Adaptive floating charge voltage control method for storage battery and related equipment

Abstract

The embodiment of the invention discloses a storage battery self-adaptive floating charge voltage control method and related equipment, and the method comprises the steps: carrying out the modeling of a multi-dimensional and nonlinear training data set which comprises a plurality of historical feature vectors and a plurality of historical floating charge voltage values, thereby capturing a complex mapping relation between the state of a battery body and the floating charge voltage, and carrying out the calculation of the floating charge voltage. The appropriate floating charge voltage is predicted according to the real-time feature vector, so that the method can adapt to the characteristic deviation of the battery body caused by environmental parameter change, battery aging and the like in the actual use process of the battery body, and the floating charge voltage within an appropriate range matched with the current state of the battery body is dynamically obtained; therefore, the battery body can obtain appropriate floating charge voltage for floating charge at any time, the possibility of overcharge or undercharge of the battery body can be reduced, the aging of the battery body can be delayed, the service life of the battery body can be prolonged, and the use safety of the battery body can be guaranteed. The method is widely applied to the technical field of battery management.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of battery management, and particularly to a storage battery adaptive float voltage control method and related equipment. BACKGROUND

[0002] In communication machine rooms, base stations and other occasions, in order to guarantee the reliability of power supply, storage batteries or storage battery groups are usually used as backup power sources for power supply. In order to guarantee that the storage battery is always available, the storage battery needs to be floated. Since the float voltage will affect the service life and performance of the storage battery, a suitable float voltage needs to be used.

[0003] Related float voltage control technologies usually use fixed voltage or set the float voltage based on the experience of operation and maintenance personnel. For example, a fixed size of 2.15V is used for each storage battery, or a size of 2.17V-2.25V is used for storage batteries with lead-calcium alloy grids, and a size of 2.15V-2.17V is used for storage batteries with lead-antimony alloy grids. However, during the use of the storage battery, changes in the use environment and aging of the storage battery itself will cause the characteristics of the storage battery to deviate, thereby causing the suitable range of the float voltage to change. The fixed voltage or the experience of the operation and maintenance personnel cannot adapt to such changes, so the related float voltage control technology can only set a suitable float voltage in the initial short time. After that, due to the change in the suitable range of the float voltage caused by the deviation of the characteristics of the storage battery, the float voltage deviates from the suitable range when the storage battery is floated, so that the storage battery is in an overcharged or undercharged state for a long time, which accelerates the aging of the storage battery, and thus the capacity attenuation and even the heating and fire accidents are more likely to occur. SUMMARY

[0004] To solve at least one of the above technical problems, the purpose of an embodiment of the present application is to provide a storage battery adaptive float voltage control method based on Gaussian process regression and related equipment.

[0005] In one aspect, an embodiment of the present application includes a storage battery adaptive float voltage control method based on Gaussian process regression, which comprises: obtaining a training data set; the training data set includes a plurality of historical feature vectors and a plurality of historical float voltage values measured from a battery body; establishing a Gaussian process regression model according to the training data set; detecting a real-time feature vector from the battery body; processing the real-time feature vector using the Gaussian process regression model to obtain a float voltage prediction value of the battery body; According to the float voltage prediction value, the float voltage of the battery body is controlled.

[0006] Further, the training data set is obtained, including: An initial training data set is established. A sliding window mechanism is used to periodically update the training data set online.

[0007] Further, the Gaussian process regression model is established according to the training data set, including: The hyperparameters are determined according to the historical feature vectors and the historical float voltage values. According to the hyperparameters and the historical feature vectors, the kernel function in the Gaussian process regression model is determined; the kernel function has the form of a radial basis function.

[0008] Further, the hyperparameters are determined according to the historical feature vectors and the historical float voltage values, including: The marginal likelihood function is established according to the historical feature vectors and the historical float voltage values. The length scale, signal variance and noise variance that maximize the marginal likelihood function are solved by using the conjugate gradient method. The length scale, signal variance and noise variance solved are used as the hyperparameters.

[0009] Further, the kernel function in the Gaussian process regression model is determined according to the hyperparameters and the historical feature vectors, including: The kernel function is established according to the formula

[0010] The kernel function is established; wherein, is the kernel function, is the i-th historical feature vector in the training data set, is the i-th historical feature vector in the training data set, is the length scale, is the signal variance, is the noise variance, is the Kronecker function. Further, the real-time feature vector is processed using the Gaussian process regression model to obtain the float voltage prediction value of the battery body, including: The float voltage prediction value is obtained according to the formula

[0011]

[0012] ​​performing the calculation of the float voltage prediction value; wherein, , , is the real-time feature vector, , , is each of the historical feature vectors, is each of the historical float voltage values, denotes matrix transposition, is a unit vector, is the float voltage prediction value.

[0013] Further, the controlling the float voltage of the battery according to the float voltage prediction value comprises: obtaining a confidence interval of the float voltage prediction value; when the confidence interval is within a preset safety range, setting the float voltage of the battery to be equal to the float voltage prediction value; when the confidence interval is outside the preset safety range, obtaining a float voltage empirical compensation value, and setting the float voltage of the battery to be equal to the float voltage empirical compensation value.

[0014] Further, the obtaining the confidence interval of the float voltage prediction value comprises: calculating the variance of the float voltage prediction value according to the formula

[0015] is the variance; calculating the confidence interval according to the formula

[0016] wherein p is a positive integer.

[0017] In another aspect, an embodiment of the present application further includes a computer device comprising a memory and a processor, the memory being configured to store at least one program, and the processor being configured to load the at least one program to perform the adaptive float voltage control method for storage battery based on Gaussian process regression in the embodiment.

[0018] In another aspect, an embodiment of the present application further includes a computer program product comprising a computer program, the computer program being configured to implement the adaptive float voltage control method for storage battery based on Gaussian process regression in the embodiment when executed by a processor.

