A battery temperature accurate estimation method based on optical fiber sensing

Abstract

The present application relates to the technical field of energy storage battery optical fiber sensing, and particularly relates to a battery temperature accurate estimation method based on optical fiber sensing, comprising: establishing a first model based on an FCM algorithm, setting a condition change value for iteration termination, training the first model by using a first sample set, obtaining a linear relationship between temperature change and first wavelength offset through repeated iteration, establishing a second model based on an EMD algorithm, setting a condition change value for iteration termination, training the second model by using a second sample set, obtaining a nonlinear relationship between battery monitoring data change and second wavelength offset through repeated iteration, superimposing the linear relationship between temperature change and first wavelength offset and the nonlinear relationship between battery monitoring data change and second wavelength offset, and generating a calibration relationship formula of temperature measurement value and wavelength offset, the estimation method provided by the present application does not need to be calibrated offline, and meanwhile ensures the accuracy and reliability of temperature measurement.

Description

Technical Field

[0001] This invention relates to the field of fiber optic sensing technology for energy storage batteries, and in particular to a method for accurate estimation of battery temperature based on fiber optic sensing. Background Technology

[0002] With the rapid development of battery energy storage technology, my country is paying increasing attention to battery energy storage technology in energy storage power stations. Whether for state estimation or fault early warning, energy storage batteries rely heavily on signal data. While the acquisition of electrical signals in energy storage power stations is mature and relatively accurate, the acquisition of temperature signals presents a challenge. Currently, commonly used temperature measurement devices include thermocouples, thermistors, and infrared thermal imaging. However, electrical temperature sensors such as thermistors and thermocouples require insulated encapsulation when measuring energy storage batteries, and each temperature measurement point occupies a separate information channel. Furthermore, the metal measuring probes are prone to oxidation over long-term use. Therefore, electrical temperature sensors are not suitable for measuring large-scale energy storage battery modules.

[0003] In recent years, fiber optic temperature sensors, a new type of temperature measurement device, have attracted attention for energy storage battery sampling due to their advantages such as small size, light weight, resistance to electromagnetic interference, natural insulation, and ease of reuse and networking. Measuring the temperature of a battery pack using FBG sensors can not only provide accurate temperature parameters for state estimation but also offer more standardized temperature indicators for early warning of thermal runaway. However, decoupling the measurement signal from the fiber optic temperature sensor is a key technical challenge. Traditional temperature calibration methods require the thermal and photosensitive coefficients of the fiber optic temperature sensor, which cannot be obtained under actual operating conditions and necessitate offline testing in specific environments, failing to meet the timely needs of field applications.

[0004] The information disclosed in this background section is intended only to enhance the understanding of the general background of this disclosure and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method for accurate estimation of battery temperature based on fiber optic sensing, which does not require offline calibration, while ensuring the accuracy and reliability of temperature measurement.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: a method for accurate estimation of battery temperature based on fiber optic sensing, comprising:

[0007] S1: Collect the first wavelength offset corresponding to different temperature values ​​of the fiber optic temperature sensor and construct the first sample set;

[0008] S2: Establish a first model based on the FCM algorithm, set the conditional change value for the termination of the iteration, train the first model using the first sample set, and obtain the linear relationship between temperature change and first wavelength offset through repeated iterations;

[0009] S3: During the battery charging and discharging process, collect battery monitoring data and the corresponding second wavelength offset. The battery monitoring data includes at least the battery SOC value and the number of cycles to construct a second sample set.

[0010] S4: Establish a second model based on the EMD algorithm, set the conditional change value for the termination of the iteration, train the second model using the second sample set, and obtain the nonlinear relationship between the changes in battery monitoring data and the second wavelength offset through repeated iterations;

[0011] S5: The linear relationship between the temperature change and the first wavelength offset obtained in step S2 and the nonlinear relationship between the battery monitoring data change and the second wavelength offset obtained in step S4 are superimposed to generate a calibration formula for the temperature measurement value and the wavelength offset.

[0012] Furthermore, in step S2, the objective function designed based on the FCM algorithm is shown in the following equation:

[0013]

[0014] The constraints are as follows:

[0015]

[0016] Where N is the number of input samples, P is the cluster center matrix, U is the membership function matrix, and u ij Let c be the membership degree of sample j belonging to the i-th class, m be the fuzzy index, and x be the number of clusters. j For sample values, c i It serves as the cluster center.

