cuboid heating element 3, air bag 4, fiber optic grating protective cover 5, and fiber optic grating 6.

[0059] The aluminum battery casing 1 measures 210mm in length, 70mm in width, and 350mm in height.

[0060] Filler material 2 is expandable microspheres with an average particle size of 25–35 micrometers, an initial expansion temperature of 150–160℃, and a peak expansion temperature of 190–200℃.

[0061] The ceramic rectangular heating element 3 has a length*width*thickness of 40mm*40mm*2mm, a rated voltage of 12V and a rated power of 96W, and a surface temperature of up to 330℃ under the rated power.

[0062] The aluminum battery casing 1 is filled with a heat transfer medium, a filler material 2, which has a thermal conductivity of 0.26 W / m·K, similar to that of sand. For example, it can be made of quartz sand, alumina powder, or boron nitride powder. The fiber optic grating 6's sensing area is attached to the outer surface of the aluminum battery casing 1 using thermally conductive silicone grease. Figure 1 The center of the side of the aluminum battery casing 1, with a length * width * height of 210mm * 70mm * 350mm, is used to sense the temperature and strain information of the battery's exterior in real time and transmit the light signal to the control module; the heating element 3 is used to receive instructions from the control module to generate heat to simulate battery heating; the airbag 4 is preferably made of high-temperature resistant elastomer material silicone rubber, which must ensure that it can maintain structural integrity and high airtightness at working temperatures up to 350℃, and its elastic modulus is preferably in the range of 1MPa to 100MPa, with a length * width * height of 10mm * 65mm * 200mm in its natural state.

[0063] To achieve active and precise control over the inflation degree of the airbag 4, the control module further includes an active airbag pressure regulation unit. This unit includes a precision miniature air pump 12 controlled by instructions from the industrial control computer 7.

[0064] The industrial control computer 7 predicts the target strain ε at the next moment based on the neural network model. target (t+△t), using a pre-established strain-pressure conversion model, calculates the target pressure P that needs to be applied inside the airbag 4. target (t+△t).

[0065] This strain-pressure conversion model describes the nonlinear relationship between the strain on the outer surface of the simulated battery casing 1 and the internal pressure of the air bladder 4. Since this relationship is closely related to the specific structure and material properties of the system, the model is pre-established through physical prototype experimental calibration. After obtaining a series of corresponding data points for pressure P and strain ε using the above method, the conversion model can be established using polynomial fitting, and its mathematical expression can be:

[0066]

[0067] In the formula, P target To calculate the target pressure, ε target For the target strain predicted by the neural network, c0, c1, c2, ..., c n These are model coefficients that characterize the mechanical response properties of this specific simulation module, obtained through data fitting. These coefficients are pre-stored in the industrial control computer 7.

[0068] In obtaining the target pressure P target Subsequently, the industrial control computer 7, through the airbag pressure active control unit, adjusts the airbag pressure based on the target pressure P. target The deviation between the pressure and the actual pressure fed back by the pressure sensor is precisely adjusted by on / off control logic to regulate the operation of the micro-pump, thereby enabling the internal pressure of the airbag to quickly and stably reach the target pressure value. Through this active and precise pressure control, the airbag 4 can apply a controllable force to the aluminum battery casing 1 to reproduce the physical deformation of a real battery under various working conditions.

[0069] The control module includes an industrial computer 7, a fiber Bragg grating demodulator 8, a programmable power supply 9, a network cable 10, a power cord 11, a miniature air pump 12, and a rubber hose 13. The fiber Bragg grating demodulator 8 is connected to the fiber Bragg grating 6 in the simulation module and is used to demodulate the received optical signal. It uses dual-grating decoupling technology to separate temperature and strain data, and sends the temperature and strain data to the industrial computer 7 via the network cable 10. The programmable power supply 9 is connected to the heating element 3 in the simulation module and receives control signals from the industrial computer 7 via the network cable 10. The industrial computer 7 receives real-time temperature and strain data from the fiber Bragg grating demodulator 8 and combines it with user-defined simulation operating condition parameters characterizing the current target operating state of the simulated battery. These simulation operating condition parameters are represented by a specific current value, for example, used to characterize the operation of an 18650 lithium iron phosphate battery at a constant current of 2A, or to characterize a specific charge / discharge process or standard test cycle. The hybrid neural network model fuses the simulated operating parameters with the real-time temperature and strain data fed back by the fiber optic grating 6, and calculates the precise current output command required to drive the heating element 3 based on this, so as to ensure that the simulated battery can reproduce the thermal and strain characteristics of the real battery under the set operating conditions.