[0019] ​The embodiment has the following beneficial effects: the battery adaptive floating charging voltage control method based on Gaussian process regression and the related device in the embodiment model the multi-dimensional and nonlinear training data set containing multiple historical feature vectors and multiple historical floating charging voltage values, capture the complex mapping relationship between the state of the battery body and the floating charging voltage, and predict the appropriate floating charging voltage according to the real-time feature vector, i.e., the real-time state of the battery body, so as to adapt to the characteristic deviation of the battery body caused by the change of environmental parameters and battery aging in the actual use process of the battery body, dynamically obtain the floating charging voltage in the appropriate range adapted to the current state of the battery body, and make the battery body obtain the appropriate floating charging voltage at any time, which is beneficial to reducing the possibility of overcharging or undercharging of the battery body, delaying the aging of the battery body, prolonging the service life of the battery body, reducing the possibility of capacity attenuation or fire accidents, and ensuring the use safety of the battery body. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A schematic diagram of a battery power supply system to which the battery adaptive floating charging voltage control method based on Gaussian process regression in the embodiment can be applied; Figure 2 A schematic diagram of the internal structure of a battery body in the battery power supply system shown in FIG. 1; Figure 1 Figure 3 A schematic diagram of a first working mode of the battery power supply system shown in FIG. 1; Figure 1 Figure 4 A schematic diagram of a second working mode of the battery power supply system shown in FIG. 1; Figure 1 Figure 5 A schematic diagram of a third working mode of the battery power supply system shown in FIG. 1; Figure 1 Figure 6 A schematic diagram of steps of the battery adaptive floating charging voltage control method based on Gaussian process regression in the embodiment; Figure 7 A schematic diagram of the principle of the sliding window mechanism in the embodiment; Figure 8 A schematic diagram of the specific flow of step S5 in the embodiment; Figure 9 A schematic diagram of the structure of the computer device in the embodiment; Figure 10 A schematic diagram of the principle of the computer readable storage medium in the embodiment. DETAILED DESCRIPTION

[0021] ​​​​Before the embodiments of the present application are described in detail, first, some of the terms and terminology involved in the embodiments of the present application are explained, and the terms and terminology involved in the embodiments of the present application are applicable to the following explanations: Battery: a device capable of converting chemical energy and electrical energy; when charging, electrical energy is converted into chemical energy between internal active substances, and when discharging, chemical energy is converted into electrical energy through chemical reactions between internal active substances, also known as secondary battery; according to the type of internal active substance, it is divided into lead-acid battery, lithium battery, etc.; Lead-acid battery: a battery using lead dioxide material as the positive electrode, using sponge metal lead material as the negative electrode, and using aqueous sulfuric acid solution as the electrolyte, which has the advantages of high working voltage, wide temperature range, good high and low rate discharge performance, abundant raw material source, and low price; Valve-regulated lead-acid battery: a special lead-acid battery, characterized by being provided with a one-way exhaust valve (also called safety valve) that communicates the inside and outside of the battery, which automatically opens when the amount of gas inside the battery exceeds a certain value, thereby exhausting the gas generated by the chemical reaction inside the lead-acid battery, and then automatically closes the valve to prevent air from entering the inside of the battery, so it will not leak acid or emit acid mist, and it does not need to be maintained by adding acid and water during use; Battery pack: a single battery is called a battery monomer, since the discharge current and discharge voltage of the battery monomer are limited, in order to obtain sufficient discharge current and discharge voltage, a plurality of battery monomers are usually connected in series and parallel to form a whole, i.e. a battery pack; Float charging: a working mode in which the battery monomer or battery pack is connected in parallel with the power supply circuit to the load circuit, a constant voltage slightly higher than the terminal voltage of the battery is applied, and the power supply circuit continuously provides compensation current, which is mainly used to compensate for self-discharge loss, share part of the DC load, and maintain the full charge state of the battery, and only in the case of sudden load increase or power failure, the battery provides large current output; this working mode can extend the life of the battery by 30%-50%, and is widely used in communication base stations, substations, and uninterruptible power supply systems; Float charging voltage: the voltage applied by the charging device to the positive and negative electrodes of the battery monomer or battery pack when the battery is being floated; if the float charging voltage is too high, it will easily cause decomposition and loss of the electrolyte inside the battery, and if the float charging voltage is too low, it will easily cause undercharging and lead to capacity decay of the battery, so the float charging voltage needs to be controlled within an appropriate range; the appropriate range of float charging voltage control is usually related to factors such as battery type and ambient temperature; Adaptive float charging voltage control: a technology that dynamically calculates the appropriate range of float charging voltage according to changes in factors that can affect the appropriate range of float charging voltage, and controls the float charging voltage of the battery within the appropriate range; Gaussian Process Regression, GPR, is a non-parametric model that uses Gaussian Process (GP) prior to regress data; Kronecker delta, also known as Kronecker function, Kronecker is an inner chain binary function, named after German mathematician Leopold Kronecker, and the independent variable (input value) of the Kronecker function is usually two integers. If the two are equal, the output value is 1, otherwise 0; Identity matrix: In the multiplication of matrices, there is a matrix that plays a special role, like 1 in the multiplication of numbers. This matrix is called the identity matrix. The identity matrix is a square matrix, and the elements on the diagonal from the top left to the bottom right (called the main diagonal) of the identity matrix are all 1, and the elements other than this are all 0.

[0022] It is difficult for fixed voltage or dynamic compensation strategies based on empirical formulas to accurately adapt to the characteristic deviation of the battery caused by aging, temperature fluctuations, internal resistance changes and other complex factors. Especially for valve-regulated lead-acid battery packs used in series, the differences between individual batteries will cause some batteries to be in overcharge or undercharge for a long time, accelerating capacity decay. Although methods based on linear regression or neural networks can partially solve this problem, they have the defects of poor model generalization, high risk of overfitting under small sample data, and high computational complexity.

[0023] As a non-parametric Bayesian model, Gaussian Process Regression can flexibly describe the non-linear relationship between battery parameters through kernel functions and quantify the uncertainty of the prediction results, making it suitable for dynamic prediction of floating voltage under complex working conditions.

[0024] Based on the above principles, in the present embodiment, a battery adaptive floating voltage control method based on Gaussian Process Regression is provided.

[0025] In the present embodiment, the battery adaptive floating voltage control method based on Gaussian Process Regression can be applied to Figure 1 the battery power supply system shown in the figure.