[0017] Furthermore, in step S2, the FCM model is optimized using the Lagrange multiplier method to obtain iterative formulas, specifically including the membership function matrix formula and the cluster center matrix formula.

[0018] Furthermore, the formula for the membership function matrix is:

[0019]

[0020] Where t is the iteration number, d ij For sample x j To cluster center c i The distance, d kj For sample x j To cluster center c kThe distance.

[0021] Furthermore, the formula for the cluster center matrix is:

[0022]

[0023] Among them, c i As the cluster center, u ij Let x be the membership degree of sample j belonging to the i-th class. j These are sample values.

[0024] Furthermore, the specific steps of the repeated iterations include:

[0025] S2.1: Set the condition change value for iteration termination, randomly generate the initial cluster center P0, and set the current iteration number t = 0;

[0026] S2.2: Calculate the membership degree using the membership function matrix formula;

[0027] S2.3: Calculate the cluster centers using the cluster center matrix formula;

[0028] S2.4: If the objective function is less than or equal to the conditional change value at the end of the iteration, stop the iteration calculation and output the final membership degree and cluster center. If the objective function is greater than the conditional change value at the end of the iteration, let t = t + 1 and repeat steps S2.2 to S2.4 to continue the iteration calculation.

[0029] Furthermore, the specific steps of step S4 include:

[0030] S4.1: The functional relationship between wavelength offset and cycle period is calculated using linear fitting.

[0031] S4.2: The functional relationship between wavelength shift and battery SOC value is calculated using polynomial fitting.

[0032] S4.3: Based on the EMD algorithm, the noise proportion is calculated. The acquired second wavelength offset is decomposed by EMD. All maximum and minimum points in the sequence are calculated. The upper and lower envelopes are fitted with cubic spline functions. The average curves of the upper and lower envelopes are calculated. The difference between the data sequence and the average envelope is used to calculate a new data sequence. It is determined whether the IMF condition is met. If not, the residual component is used as the original signal and the iteration is repeated until the condition is met. The resulting function is the first-order intrinsic mode function component.

[0033] Furthermore, the relationship between wavelength shift and cycle period is as follows:

[0034] Z c (t)=r1t 4 +r2t3 +r3t 2 +r4t+r5

[0035] Where t represents the battery cycle period, and r1-r5 are the fitting coefficients.

[0036] Furthermore, the relationship between wavelength shift and battery SOC value is as follows:

[0037]

[0038] Among them, Z ch The functional relationship representing the state of charge, Z dis The function represents the state of discharge, where s represents the SOC value of the battery, and p1-p5 and q1-q5 are both fitting coefficients.

[0039] Furthermore, in step S5, the calibration relationship between the generated temperature measurement value and the wavelength offset is as follows:

[0040] T FBG =T0+f -1 ((λ-λ0)-(Z s +Z c ))

[0041] Where T FBG T0 is the measured temperature, T0 is the compensated temperature, λ represents the measured peak wavelength of the fiber optic temperature sensor, λ0 is the characteristic wavelength of the fiber optic temperature sensor, and Z is the measured temperature. s The effect of SOC variation on wavelength shift; Z c This represents the effect of the number of cycles on the wavelength offset.

[0042] The beneficial effects of this invention are as follows: This invention processes the optical signal collected by the fiber optic temperature sensor by superimposing a linear model with a nonlinear compensation model, calibrates the temperature measurement value, and finally obtains a high-precision and high-reliability temperature measurement value. This avoids the offline verification of energy storage batteries, meets the on-site use requirements of energy storage power stations, and has great promotional value. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart illustrating the effect of temperature change on wavelength shift based on the FCM algorithm in an embodiment of the present invention.

[0045] Figure 2 This is a flowchart of the remaining wavelength offset established based on the EMD algorithm in an embodiment of the present invention. Detailed Implementation

[0046] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0047] All technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0048] It should be noted that the principle of this invention is described as follows:

[0049] Fiber optic temperature sensors reflect a specific narrow band of the spectrum, the central peak wavelength of which is called the Bragg wavelength. The Bragg wavelength varies due to various factors such as thermal and mechanical effects, and can be expressed as:

[0050] λ B =2n e Λ

[0051] Where, n e Λ is the equivalent refractive index of the fiber core, and Λ is the grating period.