[0070] The fiber grating 6 used for monitoring temperature and strain is actually composed of a sensing fiber grating FBG. s and a reference fiber optic grating (FBG) r The sensor pair is used to achieve accurate decoupled measurement of surface temperature and strain of the simulated battery casing 1.

[0071] Fiber Bragg Grating (FBG) s The center wavelength shift Δλ is measured at the center position of the test location, where thermally conductive silicone grease is tightly adhered to the outer wall of the aluminum battery casing 1. r It is simultaneously affected by both temperature change ΔT and strain ε.

[0072] Reference Fiber Bragg Grating (FBG) r It is encapsulated in a tiny protective tube, thus achieving mechanical decoupling from stress and connecting with the sensing fiber Bragg grating (FBG). s They are arranged side-by-side and adjacent to each other to ensure that they are in the exact same temperature field. Therefore, the drift Δλ of their center wavelength r It is caused solely by the temperature change ΔT.

[0073] The relationship between the wavelength shift of a fiber Bragg grating and temperature and strain can be described by the following fundamental equation:

[0074]

[0075] In the formula, λ is the center wavelength of the grating, Δλ is the wavelength shift, ε is the strain applied to the grating, and K ∈ Here, K represents the strain sensitivity coefficient, ΔT represents the ambient temperature change, and K represents the K value. T This is the temperature sensitivity coefficient. Both coefficients are inherent properties of the fiber grating material and structure, and can be obtained through prior experimental calibration.

[0076] For the sensing pair in this invention, the following set of equations applies:

[0077] Reference grating: Δλ r =K T,r ΔTλ r ;

[0078] Sensing grating: Δλ s =K ε,s ελ s +K T,s ΔTλ s ;

[0079] Since the two gratings are adjacent and have similar characteristics, their temperature sensitivity coefficients can be considered approximately equal (K). T,s ≈K T,r Therefore, the decoupling process is as follows:

[0080] The fiber optic grating demodulator 8 acquires the wavelength shift Δλ of the two gratings in real time. s and Δλ r .

[0081] First, the real-time temperature change ΔT is calculated using the readings of the reference grating:

[0082] Then, substituting the calculated ΔT into the equation of the sensing grating, the strain ε is solved:

[0083] Using the above method, the fiber optic grating demodulator 8 can accurately separate the real-time temperature change ΔT and strain ε on the surface of the simulated battery casing 1, and send these two independent physical quantities to the industrial control computer 7, providing high-precision input for the subsequent hybrid neural network model. This decoupling method using a reference grating for temperature compensation is simple in structure and the calculation process is intuitive and reliable.

[0084] The heating element 3 in the simulation module is a single heating element, placed inside the air bag 4 and tightly attached by thermal grease. It is independently controlled by a channel of the programmable power supply 9 to simulate and regulate the non-uniform distribution of the surface temperature field of the simulated battery case 1.

[0085] The air bladder 4 in the simulation module is used to generate a certain volume change under the action of heating and micro air pump 12, which helps to generate strain signals that can be measured by fiber optic grating 6, and simulate some mechanical behaviors of real batteries.

