SOC denominator array solving algorithm in data loss mode

Abstract

The invention discloses an SOC denominator array solving algorithm in a data loss mode, which comprises the following steps of: obtaining an initial denominator array VD0 by calculating an original capacity variation corresponding to each SOC change position in a data loss scene, clustering the VD0, marking a numerical value of a clustering result as aL, and assigning aL to VDL; and carrying out capacity value correction by utilizing a difference value between the VDL and the VD0, calculating a corresponding virtual # imgabs0 # in a one-dimensional array distribution state based on the corrected capacity value, calculating a mean square error J between the virtual # imgabs0 # and the real SOCi, and when the mean square error J is greater than a threshold value m1, retrieving a position where the difference value of the # imgabs1 # is greater than a threshold value m2, and marking elements of the VDL in a range from the position to the last serial number, re-starting clustering to obtain aL + 1, assigning aL + 1 to VDL + 1, continuously adjusting the SOC denominator array through multiple iterations until L is smaller than a threshold value m1, and outputting a one-dimensional array VDG of the SOC denominator. According to the method, the SOC denominator solving problem in a data loss scene is solved, and the problem of battery diagnosis failure caused by large ampere-hour integral capacity error is avoided.

Description

Technical Field

[0001] The present invention relates to the field of battery application technology, and more specifically to an SOC denominator array solution algorithm in a lost data mode. Background Art

[0002] The battery system of new energy vehicles has the function of displaying SOC, and its value represents the percentage of the remaining available capacity of the battery system to the full capacity. It is calculated by the mainboard of the battery management system (BMS) through the current value collected by the internal Can data of the battery system. The calculation is mainly based on ampere-hour integration, voltage or open-circuit voltage correction. During the entire discharge process of the battery, especially in the lithium iron phosphate battery system, the battery voltage platform is too flat, and voltage or open-circuit voltage correction cannot be implemented. SOC usually relies on ampere-hour integration to maintain a fixed rate of decline, that is, the capacity value released for each unit reduction in SOC is usually a constant, and this capacity value is marked as the SOC denominator.

[0003] The setting of the SOC denominator is mainly determined by the capacity of the battery system or the health of the battery cells. The SOC denominator of most vehicles remains unchanged during the discharge process. Based on this principle, capacity correction when the battery loses data or battery health calculation can be implemented. However, in some special scenarios, such as when the battery system has serious imbalance problems or contains special cells with abnormal self-discharge, after the vehicle's power consumption reaches a certain level, the voltage drops to the end of the SOC, triggering the terminal voltage correction of the BMS. At this time, the SOC denominator will change significantly, and a smaller denominator value will be used to guide the rapid decline of SOC. In this scenario, the SOC denominator of the discharge process is an array composed of several constants, and each constant is used for different durations. Because the SOC denominator is crucial for capacity correction when data is lost and battery health calculation, accurately solving the SOC denominator array is a necessary condition to ensure the accuracy of battery diagnosis.

[0004] At present, the existing methods for compensating for missing data scenarios include window average method, nearest filling method, interpolation fitting method, etc. However, these methods are not suitable for compensating for current data missing scenarios during battery discharge, because the current in this scenario changes rapidly, which does not meet the characteristics of the same window average value, the same nearest value, and linear interpolation change, and cannot achieve the purpose of accurately correcting the capacity. To this end, we provide an algorithm for solving the SOC denominator array in the missing data mode. Summary of the invention

[0005] The present invention provides an algorithm for solving the SOC denominator array in a data loss mode, aiming to solve the shortcomings of existing data loss scenario compensation methods, such as large correction errors and failure to accurately correct capacity.