[0052] The wavelength shift caused by temperature and stress changes can be expressed as:

[0053]

[0054] Among them, P e Let α represent the photoelastic coefficient, ξ be the thermal expansion coefficient, and ξ be the thermo-optic coefficient of the optical fiber material. ΔT and Δε represent unknown stress changes. The above equation can be simplified to:

[0055] Δλ B =f(ΔT)+Z

[0056] Where, f(ΔT)=λ B (α+ξ)ΔT represents the effect of temperature change on wavelength shift, and Z represents the remaining wavelength shift.

[0057] In this embodiment, a linear function model is proposed for the effect of temperature change on wavelength shift, while a nonlinear function model is proposed for wavelength shift caused by other factors, which is represented by Z in the above formula.

[0058] To address the impact of temperature changes on wavelength shift, this embodiment uses the FCM algorithm for modeling. The FCM algorithm iteratively calculates and minimizes the objective function of the dissimilarity index, obtaining the membership degree of each sample to each class, which is used to determine the class of each sample. First, a sample set X is defined for applying the FCM algorithm. This sample set contains the collected temperature and wavelength values, and can be represented as x. j If the sample set is divided into C classes, then there are C class centers. The objective function of the FCM algorithm is to minimize the sum of the Euclidean distances from each sample to each class center, as shown in the following formula:

[0059]

[0060] The constraints are as follows:

[0061]

[0062] Where N is the number of input samples, P is the cluster center matrix, U is the membership function matrix, and u ij Let c be the membership degree of sample j belonging to the i-th class, m be the fuzzy index, and x be the number of clusters. j For sample values, c i It serves as the cluster center.

[0063] J M The smaller the value, the better the linear fit.

[0064] The FCM model is optimized using the Lagrange multiplier method, yielding the membership function matrix formula and the cluster center matrix. The membership function matrix formula can be expressed as:

[0065]

[0066] Where t is the iteration number, d ij For sample x j To cluster center c i The distance, d kj For sample x j To cluster center c k The distance.

[0067] The formula for the cluster center matrix is ​​expressed as:

[0068]

[0069] Among them, c i As the cluster center, u ij Let x be the membership degree of sample j belonging to the i-th class. j These are sample values.

[0070] Through repeated iterations, we can obtain J. MThe smallest cluster center c i The value of is determined by c. i This allows us to obtain a linear relationship between temperature change and wavelength shift.

[0071] A nonlinear function model is proposed to address wavelength shift caused by other factors, using a combination of short-time and long-time effects. The short-time effect primarily considers electrode changes, which can be reflected by variations in the state of charge (SOC); therefore, SOC variation can be selected as the influencing factor for the short-time effect. For the long-time effect, the electrode volume change is affected by the cycle period, and its maximum stress decreases with increasing cycle count; therefore, the cycle count can be selected as the influencing factor for the long-time effect.

[0072] Based on the above analysis, Z can be decomposed into three parts:

[0073] Z = Z n +Z s +Z c

[0074] Among them, Z n Z represents the measurement noise. s Z represents the impact of changes in monomer SOC. c This represents the impact of the cycle.

[0075] The acquired second wavelength offset is decomposed using EMD. All maxima and minima in the sequence are calculated. The upper and lower envelopes are fitted using cubic spline functions. The average curves of the upper and lower envelopes are calculated. The difference between the data sequence and the average envelope is used to calculate a new data sequence. It is then determined whether the IMF condition is met. If not, the residual component is used as the original signal, and the iteration is repeated until the condition is met. The resulting function is the first-order intrinsic mode function component.

[0076] For Z n Similarly, the EMD denoising algorithm based on the correlation coefficient is used for decomposition. We define m as the number of IMFs when the correlation coefficient between Z and Z-Zn decreases significantly, and Zn can be expressed as:

[0077]

[0078] Among them, h i Let r represent the i-th order IMF, r be the final residual, and L be the number of IMFs in the decomposition.