[0086] The industrial control computer 7 in the control module runs a Physical Information Recurrent Neural Network (PI-RNN) based on GRU / LSTM, which has been pre-trained offline. This hybrid neural network model was obtained through pre-training offline. The training data is based on experimental tests of 18650 lithium iron phosphate batteries under various operating conditions. These various operating conditions include, but are not limited to: constant current constant voltage CC-CV charging process, such as charging to the nominal voltage at 0.5C and then maintaining constant voltage to the cutoff current; constant current discharge processes at different rates, such as discharging to the cutoff voltage at 0.5C, 1C, and 2C; and the above tests at different ambient temperatures, such as 15℃, 25℃, and 45℃. Under these operating conditions, the following high-precision data were simultaneously collected for model training: a) real-time temperature distribution data of the side surface of the 18650 lithium iron phosphate battery casing was obtained through a thermocouple array with multiple probes arranged in the central area of ​​the side surface of the 18650 lithium iron phosphate battery casing; b) real-time strain distribution data of the side surface of the 18650 lithium iron phosphate battery casing was obtained by attaching high-precision strain gauges to multiple probes in the central area of ​​the side surface of the 18650 lithium iron phosphate battery casing; c) corresponding real-time electrical parameters, including the actual charge and discharge current curves, voltage curves, and real-time power curves calculated from these curves; d) and other relevant parameters that significantly affect the battery's thermo-mechanical behavior, such as ambient temperature during the experiment. By learning the complex nonlinear relationships between these multi-dimensional, time-series data, the model can accurately predict and generate a precise current output control strategy for the heating element 3 based on the received target simulation operating condition parameters, such as the set target current value, power sequence, or the current operating point in a specific test cycle, as well as the real-time temperature and strain sensing feedback from the fiber optic grating 6 of this system.

[0087] The industrial PC 7 runs a built-in hybrid neural network model, the architecture of which is as follows: Figure 2 As shown. This model utilizes real-time temperature and strain data received from the fiber Bragg grating demodulator 8, along with set simulation parameters, to extract spatiotemporal features through its internal bidirectional LSTM-BiGRU structure. Combined with constraints from the Physical Information Neural Network (PINN), a dual-head fully connected network predicts the next control cycle, for example, t + Δt, where Δt is the control cycle, such as the target temperature T that the surface of the aluminum battery casing 1 should reach in 10 ms. predicted (t+△t) and target strain ε predicted (t+△t).

[0088] Subsequently, the industrial control computer 7 will predict the target temperature T. predicted (t+Δt) and target strain ε predicted (t+Δt) are respectively compared with the current actual measured temperature T fed back in real time by fiber grating 6. actual (t) and the current actual measured strain ε actualBy comparing (t), the temperature deviation Error is obtained. T =T predicted (t+Δt)-T actual (t); and strain deviation Error ε =ε predicted (t+Δt)-ε actual (t);

[0089] The industrial computer 7 runs two parallel PID control loops, which independently control temperature and strain through closed-loop regulation:

[0090] Temperature control loop: This will control temperature deviation errors. T As input, the first PID controller calculates the target current value to be applied to the heating element 3, and the industrial computer 7 then sends this target current signal to the programmable power supply 9, which precisely drives the heating element 3 to work.

[0091] Strain control loop: This will control strain deviation errors. ε As input, the required power-on time for the micro air pump 12 is calculated through the on / off control logic, and the industrial control computer 7 then sends this instruction to the airbag pressure regulation unit to drive the micro air pump 12 to work.

[0092] By implementing parallel and independent closed-loop control of temperature and strain as described above, it is ensured that the temperature and strain of the simulated battery can accurately track and reproduce the dynamic characteristics of the real battery under corresponding operating conditions.

[0093] The overall realism of the simulator in this invention is a quantitative index used to comprehensively evaluate the degree to which the simulated battery reproduces the real battery in two physical fields: temperature and strain. This overall realism is calculated by a weighted average of the simulation accuracy of each physical field. The specific calculation steps are as follows:

[0094] Calculate the normalized root mean square error (NRMSE) for each physical field: First, for specific operating conditions... Figures 3 to 6 The data collected at N time points throughout the entire process under the 8V / 2A operating condition are shown. The root mean square error (RMSE) of temperature and strain is calculated respectively.