[0026] Referring to Figure 1The battery power supply system includes a power supply, a load, a battery management system, and a battery body. The power supply can be a commercial power supply or a self-provided generator set and related voltage conversion, voltage stabilization, and rectification devices. The load includes communication room equipment, communication base station equipment, lighting, air conditioning, security systems, and other electrical equipment. The battery body can be a battery monomer or a battery pack. The battery management system can detect parameters of the battery body and manage the charging and discharging of the battery body. For example, the battery management system can control the current flow direction of the battery body, i.e., the current from the power supply flows into the battery body, thereby charging the battery body. The battery management system connects the battery body and the load, so that the battery body discharges to supply power to the load. Figure 1 The battery management system and the battery body can be regarded as a whole, and the whole is in parallel with the power supply relative to the load.

[0027] In this embodiment, Figure 1 the internal structure of the battery body in the battery power supply system is as shown in Figure 2 . Specifically, the battery body can be a battery monomer as shown in Figure 2 (a), a battery pack composed of multiple battery monomers connected in series as shown in Figure 2 (b), a battery pack composed of multiple battery monomers connected in parallel as shown in Figure 2 (c), or a structure in which multiple battery monomers are connected in series and then in parallel. In this embodiment, when a battery pack is used, it can be equivalent to a battery monomer, and the overall performance of the battery body can also be represented by the parameters of the battery monomer, such as discharge voltage (terminal voltage), discharge current, working temperature, and internal resistance.

[0028] In this embodiment, several working modes of the battery power supply system are as shown in Figure 3 , Figure 4 and Figure 5 .

[0029] Figure 3 The first working mode of the battery power supply system is shown in Figure 3 . In this working mode, the power supply and the load are both in normal state, i.e., the power supply has sufficient power supply capacity, and the power consumption of the load is within the rated range. At this time, the power supply completely bears the power supply to the load, i.e., the power supply current received by the load is completely output by the power supply. At the same time, the power supply also outputs power to the battery management system, thereby charging the battery body. At this time, the charging voltage of the battery body, i.e., the float voltage, is slightly higher than the terminal voltage of the battery body, thereby forming a small current input to the battery body for floating charging.

[0030] Figure 4 The second working mode of the battery power supply system is shown in .Figure 4 In this working mode, the power supply is in normal state, and the load is in load-increase state, that is, the power supply has enough power supply power, and the power consumption of the load is greater than the rated range. At this time, the power supply bears part of the power supply for the load, and at the same time, the battery management system also controls the battery body to discharge, and the discharge current of the battery body is delivered to the load, that is, the battery body bears another part of the power supply for the load, thereby forming a supplement to the power supply.

[0031] Figure 5 The third working mode is for the battery power supply system. Referring to Figure 5 In this working mode, the power supply is in a fault state, and the load is in a normal state or a load-increase state, that is, the power supply cannot output enough power supply power (usually, in order to ensure safety, the power supply will be suspended, that is, the power supply does not output current), at this time, the battery management system controls the battery body to discharge, and the discharge current of the battery body is delivered to the load, that is, the battery body bears the power supply for the load.

[0032] In this embodiment, the battery adaptive floating voltage control method based on Gaussian process regression applied to the battery power supply system is as shown in Figure 6 The method comprises the following steps: S1. Obtain a training data set; S2. Establish a Gaussian process regression model according to the training data set; S3. Obtain a real-time feature vector from the battery body; S4. Process the real-time feature vector using the Gaussian process regression model to obtain a floating voltage prediction value of the battery body; S5. Control the floating voltage of the battery body according to the floating voltage prediction value.

[0033] In this embodiment, each step S1-S5 in the battery adaptive floating voltage control method based on Gaussian process regression can be performed by the battery management system in the battery power supply system. The battery management system has the performance of data acquisition, data processing and control. When the battery management system performs some steps, the battery management system can process the data detected from the external corresponding components.

[0034] In this embodiment, one round of steps S1-S5 can be performed in a time period When the time period ends and enters the next time period , a new round of steps S1-S5 is performed, and the cycle is repeated to realize dynamic control of the floating voltage of the battery body. The time period The length of the time period can be 10 min, 1 min, etc. a longer value, so that the battery management system has enough time to complete a round of steps S1-S5; in the case of sufficient performance of the battery management system, the length of the time period can be set to 10 s, 1 s, etc. a shorter value, so as to realize real-time control of the float voltage of the battery body. In the present embodiment, a round of steps S1-S5 performed in the time period is taken as an example for description.

[0035] In step S1, the battery management system obtains a plurality of historical feature vectors and a plurality of historical float voltage values by detecting the battery body.

[0036] Specifically, the content of the historical feature vector can be a feature vector describing the battery body before the execution of steps S1-S5 (i.e. before the time period ). For example, a historical feature vector includes parameters such as terminal voltage , charging current , discharging current , working temperature , internal resistance , number of cycles , cumulative charge-discharge capacity , etc., wherein the terminal voltage represents the terminal voltage of the battery body at time , the charging current represents the charging current of the battery body at time , the discharging current represents the discharging current of the battery body at time , the working temperature represents the working temperature of the battery body at time , the internal resistance represents the internal resistance of the battery body at time , the number of cycles represents the number of complete charge-discharge cycles of the battery body from a certain reference time (e.g. initial activation or re-commissioning after maintenance) to time , and the cumulative charge-discharge capacity represents the cumulative charge-discharge capacity completed by the battery body from a certain reference time (e.g. initial activation or re-commissioning after maintenance) to time . The parameters at the same time constitute a feature vector (V1, V2, V3, V4, V5, V6, V7), and when the time , , , , , , , the battery management system obtains a plurality of historical feature vectors and a plurality of historical float voltage values by detecting the battery body. ​ If the time point t is before the time point t0, then the feature vector becomes a historical feature vector, and if the time point t is after the time point t0, then the feature vector becomes a current feature vector. Taking different time points, a plurality of historical feature vectors can be obtained.

[0037] Specifically, the content of the historical float charging voltage value can be a float charging voltage used by the battery management system to actually float the battery body before the execution of steps S1-S5 (i.e., the time period before t0). For example, the battery management system can record the size of the float charging voltage each time the float charging voltage used to actually float the battery body is updated, and the time point t0 is recorded as the time point at which the float charging voltage is updated. The historical float charging voltage value used at the time point t0 is recorded as If the time point t is before the time point t0, then the feature vector becomes a historical feature vector, and if the time point t is after the time point t0, then the feature vector becomes a current feature vector. Taking different time points, a plurality of historical float charging voltage values can be obtained.

