A real-time soc prediction method for a vanadium redox flow battery system

Abstract

The application discloses a real-time SOC prediction method for a vanadium redox flow battery system, which comprises the following steps: sampling positive and negative electrolytes in the charging and discharging cycle of the vanadium redox flow battery system to determine the concentrations of the four-valent and five-valent vanadium ions in the positive electrolyte and the concentrations of the two-valent and three-valent vanadium ions in the negative electrolyte; recording the charging and discharging parameters at the sampling time of the positive and negative electrolytes; taking the charging and discharging parameters as characteristic vectors X; taking the concentrations of the four-valent and five-valent vanadium ions in the positive electrolyte and the concentrations of the two-valent and three-valent vanadium ions in the negative electrolyte as a target function y, and establishing a database; dividing the data in the database into a training set and a test set; training the data in the training set by using an artificial intelligence algorithm to establish a prediction model for the concentrations of the four-valent and five-valent vanadium ions in the positive electrolyte and the concentrations of the two-valent and three-valent vanadium ions in the negative electrolyte of the vanadium redox flow battery. The method can be used for real-time prediction of the concentrations of active substances in the positive and negative electrolytes of the vanadium redox flow battery system in the charging and discharging cycle, and real-time monitoring of the SOC of the vanadium redox flow battery system.

Description

TECHNICAL FIELD

[0001] The application relates to a real-time SOC prediction method for a vanadium flow battery system and belongs to the technical field of vanadium flow battery systems. BACKGROUND

[0002] The state-of-charge (SOC) monitoring of a vanadium flow battery system is very important for system safety and stability. A reliable real-time SOC prediction system for a vanadium flow battery can significantly improve the stability of the system and reduce operation and maintenance costs. At present, the SOC estimation methods for a vanadium flow battery system mainly include a discharge experiment method, an ampere-hour integration method, an open-circuit voltage method and a Kalman filtering method. The difficulty of real-time SOC prediction for a vanadium flow battery system lies in that, during the operation of the vanadium flow battery system, the migration of electrolyte can cause the imbalance of positive and negative electrolyte, so that the real-time estimated SOC is inaccurate. The discharge experiment method has high precision but cannot be measured online; the ampere-hour integration method and the open-circuit voltage method cannot consider the precision decline caused by electrolyte migration; the Kalman filtering method greatly depends on the accuracy of the battery model and cannot consider the influence of electrolyte migration during the operation of the vanadium flow battery, so these methods cannot realize real-time prediction of the SOC of the system in the large-scale demonstration and commercial operation of the vanadium flow battery system. SUMMARY

[0003] According to one aspect of the application, a real-time SOC prediction method for a vanadium flow battery system is provided. In view of the limitations and complexities of existing SOC detection methods, the application only uses the charge and discharge parameter data of the charge and discharge process of the vanadium flow battery system to realize real-time online prediction of the SOC without introducing additional complex collection instruments and measuring devices, thereby solving the problem that the high precision and real-time performance of the existing SOC detection method are difficult to be considered.

[0004] The application adopts the following technical scheme:

[0005] A real-time SOC prediction method for a vanadium flow battery system comprises the following steps:

[0006] S1. A charge and discharge cycle test is performed on the vanadium flow battery system, the electrolyte in the electrolyte storage tank during the charge and discharge cycle process is sampled for several times, the vanadium ion concentration in the electrolyte is obtained, and the charge and discharge parameters at the sampling time are recorded;

[0007] S2. The charge and discharge parameters recorded in step S1 are taken as feature vectors X, the vanadium ion concentration obtained in step S1 is taken as a target function y, a database is established, the data in the database are divided into a training set and a test set, an artificial intelligence algorithm is used to train the data in the training set and model the data, after data verification and model parameter adjustment, a prediction model for predicting the vanadium ion concentration in the electrolyte is obtained.