[0095]

[0096] In the formula:

[0097] RMSE T This represents the root mean square error of the temperature.

[0098] RMSE ε The root mean square error of the strain;

[0099] N is the total number of data points collected;

[0100] i is the index of the data point, from 1 to N;

[0101] T sim,i This represents the simulated temperature value at the i-th time point;

[0102] T real,i This represents the actual temperature value at the i-th time point;

[0103] ∈ sim,i Let be the simulated strain value at the i-th time point.

[0104] ∈ real,i This represents the actual strain value at the i-th time point;

[0105] Then, the RMSE is normalized using the difference between the maximum and minimum values ​​of the measurement range throughout the entire test using real data, resulting in the NRMSE:

[0106]

[0107] In the formula:

[0108] NRMSE T This represents the normalized root mean square error of temperature.

[0109] NRMSE ε The normalized root mean square error of the strain;

[0110] T real,max This represents the maximum actual temperature during the entire test process;

[0111] T real,min This represents the minimum actual temperature during the entire test process;

[0112] ∈ real,max This represents the maximum actual strain during the entire testing process;

[0113] ∈ real,min This represents the minimum actual strain during the entire testing process;

[0114] Calculate the individual simulation accuracy for each physics field: The individual simulation accuracy is derived from the normalized root mean square error, and the calculation formula is as follows:

[0115]

[0116] Calculating the weighted comprehensive fidelity: Considering that temperature and strain may have different importance in battery thermal management and structural safety assessments, a weighted average method is used to calculate the comprehensive fidelity. The calculation formula is as follows:

[0117] Fidelity = w T ×AccuracyT +w ε ×Accuracy ε ;

[0118] In the formula, Accuracy T Accuracy indicates the accuracy of a single simulation of the temperature physical field. ε NRMSE represents the accuracy of a single simulation of the strain physical field. T Normalized Root Mean Square Error (NRMSE) represents the normalized root mean square error of temperature. ε The normalized root mean square error represents the strain; Fidelity represents the weighted comprehensive Fidelity. T and w ε The weighting coefficients for temperature simulation accuracy and strain simulation accuracy, respectively, w T +w ε =1.

[0119] In one specific embodiment of the present invention, by... Figures 3 to 6 The calculations based on the experimental data show that the single-term simulation accuracy for temperature is approximately 98.1%, and the single-term simulation accuracy for strain is approximately 96.8%. With the weight w set... T =0.75 and w ε When the value is 0.25, the overall realism of this simulator is calculated to be 97.78%, where the weighting coefficient w T and w ε The system can be preset or adjusted according to different emphases on temperature and strain simulation accuracy in actual application scenarios. This indicates that the control system of the present invention can reproduce the electro-thermal-mechanical multiphysics coupling behavior of real batteries with high fidelity.

[0120] The fiber optic grating demodulator 8 has high-speed demodulation capability, which can meet the real-time monitoring requirements of rapid temperature and strain changes during battery charging and discharging, and transmit the decoupled data to the industrial control computer 7 through the standard communication interface Ethernet.