[0038] In this embodiment, a sliding window mechanism can be used to update the training data set online, as shown in FIG. 2. Figure 7 Specifically, an initial training data set with a size of n can be established at the beginning (e.g., the time period before t0), which contains n historical feature vectors and n historical float charging voltage values, and in the subsequent time period t0, new feature vectors and new float charging voltage values are detected, the new feature vectors and the new float charging voltage values are added to the training data set, and the same number of earliest detected feature vectors and float charging voltage values in the training data set are deleted, so that the size of the training data set is always maintained as n, thereby realizing online updating of the training data set. By updating the training data set online, the dynamic tracking capability of the Gaussian process regression model to the battery body usage environment and the battery body aging process can be maintained, which is beneficial to making the float charging voltage prediction value calculated by each time period execution step S4 close to the appropriate range of the float charging voltage under the current usage environment.

[0039] In this embodiment, the historical feature vectors in the training data set (x1, x2, x3, x4, x5, x6, x7) can be represented as

[0040] In this embodiment, the historical feature vectors in the training data set (x1, x2, x3, x4, x5, x6, x7) can be represented as ​​​​​​​​​​​​​​​​). Higher dimensional history feature vectors can also be used, i.e. the history feature vector can also contain more data describing the state of the battery body. Lower dimensional history feature vectors can also be used, e.g. only or are used as the history feature vector. Using higher dimensional history feature vectors is advantageous for more finely utilizing the state data of the battery body, thus more finely calculating the floating voltage prediction value. Using lower dimensional history feature vectors is advantageous for reducing the data processing amount and lowering the performance requirement on the battery management system.

[0041] For the history feature vector = ( , , , , , , ) and the history floating voltage value in the training data set, the original detected data can be used, i.e. , , ... are the values actually detected at time . Such values will have dimensions, e.g. has the dimension of volt, and have the dimension of ampere, etc.

[0042] In this embodiment, the history feature vector = ( , , , , , , ) and the history floating voltage value can be normalized. For example, each data can be normalized by the following formula:

[0043] wherein represents any type of data , , , , , , and corresponding to a specific time , represents the value of the data at the time of determining the specific type of data represented (for example ), the mean value calculated from a plurality of data (for example , … ) obtained by taking different values according to the time , represents the standard deviation calculated from a plurality of data obtained by taking different values according to the time , . represents the result after normalization processing is performed on .

[0044] After normalization processing is performed on , the is replaced by the corresponding in the training data set.

[0045] By performing standardization processing, the parameter difference between different types of data can be eliminated, thereby facilitating large-scale data processing and improving the speed of data processing.

[0046] In this embodiment, when performing step S2, that is, establishing a Gaussian process regression model according to the training data set, the following steps can be performed: S201. Determine the hyperparameters according to the historical feature vectors and the historical floating voltage values; S202. Determine the kernel function in the Gaussian process regression model according to the hyperparameters and the historical feature vectors.

[0047] In this embodiment, by performing steps S201-S202, the kernel function can be established as the basis of the Gaussian process regression model to be used.

[0048] In step S201, the types of hyperparameters to be used include length scale , signal variance , and noise variance . When setting the hyperparameters, the initial values of each of the length scale , signal variance , and noise variance can be set, for example, the length scale , signal variance , and noise variance are set as fixed values or random values.

[0049] In this embodiment, the meaning of the signal variance is that the historical feature vectors = ( , , , , , , ) and the variance of noise contained in each historical float voltage value The meaning of the signal variance is the variance of each historical feature vector = ( , , , , , , ) and the variance of noise contained in each historical float voltage value The length scale will affect the exponent in the exponential function in the kernel function, thereby affecting the size of the kernel function, so the value of the length scale can be selected to limit the size of the kernel function within an appropriate range, avoiding the consequences of data overflow and the like caused by excessively large values of the kernel function.

[0050] When performing step S201, a step of hyperparameter optimization can be performed to determine the values of the length scale , the signal variance , and the noise variance .

[0051] In this embodiment, when performing step S201, the hyperparameter optimization step performed includes the following steps: S20101. Establish an edge likelihood function based on each historical feature vector and each historical float voltage value; S20102. Use the conjugate gradient method to solve the length scale, signal variance, and noise variance that maximize the edge likelihood function; S20103. Use the solved length scale, signal variance, and noise variance as hyperparameters.

[0052] In step S20101, an edge likelihood function is established based on each historical feature vector = ( , , , , , , ) and each historical float voltage value . In this embodiment, the edge likelihood function has the following form:

[0053] In step S20102, the edge likelihood function Maximizing the desired length is used as the optimization objective, and the conjugate gradient method is employed for iterative solution to determine the length scale. Signal variance and noise variance The respective values ​​are used to obtain the optimal hyperparameters. ={ , , }

[0054] Determining the optimal hyperparameters ={ , , After that, in step S202, the optimal hyperparameters are used. ={ , , } and each historical feature vector =( , , , , , , Establish the kernel function in the Gaussian process regression model.

[0055] In step S202, the radial basis function (RBF) is selected as the covariance function to establish the kernel function in the Gaussian process regression model, that is, the established kernel function has the form of a radial basis function.

[0056] Specifically, in this embodiment, when performing step S202, according to the formula

[0057] Create the kernel function.

[0058] in, For kernel function, For the first in the training dataset A historical feature vector, For the first in the training dataset A historical feature vector, namely and These represent the index numbers of the historical feature vectors in the training dataset, and and They can take independent values ​​and may be equal. It is an exponential function, where The length scale set in step S201 The signal variance set in step S201, The noise variance is set in step S201.

[0059] Kernel function middle, Let Kronecker function be the Kronecker function. Specifically, the Kronecker function is defined as follows:

[0060] In this embodiment, the kernel function established in step S202 can be expressed in matrix form, and since the kernel function established in step S202 is essentially a covariance function, it can also be called a covariance matrix. It can represent the th element in the covariance matrix. line, number The numerical values ​​of the elements in the column, i.e. and They can also represent the row and column indices in the covariance matrix, respectively.