[0079] Z s There are significant differences between the charging and discharging processes; therefore, polynomial fitting is used to fit the SOC and Z during the charging and discharging processes, respectively. s The relationship between them is as follows:

[0080]

[0081] Among them, Z ch The functional relationship representing the state of charge, Z dis The function represents the state of discharge, where s represents the SOC value of the battery, and p1-p5 and q1-q5 are both fitting coefficients.

[0082] Z c This represents the functional relationship with the cycle period, which can be achieved using linear fitting, as shown below:

[0083] Z c (t)=r1t 4 +r2t 3 +r3t 3 +r4t+r5

[0084] Where t represents the battery cycle period, and r1-r5 are the fitting coefficients.

[0085] In summary, by combining all factors that affect wavelength shift and superimposing them, the calibration relationship between the temperature measurement and wavelength shift is shown in the following formula:

[0086] T FBG =T0+f -1 ((λ-λ0)-(Z s +Z c ))

[0087] Where T FBG T0 is the measured temperature, T0 is the compensated temperature, λ represents the measured peak wavelength of the fiber optic temperature sensor, λ0 is the characteristic wavelength of the fiber optic temperature sensor, and Z is the measured temperature. s The effect of SOC variation on wavelength shift; Z c This represents the effect of the number of cycles on the wavelength offset.

[0088] The specific calculations of this invention include determining the linear relationship between temperature change and wavelength shift based on the FCM algorithm, and determining the nonlinear relationship between residual wavelength shift and battery factors based on the residual wavelength shift established by the EMD algorithm.

[0089] Among them, such as Figure 1 As shown, the specific calculations of the FCM algorithm include:

[0090] (1) Collect the first wavelength offset corresponding to different temperature values ​​of the fiber optic temperature sensor and construct the first sample set;

[0091] (2): Set the condition change value for the termination of the iteration, randomly generate the initial cluster center P0, and set the current iteration number t = 0;

[0092] (3): Calculate the membership degree using the membership function matrix formula;

[0093] (4): Calculate the cluster centers using the cluster center matrix formula;

[0094] (5): If the objective function is less than or equal to the conditional change value at the end of the iteration, stop the iteration calculation and output the final membership degree and cluster center. If the objective function is greater than the conditional change value at the end of the iteration, let t = t + 1 and repeat steps (2) to (4) to continue the iteration calculation.

[0095] like Figure 2 As shown, the residual wavelength offset established based on the EMD algorithm includes the following steps:

[0096] (1) Determine the functional relationship between wavelength offset and cycle period using linear fitting, as shown below:

[0097] Z c (t)=r1t 4 +r2t 3 +r3t 2 +r4t+r5

[0098] Where t represents the battery cycle period, and r1-r5 are the fitting coefficients.

[0099] (2) Determine the functional relationship between wavelength offset and SOC, and obtain the polynomial equation by polynomial fitting:

[0100]

[0101] (3) Regarding Z n The correlation coefficient with Z was obtained using the EMD algorithm:

[0102] (4) Perform EMD decomposition on the Z(t) signal, setting i=1, n=1, r0=Z;

[0103] (5) Calculate all minimum and maximum points in the sequence h(t), fit the upper and lower envelopes using cubic spline functions, and calculate the average curve m(t) of the upper and lower envelopes.

[0104] (6) Calculate the difference between the data sequence h(t) and the average envelope m(t) to obtain a new data sequence h(t);

[0105] (7) Repeat steps (5) and (6) until the conditions of IMF are met. The resulting function is the first-order intrinsic mode function component imf(t).

[0106] The influence of each factor on the wavelength shift was calculated using the two models described above. All factors were then superimposed to generate a calibration relationship between the temperature measurement and the wavelength shift, as follows:

[0107] T FBG =T0+f -1 ((λ-λ0)-(Z s +Z c ))

[0108] Where T FBG T0 is the measured temperature, T0 is the compensated temperature, λ represents the measured peak wavelength of the fiber optic temperature sensor, λ0 is the characteristic wavelength of the fiber optic temperature sensor, and Z is the measured temperature. s The effect of SOC variation on wavelength shift; Z c This represents the effect of the number of cycles on the wavelength offset.