[0121] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A power storage analog battery regulation system based on fiber Bragg grating sensors, characterized in that, include: Control module and simulation module; The simulation module includes an aluminum battery shell (1) for simulating the physical form of a battery, filled with a filling material (2) as a heat transfer medium, and equipped with a heating element (3) for simulating heat generation, an air bladder (4) for generating controllable strain, and a fiber optic grating body (6) for monitoring temperature and strain. The fiber optic grating body (6) is provided with a protective cover (5) for protecting it. The control module includes an industrial computer (7) for running algorithms and models, a fiber optic demodulator (8) for demodulating fiber optic signals, a programmable power supply (9) for driving the heating element (3), and an airbag pressure control unit. The fiber grating body (6) includes a sensing fiber grating and a reference fiber grating. The fiber grating demodulator (8) is connected to the two fiber gratings and transmits data to the industrial control computer (7) via a network cable (10). The industrial control computer (7) is used to receive independent temperature and strain data and run the built-in hybrid neural network model. The model calculates in parallel the target heating command for temperature control and the target strain command for strain control based on the input real-time sensor data and preset simulation working condition parameters. The industrial control computer (7) sends the target heating command to the programmable power supply (9) and sends the target strain command to the airbag pressure control unit. The programmable power supply (9) precisely controls the current output to the heating element (3) through the wire (11) according to the target heating command, so as to regulate the temperature field of the simulation module. The airbag pressure control unit performs active closed-loop control on the airbag (4) according to the target strain command, so as to regulate the physical deformation of the simulation module; Thus, through parallel closed-loop control of temperature field and physical deformation, the dynamic change process of simulated external battery temperature and strain measured by fiber optic grating body (6) can accurately reproduce the physical characteristics of real battery under corresponding charging and discharging conditions.

2. The energy storage analog battery regulation system based on a fiber Bragg grating sensor according to claim 1, characterized in that, The fiber grating body (6) in the simulation module is deployed at a single point to obtain temperature and strain distribution information of key areas on the surface of the simulated battery case (1).

3. The energy storage analog battery regulation system based on a fiber Bragg grating sensor according to claim 1, characterized in that, The heating element (3) in the simulation module consists of multiple partitioned heating elements (3), which are independently controlled by multiple channels of the programmable power supply (9) to simulate and regulate the non-uniform distribution of the surface temperature field of the simulated battery case (1).

4. The energy storage analog battery regulation system based on a fiber Bragg grating sensor according to claim 1, characterized in that, The airbag pressure control unit includes a micro air pump (12), which is connected to the airbag (4) and battery casing (1) via a rubber tube (13). The industrial control computer (7) actively controls the internal pressure of the airbag (4) by controlling the micro air pump (12).

5. The energy storage analog battery regulation system based on a fiber Bragg grating sensor according to claim 1, characterized in that, The hybrid neural network model running on the industrial computer (7) in the control module is a physical information recurrent neural network PI-RNN based on GRU / LSTM. This model is pre-trained offline. The training is based on the measured temperature and strain data of real batteries under various working conditions and the corresponding electrical and thermal parameters, so that the model can learn and predict the electrical-thermal-mechanical coupling response relationship of real batteries.

6. The energy storage analog battery regulation system based on a fiber Bragg grating sensor according to claim 1, characterized in that, The industrial control computer (7) operates two parallel control loops, which independently control the temperature and strain in a closed loop.

7. The energy storage analog battery regulation system based on a fiber Bragg grating sensor according to claim 6, characterized in that, The industrial control computer (7) compares the target temperature predicted by the neural network model with the actual measured temperature fed back by the fiber optic grating to obtain the temperature deviation, and calculates the target current value applied to the heating element (3) through a temperature control loop based on the temperature deviation.

8. The energy storage analog battery regulation system based on a fiber Bragg grating sensor according to claim 1, characterized in that, The industrial control computer (7) compares the target strain predicted by the neural network model with the actual measured strain fed back by the fiber optic grating to obtain the strain deviation, and calculates the control command to drive the micro air pump (12) through a strain control loop based on the strain deviation.

9. The energy storage analog battery regulation system based on a fiber Bragg grating sensor according to claim 1, characterized in that, The industrial control computer (7) pre-stores a strain-pressure conversion model, which is used to calculate the target pressure to be applied to the airbag (4) based on the target strain predicted by the neural network. Its mathematical expression is a polynomial fitting function.

10. The energy storage analog battery regulation system based on a fiber Bragg grating sensor according to claim 1, characterized in that, The filler material (2) is an expandable microsphere, which is a composite microsphere with an acrylonitrile copolymer as the shell and coated with isobutane or isopentane as a foaming agent; the air bag (4) is made of high-temperature resistant silicone rubber material, which is methyl vinyl silicone rubber with added ferric oxide or cerium oxide heat resistant agent.

Images