[0008] S3, when the all-vanadium redox flow battery system is running, real-time charging and discharging parameters are collected, the concentration of vanadium ions in the real-time electrolyte is predicted through the prediction model in step S2, and then the real-time SOC of the all-vanadium redox flow battery system is calculated according to formula (1):

[0009]

[0010] In the formula, [V 2+ ], [V 3+ ], [VO 2+ ] and represent the concentrations of divalent, trivalent, tetravalent and pentavalent vanadium ions, respectively.

[0011] Optionally, in step S1, the concentration of vanadium ions in the electrolyte is obtained by:

[0012] The concentration of tetravalent and pentavalent vanadium ions in the positive electrolyte tank is obtained, and / or the concentration of divalent and trivalent vanadium ions in the negative electrolyte tank is obtained.

[0013] Optionally, in step S1, the charging and discharging parameters are selected from at least one of the current, voltage, average voltage, charging capacity during charging, and discharging capacity during discharging.

[0014] Optionally, in step S1, the charging and discharging cycle includes n groups of charging and discharging cycles, and n>1.

[0015] The sampling process includes:

[0016] In the charging cycle, at least 4 samplings are performed, wherein the first sampling is performed within 1 minute after the start of charging, and the last sampling is performed within 1 minute before the end of charging.

[0017] In the discharging cycle, at least 4 samplings are performed, wherein the first sampling is performed within 1 minute after the start of charging, and the last sampling is performed within 1 minute before the end of charging.

[0018] Optionally, in step S1, after sampling, the process of obtaining the concentration of vanadium ions in the electrolyte is titration detection of the concentration of vanadium ions in the electrolyte, and for the concentration of vanadium ions corresponding to the time not sampled, the concentration of vanadium ions is obtained by interpolation method according to formula (2) through the concentration of vanadium ions determined by titration of adjacent two samplings and the capacity difference of charging and discharging:

[0019]

[0020] In the formula, c(t1), c(t2), c(t i ) respectively represent the first sampling point, the second sampling point and t iVanadium ion concentration at time C(t1), C(t2), C(t i ) respectively represent the first sampling point, the second sampling point and the charging or discharging capacity at time t i .

[0021] Optionally, in step S2, the n groups of charge and discharge parameters recorded in step S1 are taken as feature vectors X, and the vanadium ion concentration in the electrolyte corresponding to the nth group of charge and discharge cycles is taken as the target function y to establish a database.

[0022] Optionally, the data of any m1 groups of cycles in the database are taken as a training set of charging cycles;

[0023] The data of any z1 groups of cycles in the database are taken as a test set of charging cycles;

[0024] The data of any m2 groups of cycles in the database are taken as a training set of discharging cycles;

[0025] The data of any z2 groups of cycles in the database are taken as a test set of discharging cycles;

[0026] Wherein, m1, z1, m2, z2 are independently less than n.

[0027] Optionally, in step S2, the artificial intelligence algorithm is selected from one of autoregressive integrated moving average algorithm (ARIMA), Prophet algorithm, gradient boosting machine algorithm (GBM), Seq2Seq type algorithm, TCN algorithm, WaveNet algorithm, Informer algorithm and Transformer algorithm.

[0028] Optionally, the Seq2Seq type algorithm is selected from one of recurrent neural network RNN, long short-term memory neural network LSTM and autoregressive recurrent neural network DeepAR.

[0029] Optionally, when the artificial intelligence algorithm is a long short-term memory neural network (LSTM) algorithm, the activation function is “relu”, the optimizer of the neural network is “Adam”, the number of hidden layers is at least 1 layer, the number of neurons in each layer is at least 8, and a part of the neurons are randomly discarded in the training process to prevent overfitting.

[0030] Optionally, in step S2, the data modeling process further comprises adopting a regularization method to prevent overfitting.

[0031] The regularization method is selected from one or a composite regularization method of at least any two of L1 regularization, L2 and Dropout.

[0032] Optionally, the parameters of the regularization method are (0,1).

[0033] Optionally, the range of randomly discarded neurons is (0,1).