[0061] In this embodiment, since the covariance matrix is ​​established based on historical feature vectors in the training dataset, its size is related to the size of the training dataset. For example, the training dataset in this embodiment contains n historical feature vectors, and a sliding window mechanism is used to periodically update the training dataset online, ensuring that the training dataset always contains n historical feature vectors; that is, the size of the training dataset is n. , =1,2,3……n, i.e., the covariance matrix Specifically, it can be represented as an n x n matrix.

[0062] In this embodiment, the kernel function established in step S202 is executed, and the Kronecker function is used. Modeling of the noise term was achieved. Specifically, due to the Kronecker function... Having "if the index number of the historical feature vector" and If they are equal, then the Kronecker function is... The output value is 1; otherwise, if the index number of the historical feature vector is... and If they are not equal, then the Kronecker function The output value is 0, therefore, the historical feature vector index number is 0 if and only if the output value is 0. and When they are equal, the noise variance Only then will it be added to the covariance matrix. On the diagonal elements, that is, the covariance matrix Bank of China index number and column index number Those elements that are equal. That is, the covariance matrix. The diagonal elements of the covariance matrix = ( , , , , , , ) represent the independent observation noise of each data point [including each historical feature vector The non-diagonal elements of the covariance matrix , i.e. those elements whose row index number and column index number are not equal, do not reflect the independent observation noise of the data points.

[0063] After the step S2 of establishing the kernel function, i.e. the covariance matrix , is performed, the step S3 of obtaining the real-time feature vector from the battery detection can be performed.

[0064] In the embodiment, the historical feature vectors = ( , , , , , , ) and the historical float voltage values obtained in the step S1 in the time period represent the state of the battery before the time period . Accordingly, the real-time feature vector obtained in the step S3 in the time period represents the current state of the battery in the time period .

[0065] In the embodiment, the form of the real-time feature vector to be detected is the same as that of any historical feature vector. For example, the real-time feature vector also includes the terminal voltage , the charging current , the discharging current , the working temperature , the internal resistance , the number of cycles , the cumulative charge and discharge capacity , etc. The terminal voltage represents the terminal voltage of the battery in the time period , the charging current represents the charging current of the battery in the time period , the discharging current represents the discharging current of the battery in the time period , and the working temperature represents the working temperature of the battery in the time period .represents the battery body at a time period current working temperature, internal resistance represents the battery body at a time period current internal resistance, number of cycles represents the battery body at a certain reference time (e.g. initial activation or re-commissioning after maintenance) to a time period current number of complete charge-discharge cycles, cumulative charge-discharge capacity represents the battery body at a certain reference time (e.g. initial activation or re-commissioning after maintenance) to a time period current cumulative completed charge-discharge capacity.

[0066] Therefore, the real-time feature vector and any historical feature vector can be vectors of the same dimension, and the real-time feature vector can also be marked as = (x1, x2, x3, x4, x5, x6, x7) , , , , , , ) in the form, the difference is that the time in the real-time feature vector is equivalent to taking the time in the time period , and the time in the historical feature vector is equivalent to taking the time before the time period .

[0067] After the battery management system obtains the kernel function, i.e. the covariance matrix in step S2 and obtains the real-time feature vector in step S3, it can perform step S4, processing the real-time feature vector using the kernel function, i.e. the covariance matrix to obtain the float voltage prediction value of the battery body .That is, the Gaussian process regression model in this embodiment is equivalent to the process of processing the real-time feature vector using the kernel function, i.e. the covariance matrix .

[0068] In this embodiment, when the battery management system performs step S4, i.e. processes the real-time feature vector using the Gaussian process regression model to obtain the float voltage prediction value of the battery body, it can specifically perform the following steps: S401. Calculate according to the formula

[0069] . Wherein, , , is a real-time feature vector, , , is each historical feature vector, is each historical float voltage value, denotes matrix transpose, is a unit vector, is noise variance, is a kernel function, is a float voltage prediction value.

[0070] In step S401, means a matrix composed of the covariance of the real-time feature vector with all the historical feature vectors in the training data set, which contains , , and so on n rows. Taking for example, the vector can be calculated by letting the kernel function, i.e. the covariance matrix , and letting , , and the vectors , , and so on can be calculated based on the same principle, thereby composing the matrix .

[0071] In step S401, denotes the transpose matrix of the matrix .

[0072] In step S401, is the kernel function, i.e. the covariance matrix itself established in step S202, i.e. .

[0073] In step S401, is a unit vector, i.e. is a matrix of the same size as the kernel function, i.e. the covariance matrix (e.g. in the present embodiment, both are n-row, n-column square matrices), and the unit matrix has elements of 1 on the diagonal from the top-left corner to the bottom-right corner (referred to as the main diagonal) and elements of 0 elsewhere.

[0074] In step S401, , i.e. is composed of each historical float voltage value ​​vector of components.

[0075] By performing step S401, the kernel function, i.e., the covariance matrix is utilized to process the real-time feature vector , so as to obtain the float voltage prediction value predicted by running the Gaussian process regression model. .

[0076] In this embodiment, the float voltage prediction value means that the Gaussian process regression model (GPR) is used to model the multi-dimensional and nonlinear training data set containing multiple historical feature vectors and multiple historical float voltage values, so as to capture the complex mapping relationship between the state of the battery body and the float voltage, and then predict the appropriate float voltage according to the real-time feature vector, i.e., the real-time state of the battery body. Compared with related technologies such as using a fixed float voltage or a float voltage set based on the experience of operation and maintenance personnel, the float voltage prediction value obtained in this way has a greater possibility of being in the appropriate range adapted to the current state of the battery body. Moreover, since the training data set is obtained by actually measuring the battery body, the sliding window mechanism can also be used to update the training data set online, so that the process of obtaining the float voltage prediction value is dynamic, thereby being able to adapt to the characteristic drift of the battery body caused by changes in environmental parameters and battery aging during actual use of the battery body, so as to dynamically obtain the float voltage in the appropriate range adapted to the current state of the battery body, so that the battery body can obtain the appropriate float voltage at any time, which is conducive to reducing the possibility of overcharging or undercharging of the battery body, delaying the aging of the battery body, prolonging the service life of the battery body, reducing the possibility of capacity attenuation or fire accidents, and ensuring the safe use of the battery body.