[0109] Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to this invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for accurate estimation of battery temperature based on fiber optic sensing, characterized in that, include: S1: Collect the first wavelength offset corresponding to different temperature values ​​of the fiber optic temperature sensor and construct the first sample set; S2: Establish a first model based on the FCM algorithm, set the conditional change value for the termination of the iteration, train the first model using the first sample set, and obtain the linear relationship between temperature change and first wavelength offset through repeated iterations; S3: During the battery charging and discharging process, collect battery monitoring data and the corresponding second wavelength offset. The battery monitoring data includes at least the battery SOC value and the number of cycles to construct a second sample set. S4: Establish a second model based on the EMD algorithm, set the conditional change value for the termination of the iteration, train the second model using the second sample set, and obtain the nonlinear relationship between the changes in battery monitoring data and the second wavelength offset through repeated iterations; S5: The linear relationship between the temperature change and the first wavelength offset obtained in step S2 and the nonlinear relationship between the battery monitoring data change and the second wavelength offset obtained in step S4 are superimposed to generate a calibration formula between the temperature measurement value and the wavelength offset. In step S2, the objective function designed based on the FCM algorithm is shown in the following equation: The constraints are as follows: Where N is the number of input samples, P is the cluster center matrix, and U is the membership function matrix. Let be the membership degree of sample j belonging to the i-th class, c be the number of clusters, and m be the fuzzy index. For sample values, As cluster center; The specific steps of step S4 include: S4.1: The functional relationship between wavelength offset and cycle period is calculated using linear fitting. S4.2: The functional relationship between wavelength shift and battery SOC value is calculated using polynomial fitting. S4.3: Based on the EMD algorithm, the noise proportion is calculated. The acquired second wavelength offset is decomposed by EMD. All maximum and minimum points in the sequence are calculated. The upper and lower envelopes are fitted with cubic spline functions. The average curves of the upper and lower envelopes are calculated. The difference between the data sequence and the average envelope is used to calculate the new data sequence. It is determined whether the IMF condition is met. If not, the residual component is used as the original signal. The iteration is repeated until the condition is met. The resulting function is the first-order intrinsic mode function component. The relationship between wavelength shift and cycle period is: Where t represents the battery cycle period, and r1-r5 are the fitting coefficients; The relationship between wavelength shift and battery SOC value is as follows: in, The functional relationship representing the charging state. The function represents the state of discharge, where s represents the SOC value of the battery, and p1-p5 and q1-q5 are both fitting coefficients. In step S5, the calibration relationship between the generated temperature measurement value and the wavelength offset is as follows: in For actual measured temperature, To compensate for temperature, This indicates the peak wavelength measured by the fiber optic temperature sensor. The characteristic wavelength of the fiber optic temperature sensor. The effect of SOC variation on wavelength shift; This represents the effect of the number of cycles on the wavelength offset.

2. The method for accurate battery temperature estimation based on fiber optic sensing according to claim 1, characterized in that, In step S2, the FCM model is optimized using the Lagrange multiplier method to obtain iterative formulas, specifically including the membership function matrix formula and the cluster center matrix formula.

3. The method for accurate battery temperature estimation based on fiber optic sensing according to claim 2, characterized in that, The formula for the membership function matrix is: Where t is the number of iterations. For the sample To the cluster center distance, For the sample To the cluster center The distance.

4. The method for accurate battery temperature estimation based on fiber optic sensing according to claim 3, characterized in that, The formula for the cluster center matrix is: in, As cluster center, Let be the membership degree of sample j belonging to the i-th class. These are sample values.

5. The method for accurate battery temperature estimation based on fiber optic sensing according to claim 4, characterized in that, The specific steps of the repeated iterations include: S2.1: Set the condition change value for iteration termination, randomly generate the initial cluster center P0, and set the current iteration number t=0; S2.2: Calculate the membership degree using the membership function matrix formula; S2.3: Calculate the cluster centers using the cluster center matrix formula; S2.4: If the objective function is less than or equal to the conditional change value at the end of the iteration, stop the iteration calculation and output the final membership degree and cluster center. If the objective function is greater than the conditional change value at the end of the iteration, let t=t+1 and repeat steps S2.2 to S2.4 to continue the iteration calculation.

Images