[0034] Optionally, the prediction model for the concentration of vanadium ions in the electrolyte is a concentration prediction model for tetravalent and pentavalent vanadium ions, or a concentration prediction model for divalent and trivalent vanadium ions.

[0035] The beneficial effects that this application can produce include:

[0036] A real-time SOC prediction method for vanadium redox flow battery systems uses parameters such as current, voltage, and charge / discharge capacity during the charging and discharging process as input features. An artificial intelligence algorithm is used to establish a prediction model for the concentration of vanadium ions in various valence states at the positive and negative electrodes. This method can utilize real-time data from the charging and discharging process to predict the electrolyte SOC state of the vanadium redox flow battery after electrode imbalance, exhibiting high sensitivity and accuracy. The method is low-cost, requires no additional data acquisition devices or testing instruments, and is easy to implement and maintain. The vanadium redox flow battery positive electrode SOC prediction model established using parameter data from the charging and discharging process of a single cell can be applied to large-scale demonstration and commercial systems. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating the implementation steps of the real-time SOC prediction method for the all-vanadium redox flow battery system of this application.

[0038] Figure 2 This paper compares the model prediction results and actual values ​​of the concentrations of tetravalent and pentavalent vanadium ions at the positive electrode during the charging process in Example 1 of this application.

[0039] Figure 3 This is a comparison between the model prediction results and the actual values ​​of the tetravalent and pentavalent vanadium ion concentrations at the positive electrode during the discharge process in Example 1 of this application. Detailed Implementation

[0040] The present application is described in detail below with reference to the embodiments, but the present application is not limited to these embodiments.

[0041] Unless otherwise specified, all raw materials used in the embodiments of this application were purchased through commercial channels.

[0042] Unless otherwise specified, all test methods are conventional and all instrument settings are those recommended by the manufacturer.

[0043] Example 1

[0044] like Figure 1 The implementation steps shown are as follows: Using a prediction model of the concentrations of tetravalent and pentavalent vanadium ions in the cathode of a vanadium redox flow battery system, a real-time SOC prediction method for the vanadium redox flow battery system is implemented:

[0045] Step 1: Perform charge-discharge test on the all-vanadium redox flow battery system, and sample and titrate the positive and negative electrolytes during the charge-discharge test to determine the concentrations of the positive tetravalent and pentavalent vanadium ions and the concentrations of the negative divalent and trivalent vanadium ions. In this embodiment, the SOC of the all-vanadium redox flow battery system is determined by the concentrations of the positive tetravalent and pentavalent vanadium ions; the rated power of the system is 5 kW. The charge-discharge test process is as follows: first, stand for 30 seconds, then perform 5 kW constant power charging, charge to the constant voltage charging of 1.55 V to 1.53 V, charge to 100% SOC, stand for 30 seconds; finally, perform constant power discharge, discharge to the cut-off voltage of 1 V, complete 1 charge-discharge cycle; the sampling time for titration is as follows: for one charging cycle, take at least 4 samples, take 1 sample within 1 minute at the beginning of charging, take 1 sample within 1 minute at the end of charging, and take at least 2 more samples during charging; for one discharging cycle, take at least 4 samples, take 1 sample within 1 minute at the beginning of discharging, take 1 sample within the last 1 minute at the end of discharging, and take at least 2 more samples during discharging.

[0046] Step 2: Based on the titrated concentrations of the positive tetravalent and pentavalent vanadium ions, the concentrations of the positive tetravalent and pentavalent vanadium ions in the cycle are calculated by interpolation according to the charge-discharge capacity; for the vanadium ion concentration corresponding to the time not sampled, the vanadium ion concentration is calculated by interpolation method according to formula (2) according to the difference between the vanadium ion concentrations of the adjacent two sampling titrations and the charge-discharge capacity:

[0047]

[0048] In the formula, c(t1), c(t2), c(t i ) respectively represent the vanadium ion concentrations at the 1st sampling point, the 2nd sampling point and t i time, C(t1), C(t2), C(t i ) respectively represent the charging or discharging capacity at the 1st sampling point, the 2nd sampling point and t i time.