[0077] In this embodiment, when the battery management system performs step S5, i.e., controls the float voltage of the battery body according to the float voltage prediction value, the battery management system can directly set the float voltage of the battery body to be equal to the float voltage prediction value , that is, use the float voltage equal to the float voltage prediction value to float the battery body within the current time period and before the next time period . In this case, the battery management system applies a voltage of to the positive and negative electrodes of the battery body, i.e., the battery management system is a voltage source relative to the battery body. Since the float voltage prediction value better adapts to the characteristic drift of the battery body, the float voltage prediction value is in the appropriate range adapted to the current state of the battery body, so that the battery body can be floated in the time period Obtaining a proper float voltage, thereby reducing the possibility of overcharging or undercharging the battery body.

[0078] In this embodiment, when the battery management system performs step S5, i.e., the step of controlling the float voltage of the battery body according to the float voltage prediction value, the following steps can also be performed as shown in the following: Figure 8 S501. Obtain a confidence interval of the float voltage prediction value; S502. When the confidence interval is within a preset safety range, set the float voltage of the battery body to be equal to the float voltage prediction value; S503. When the confidence interval is outside the preset safety range, obtain a float voltage empirical compensation value, and set the float voltage of the battery body to be equal to the float voltage empirical compensation value.

[0079] The process of steps S501-S503 is as follows: after the float voltage prediction value is calculated by performing steps S1-S4, the float voltage of the battery body is not directly set to be equal to the float voltage prediction value but the confidence interval of the float voltage prediction value is first calculated, and whether the float voltage of the battery body is set to be equal to the float voltage prediction value is determined according to the confidence interval of the float voltage prediction value .

[0080] Specifically, when step S501, i.e., the step of obtaining the confidence interval of the float voltage prediction value, is performed, the following steps can be performed: S50101. Calculate the variance of the float voltage prediction value according to the formula

[0081] ; wherein, is the variance; S50102. Calculate the confidence interval according to the formula

[0082] , wherein p is a positive integer.

[0083] The formula used in step S50101 is also equivalent to part of the Gaussian process regression model. In the formula used in step S50101, the meanings of various symbols are the same as those of the formula used in step S401, i.e., can be calculated by setting the kernel function, i.e., the covariance matrix , as , and setting , , ,​ is a transpose matrix of is a unit vector, is a noise variance.

[0084] By performing step S50101, the calculated variance of the float voltage prediction value obtained in steps S1-S4, which quantitatively indicates the size of the range in which the float voltage prediction value may be distributed.

[0085] After the float voltage prediction value obtained in steps S1-S4 and the variance of the float voltage prediction value obtained in step S50101 are calculated, then step S50102 is performed, and the confidence interval of the float voltage prediction value is calculated according to the formula

[0086] where p is a positive integer. In this embodiment, p can specifically take the value 2, that is, in step S50102, the confidence interval of the float voltage prediction value is calculated according to the formula

[0087]

[0088]

[0089] The confidence interval of the float voltage prediction value obtained in step S501 indicates a voltage range, and according to , it is determined whether this voltage range is within the preset safe range or outside the preset safe range, so as to determine whether to perform step S502 or step S503. Figure 8 In this embodiment, the preset safe range can be a voltage range in which the battery body (battery monomer) is guaranteed to be in a safe state when the battery body is subjected to float charging. For example, for a valve-regulated lead-acid battery, the preset safe range can be 2.23V-2.30V, that is, the valve-regulated lead-acid battery is subjected to float charging using a float voltage (voltage accepted by the battery monomer) in the range of 2.23V-2.30V, which can guarantee the safety of the float charging process and make the service life of the valve-regulated lead-acid battery within the expected service life.

[0090]

[0091] ​​​​​​​In this embodiment, when determining whether the confidence interval is within or outside the preset safety range, the following judgment criteria can be adopted: if the confidence interval belongs to a proper subset of the preset safety range, then the confidence interval is determined to be within the preset safety range; otherwise, the confidence interval is determined to be outside the preset safety range.

[0092] For example, when the preset safety range is 2.23V-2.30V, the predicted float charge voltage value is... The confidence interval is If the lower limit of the confidence interval is greater than the lower limit of the preset safety range, and the upper limit of the confidence interval is less than the upper limit of the preset safety range, that is... >2.23V and If the value is less than 2.30V, then the confidence interval is determined to be within the preset safety range, and step S502 will be triggered. Conversely, if the lower limit of the confidence interval is less than the lower limit of the preset safety range, or the upper limit of the confidence interval is greater than the upper limit of the preset safety range, that is, the confidence interval does not belong to a proper subset of the preset safety range, then the confidence interval is determined to be outside the preset safety range, and step S503 will be triggered.

[0093] If step S502 is triggered when the confidence interval is within the preset safety range, the battery management system sets the float charge voltage of the battery to be equal to the predicted float charge voltage value; if step S503 is triggered when the confidence interval is outside the preset safety range, the battery management system switches to the empirical formula compensation mode.

[0094] Reference Figure 8 In the empirical formula compensation mode, the battery management system can use the formula...

[0095] Calculate the empirical compensation value of the float charge voltage. .in, The base float charge voltage is a fixed voltage value, such as 2.23V. This is the temperature compensation coefficient. This is the internal resistance compensation coefficient. It is the reference temperature, also known as the base temperature. Specifically, it is a fixed temperature value, usually the standard temperature for battery design, such as 25°C. It is the operating temperature in the real-time feature vector, representing the battery body over a time period. The actual temperature reached inside. This is a temperature difference compensation term, which can offset the effect of temperature changes on the float charge voltage. For example, when the operating temperature of the battery increases, the empirical compensation value of the float charge voltage will be adjusted accordingly. Conversely, when the battery's operating temperature decreases, the empirical compensation value of the float charge voltage will decrease accordingly. Increase.