[0049] Take the characteristics of the first n groups of charge-discharge cycles as input features X, and take the concentrations of the positive tetravalent and pentavalent vanadium ions corresponding to the nth group of charge-discharge cycles as the target function y to establish a database. The characteristics of the charge-discharge cycles include: current, voltage, average standing voltage, charging capacity (charging process), discharging capacity (discharging process); n is an integer greater than or equal to 1.

[0050] Step 3: The data in the database is divided into a training set and a test set. In this example, the charging cycle: the data of the 9th, 15th, 55th, and 81st charging cycles are used as the training set, and the data of the 88th charging cycle is used as the test set; the discharge cycle: the data of the 2nd, 21st, 29th, and 39th discharge cycles are used as the training set, and the data of the 61st discharge cycle is used as the test set.

[0051] Step 4: The data in the training set is trained using an artificial intelligence algorithm to establish a prediction model for the concentration of vanadium ions in the positive electrode of the all-vanadium redox flow battery during charging and discharging. The artificial intelligence uses a long short-term memory (LSTM) algorithm, with one hidden layer and 32 neurons, and a "relu" activation function. The neural network optimizer is "Adam", and L2 regularization is used to prevent overfitting, with a parameter of 0.01. During training, 20% of the neurons are randomly discarded to prevent overfitting. The prediction results and actual results for the concentration of vanadium ions in the positive electrode are shown in Figure 2 , and the model training accuracy is shown in Table 1.

[0052] Step 5: The trained prediction model for the concentration of vanadium ions in the positive electrode of the all-vanadium redox flow battery is evaluated using the data in the test set. The training results are shown in Figure 2 , and the accuracy of the test set is shown in Table 1.

[0053] Step 6: During the operation of the all-vanadium redox flow battery system, real-time charging and discharging parameter data are collected, and the real-time concentration of vanadium ions in the positive electrode electrolyte is predicted using the prediction model for the concentration of vanadium ions in the positive electrode. Then, the real-time SOC of the all-vanadium redox flow battery system is calculated according to formula (1):

[0054]

[0055] where [V 2+ ], [V 3+ ], [VO 2+ ], and represent the concentrations of divalent, trivalent, tetravalent, and pentavalent vanadium ions, respectively.

[0056] Table 1: Model accuracy of the prediction model for the concentration of vanadium ions in the positive electrode of the all-vanadium redox flow battery system

[0057]

[0058] The above merely describes several embodiments of the present application, and does not limit the present application in any form. Although the present application is disclosed with the preferred embodiments, it is not intended to limit the present application. Any skilled person in the art can make some changes or modifications to the disclosed technical contents without departing from the scope of the technical solutions of the present application, and the equivalent embodiments are equivalent to the equivalent embodiments, which are within the scope of the technical solutions.