[0096] The experience formula compensation mode executed in step S503 is equivalent to a floating charge voltage which is summarized according to experience, and which considers the deviation of the working temperature of the battery body from the reference temperature to correct the basic floating charge voltage to obtain a floating charge voltage experience compensation value In the experience formula compensation mode, the battery management system only needs to consider the parameter of the working temperature of the battery body, and does not need to process high-dimensional historical feature vectors and other data, thereby having a faster response speed.

[0097] In this embodiment, the principle of executing steps S501-S503 is that: since the floating charge voltage prediction value is essentially a prediction value, there is a certain deviation between the floating charge voltage prediction value and the actual most appropriate floating charge voltage; the confidence interval of the floating charge voltage prediction value indicates the interval in which the actual most appropriate floating charge voltage has a great probability (for example, a probability of 0.9544 when p=2); if the confidence interval of the floating charge voltage prediction value is within the preset safety range, it indicates that if the floating charge voltage of the battery body is set according to the floating charge voltage prediction value , there is a great possibility to set the most appropriate floating charge voltage under the premise of guaranteeing safety, and therefore the floating charge voltage of the battery body can be set to equal the floating charge voltage prediction value , thereby exerting the effect of the floating charge voltage prediction value ; if the confidence interval of the floating charge voltage prediction value is outside the preset safety range, it indicates that the most appropriate floating charge voltage has a great possibility of deviating from the preset safety range, that is, if the most appropriate floating charge voltage is to be achieved, it may exceed the safety range of the battery body, or it is difficult to set the most appropriate floating charge voltage under the premise of guaranteeing safety, and therefore the floating charge voltage of the battery body is not set according to the floating charge voltage prediction value , but a floating charge voltage experience compensation value is calculated by the experience formula compensation mode to set the floating charge voltage of the battery body, thereby exerting the effect of the experience compensation and obtaining a usable floating charge voltage.

[0098] The computer program for executing the adaptive floating charge voltage control method for storage batteries based on Gaussian process regression in this embodiment can be written into a computer device or a storage medium, and when the computer program is read out and run, the adaptive floating charge voltage control method for storage batteries based on Gaussian process regression in this embodiment is executed, thereby achieving the same technical effects as the adaptive floating charge voltage control method for storage batteries based on Gaussian process regression in the embodiment.

[0099] Taking a 48V valve-regulated lead-acid battery pack (24 cells in series) of a communication base station as an example, a battery adaptive float voltage control method based on Gaussian process regression is implemented: 1. Initial training: load historical data (including temperature, load current, and internal resistance change records in different seasons), and build an initial GPR model.

[0100] 2. Real-time prediction: when the temperature of a certain cell rises to 35℃, the GPR model predicts that the float voltage needs to be adjusted from 2.27V to 2.25V to avoid overcharging at high temperature, based on the increase of internal resistance (from 5mΩ to 8mΩ) and current fluctuations.

[0101] 3. Abnormal processing: if the confidence interval is out of limit due to sensor noise in a certain prediction, the system automatically switches to compensation mode and triggers a self-diagnosis program to detect the sensor state.

[0102] By implementing the battery adaptive float voltage control method based on Gaussian process regression, the following beneficial effects can be obtained: 1. Life extension: compared with the fixed float voltage strategy, the battery cycle life is improved by 20%~30%; 2. Energy efficiency optimization: the float voltage dynamically matches the actual needs of the battery, and the overall energy consumption of the system is reduced by 8%~12%; 3. Safety enhancement: through confidence interval monitoring and mode switching, the false alarm rate is reduced to below 0.1%; 4. High-precision prediction: using the non-linear modeling capability of GPR for multi-dimensional input features, the complex mapping relationship between battery state and float voltage is captured; 5. Dynamic adaptability: by updating the training data set online, the model hyperparameters are continuously optimized to adapt to the characteristic drift in the battery aging process; 6. Robustness guarantee: combined with the confidence interval of the prediction result, a safety threshold mechanism is designed to avoid voltage regulation out of control due to abnormal data.

[0103] In this embodiment, the computer device shown in Figure 9 including a memory and a processor, the memory is used to store at least one program, and the processor is used to load at least one program to implement the battery adaptive float voltage control method based on Gaussian process regression, so as to obtain the effect of the battery adaptive float voltage control method based on Gaussian process regression.

[0104] In this embodiment, a computer readable storage medium can be used, which stores a program executable by a processor, such as Figure 10As shown, the processor-executable program, when executed by the processor, is used to execute the adaptive float charging voltage control method for storage battery based on Gaussian process regression, so that the effect of the adaptive float charging voltage control method for storage battery based on Gaussian process regression is obtained.

[0105] The embodiment of the present application further provides a computer program product, comprising a computer program which, when executed by a processor, implements the method described above.

[0106] It can be understood that the contents in the above method embodiments are all applicable to the present program product embodiment, the program product embodiment specifically implements the same functions as the above method embodiments, and achieves the same beneficial effects as the above method embodiments.

[0107] It should be noted that, unless otherwise specified, when a feature is said to be "fixed", "connected" to another feature, it can be directly fixed, connected to the other feature, or indirectly fixed, connected to the other feature. In addition, the up, down, left, right and other descriptions used in the present disclosure are only relative to the relative positional relationship of the components of the present disclosure in the drawings. In the present disclosure, the singular forms "a", "an" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. In addition, unless otherwise defined, all technical and scientific terms used in the present embodiment are the same as those commonly understood by those skilled in the art. The terms used in the present embodiment are only used to describe the specific embodiments, and are not intended to limit the embodiments of the present application. The term "and / or" used in the present embodiment includes any combination of one or more related listed items.

[0108] It should be understood that although the terms first, second, third, etc. can be used in the present disclosure to describe various elements, these elements should not be limited to these terms. These terms are only used to distinguish one type of element from another. For example, without departing from the scope of the present disclosure, the first element can also be referred to as the second element, and similarly, the second element can also be referred to as the first element. The use of any and all examples or exemplary language ("for example", "for example", etc.) provided in the present embodiment is only intended to better illustrate the embodiments of the present application, and unless otherwise required, does not impose any limitation on the scope of the embodiments of the present application.