Claims

1. A method for real-time SOC prediction of a vanadium redox flow battery system, characterized in that, The method comprises the following steps: S1, performing a charge-discharge cycle test on the all-vanadium redox flow battery system, sampling the electrolyte in the electrolyte storage tank for several times during the charge-discharge cycle process, obtaining the concentration of vanadium ions in the electrolyte, and recording the charge-discharge parameters at the sampling time; S2, establishing a database with the charge-discharge parameters recorded in step S1 as the feature vector X and the concentration of vanadium ions in the electrolyte obtained in step S1 as the target function y, dividing the data in the database into a training set and a test set, training the data in the training set by using an artificial intelligence algorithm, data modeling, data verification, and model parameter adjustment to obtain a prediction model for predicting the concentration of vanadium ions in the electrolyte; S3, during the operation of the all-vanadium redox flow battery system, real-time acquisition of the charge-discharge parameters, prediction of the real-time concentration of vanadium ions in the electrolyte by the prediction model in step S2, and then calculation of the real-time SOC of the all-vanadium redox flow battery system according to formula (1): Equation (1); wherein, , , and denote the concentration of divalent, trivalent, tetravalent and pentavalent vanadium ions, respectively. In step S1, the concentration of vanadium ions in the electrolyte is obtained by: obtaining the concentration of tetravalent and pentavalent vanadium ions in the positive electrolyte storage tank, and / or obtaining the concentration of divalent and trivalent vanadium ions in the negative electrolyte storage tank; In step S1, after sampling, the process of obtaining the concentration of vanadium ions in the electrolyte is: titration method is used to detect the concentration of vanadium ions in the electrolyte, for the vanadium ion concentration corresponding to the time when the sampling is not performed, the concentration of vanadium ions is calculated by interpolation method according to the concentration of vanadium ions determined by titration of adjacent two times of sampling and the capacity difference of charge and discharge according to formula (2): Formula (2); wherein , , respectively represent the vanadium ion concentration at the 1st sampling point, the 2nd sampling point and the t i th sampling point, , , respectively represent the charging or discharging capacity at the 1st sampling point, the 2nd sampling point and the t i th sampling point.

2. The method of claim 1, wherein the method is a method of predicting a real-time SOC of a vanadium redox flow battery system, the method comprising: In step S1, the charge-discharge parameters are selected from at least one of the current, voltage, average voltage, charge capacity during charging, and discharge capacity during discharging.

3. The method of claim 1, wherein the method is a method of predicting a real-time SOC of a vanadium redox flow battery system, the method comprising: In step S1, the charge-discharge cycle comprises n groups of charge-discharge cycles, and n>1. The sampling process comprises: In the charging cycle, at least 4 times of sampling are performed, wherein the first sampling is performed within 1 min after the start of charging, and the last sampling is performed within 1 min before the end of charging; In the discharging cycle, at least 4 times of sampling are performed, wherein the first sampling is performed within 1 min after the start of charging, and the last sampling is performed within 1 min before the end of charging.

4. The method of claim 3, wherein the method is a method of predicting a real-time SOC of a vanadium redox flow battery system, the method comprising: In step S2, the continuous n groups of charge-discharge parameters recorded in step S1 are used as the feature vector X, and the concentration of vanadium ions in the electrolyte corresponding to the n th group of charge-discharge cycles is used as the target function y to establish a database.

5. The method of claim 4, wherein the method is used for a vanadium redox flow battery system real-time SOC prediction, and Any m1 groups of cycle data in the database are used as the training set of the charging cycle; Any z1 groups of cycle data in the database are used as the test set of the charging cycle; Any m2 groups of cycle data in the database are used as the training set of the discharging cycle; Any z2 groups of cycle data in the database are used as the test set of the discharging cycle; Wherein, m1, z1, m2, z2 are independently less than n.

6. The real-time SOC prediction method of the all-vanadium redox flow battery system according to claim 1, wherein in step S2, the artificial intelligence algorithm is selected from one of the autoregressive integrated moving average algorithm, the Prophet algorithm, the gradient boosting machine algorithm, the Seq2Seq algorithm, the TCN algorithm, the WaveNet algorithm, the Informer algorithm, and the Transformer algorithm. The Seq2Seq algorithm is selected from one of a recurrent neural network (RNN), a long short-term memory neural network (LSTM), and a deep autoregressive recurrent neural network (DeepAR).

7. The method of claim 1, wherein the method is characterized in that, In step S2, the data modeling process further comprises using a regularization method to prevent overfitting. The regularization method is selected from one of L1 regularization, L2, and Dropout, or a composite regularization method of at least any two thereof.

8. The method of claim 1, wherein the method is a method of real-time SOC prediction for a vanadium redox flow battery system, and wherein the method comprises: The prediction model for predicting the concentration of vanadium ions in the electrolyte is a concentration prediction model for tetravalent and pentavalent vanadium ions, or a concentration prediction model for divalent and trivalent vanadium ions.

Images