[0109] It should be appreciated that embodiments of the present application can be realized by a computer hardware, a combination of hardware and software, or by computer instructions stored on a non-transitory computer-readable storage medium. The methods can be implemented in a computer program product tangibly embodied in a machine- readable storage medium (e.g., magnetic disk, optical disk, memory, etc.) including a machine-readable storage medium. The methods can be implemented using standard programming techniques - including the configuration of a non-transitory computer-readable storage medium with computer program code - which when executed by a computer, cause the computer to perform the methods described in the specific embodiments. Each program can be implemented in a high level procedural or object oriented programming language to communicate with a computer system. However, the programs can be implemented in assembly or machine language, if desired. In any case, the language can be a compiled or interpreted language. Furthermore, the programs can be stored on a computer program product which can be a dedicated hardware, or a general purpose computer which is programmed to perform the functions described in the specific embodiments.

[0110] Further, the operations of the processes described in the embodiments can be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described in the embodiments (or variations and / or combinations thereof) can be implemented under the control of one or more computer systems configured with executable instructions (e.g., computer program code, one or more computer programs or one or more applications) to perform the processes described in the specific embodiments, and can be implemented as code (e.g., executable instructions, one or more computer programs or one or more applications) running on a computer system (a digital electronic device), and / or executed by hardware or combinations thereof. The computer program includes a plurality of instructions that can be executed by one or more processors.

[0111] Further, the methods can be implemented in any suitable type of computing platform that is operatively connected to any suitable type of computing platform, including but not limited to a personal computer, a mini-computer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, and the like. Aspects of the embodiments of the present application can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated to the computing platform, such as a hard disk, an optical read and / or write storage medium, RAM, ROM, and the like, such that it can be read by a programmable computer to configure and operate the computer to perform the processes described herein. Further, the machine-readable code, or portions thereof, can be transmitted over wired or wireless networks. The present application includes these and other different types of non-transitory computer-readable storage media when the instructions or programs incorporating microprocessors or other data processors implement the steps described above. The embodiments of the present application also include the computer itself when programmed according to the methods and techniques of the embodiments of the present application.

[0112] A computer program can be applied to input data to perform the functions of the present embodiments to transform the input data to generate output data that is stored to non-volatile memory. The output information can also be applied to one or more output devices such as a display. In preferred embodiments of the present embodiments, the transformed data represents a physical and tangible object, including a particular visual depiction of a physical and tangible object produced on a display.

[0113] The above merely preferred embodiments of the present embodiments, the present embodiments are not limited to the above-described embodiments, as long as the same means to achieve the technical effects of the present embodiments, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present embodiments should be included in the scope of protection of the present embodiments. The technical solutions and / or implementation manners within the protection scope of the present embodiments can have various modifications and changes.

Claims

1. A method for adaptive float charge voltage control of a storage battery, characterized in that, The battery adaptive float charge voltage control method based on Gaussian process regression includes: Obtain a training dataset; the training dataset includes multiple historical feature vectors and multiple historical float charge voltage values ​​measured from the battery cell; Based on the training dataset, a Gaussian process regression model is established; Real-time feature vectors are obtained from the battery body detection; The Gaussian process regression model is used to process the real-time feature vector to obtain the predicted value of the float charge voltage of the battery body. The float charge voltage of the battery is controlled based on the predicted float charge voltage value.

2. The adaptive float charge voltage control method for a battery according to claim 1, characterized in that, The acquisition of the training dataset includes: Establish the initial training dataset; A sliding window mechanism is used to periodically update the training dataset online.

3. The adaptive float charge voltage control method for a battery according to claim 1, characterized in that, The step of establishing a Gaussian process regression model based on the training dataset includes: The hyperparameters are determined based on the historical feature vectors and the historical floating charge voltage values. Based on the hyperparameters and each of the historical feature vectors, the kernel function in the Gaussian process regression model is determined; the kernel function has the form of a radial basis function.

4. The adaptive float charge voltage control method for a battery according to claim 3, characterized in that, The process of determining hyperparameters based on each of the historical feature vectors and each of the historical floating charge voltage values ​​includes: Based on the historical feature vectors and the historical floating charge voltage values, establish the marginal likelihood function; The conjugate gradient method is used to solve for the length scale, signal variance, and noise variance that maximize the marginal likelihood function; The solved length scale, signal variance, and noise variance are used as the hyperparameters.

5. The adaptive float charge voltage control method for a battery according to claim 3, characterized in that, The step of determining the kernel function in the Gaussian process regression model based on the hyperparameters and each of the historical feature vectors includes: According to the formula Establish the kernel function; where, For the kernel function, For the first in the training dataset A historical feature vector, For the first in the training dataset A historical feature vector, For length scale, For signal variance, For noise variance, Let Kronecker function be used.

6. The battery adaptive float charge voltage control method according to claim 5, characterized in that, The step of processing the real-time feature vector using the Gaussian process regression model to obtain the predicted float charge voltage of the battery includes: According to the formula The predicted value of the float charge voltage is calculated; wherein, , , The real-time feature vector, , ,…, For each of the aforementioned historical feature vectors, For each of the aforementioned historical float charge voltage values, Indicates matrix transpose. It is a unit vector. This is the predicted value of the float charge voltage.

7. The battery adaptive float charge voltage control method according to claim 5 or 6, characterized in that, The step of controlling the float charge voltage of the battery body based on the predicted float charge voltage includes: Obtain the confidence interval of the predicted float charge voltage; When the confidence interval is within the preset safety range, the float charge voltage of the battery body is set to be equal to the predicted value of the float charge voltage; When the confidence interval is outside the preset safety range, the float charge voltage empirical compensation value is obtained, and the float charge voltage of the battery body is set to be equal to the float charge voltage empirical compensation value.

8. The adaptive float charge voltage control method for a battery according to claim 7, characterized in that, The confidence interval for obtaining the predicted floating charge voltage includes: According to the formula Calculate the variance of the predicted float charge voltage; where, Let the variance be the variance. According to the formula Calculate the confidence interval, where p is a positive integer.

9. A computer device, characterized in that, It includes a memory and a processor, the memory being used to store at least one program, and the processor being used to load at least one program to execute the battery adaptive float charge voltage control method based on Gaussian process regression as described in any one of claims 1-8.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the adaptive float charge voltage control method for batteries based on Gaussian process regression as described in any one of claims 1-8.

Images