[0006] The present invention adopts the following technical solution:

[0007] An algorithm for solving the SOC denominator array in a data loss mode includes the following steps:

[0008] Step 1: Collect vehicle current, time, and state of charge (SOC) i Data, the current and time data of the vehicle discharge process are integrated by ampere-hour to obtain the capacity array Q corresponding to each moment 1*N , and obtain SOC i The time Tc corresponding to the changed position 1*M , where N is the current array length, M is SOC i The number of changes that occurred;

[0009] Step 2: Calculate each SOC i The original capacity change corresponding to the changed position is used to obtain the initial denominator array VD 0 ;

[0010] Step 3: VD obtained in step 2 0 Clustering is performed, and the numerical value of the clustering result is marked as a L , a L Assign to VD L , where VD L All elements are a L ;

[0011] Step 4: Use VD L With VD 0 The capacity value is corrected based on the difference between them, and the corresponding virtual

[0012] Step 5: Calculate virtual With real SOC i The mean square error J between

[0013] Step 6: When the mean square error J is greater than the threshold m1, search The position where the difference is greater than the threshold m2 is marked as k1; for VD L The elements from k1 to M in the cluster are re-clustered to obtain a L+1 , and a L+1 Assign to VD L+1 , where VD L +1 The element at the corresponding position from k1 to the last position is a L+1 , other position elements remain the same as VD L Same, L represents the number of iterations;

[0014] Step 7: Repeat steps 4 to 7 until the mean square error J is less than the threshold m1, and output the SOC denominator one-dimensional array VDG , where G is the final number of iterations.

[0015] In a preferred embodiment, the initial denominator array VD in the above step 2 is 0 The method of obtaining is as follows: calibrate the capacity values ​​Q(Tc(k+1)) and Q(Tc(k)) corresponding to the time Tc(k+1) and Tc(k), Q(Tc(k+1))-Q(Tc(k)) is q i (k), find the q corresponding to all Tc moments i , that is, forming the initial SOC denominator array VD 0 .

[0016] In a preferred implementation scheme, the clustering in the above step three can select the average value of the class with the largest average value in the clustering results as the clustering result; or select the class with the largest average value in the clustering results as the target array, perform secondary outlier filtering, and obtain the maximum value after filtering as the clustering result; or select the class with the largest average value in the clustering results as the target array, perform probability density distribution calculation, and obtain the element with the largest probability density distribution value as the clustering result.

[0017] In a preferred embodiment, in step 4, VD L With VD 0 The difference between the virtual The formula is as follows:

[0018]

[0019] detaQ(k)=VD L (k)-q i (k)

[0020]

[0021] Where I is the current, t is the time, i is the time serial number, k is the serial number of the number of times the SOC changes, and SOC i (i) SOC i (i-1) represents the SOC at time i and i-1 respectively i Value, q i is the capacity value corresponding to the two SOC changes, detaQ is the compensation value of the lost capacity between the two SOC changes; Q L The updated capacity value is calculated for the Lth iteration, and detaQ is compensated only once when the SOC remains unchanged during the update.

[0022] In a preferred embodiment, the objective function of the mean square error J in step 5 above is as follows:

[0023] In a preferred embodiment, the objective function of the mean square error J in step 5 above can also be replaced by the sum of absolute values ​​of errors or the average value of absolute values ​​of errors.

[0024] In a preferred embodiment, the threshold m2 is greater than the threshold m1, and the threshold m1 is set to 0.01-0.05, and the threshold m2 is set to 0.02-0.05.

[0025] The solution algorithm of the present invention is applied to a new energy vehicle battery system or an energy storage battery system.

[0026] It can be seen from the above description of the present invention that, compared with the prior art, the present invention has the following advantages:

[0027] The present invention is based on the SOC calculation principle of the battery management system, and uses the change law of SOC to realize multiple iterations of the SOC denominator for accurate solution, and obtains the one-dimensional array solution value of the SOC denominator, which can be used to solve the compensation problem of the ampere-hour integral capacity in the data loss scenario, and greatly overcomes the influence of data loss on the ampere-hour integral capacity value when the battery has problems such as abnormal self-discharge and imbalance, so that accurate diagnosis and early warning of problems such as abnormal self-discharge and imbalance can still be achieved in the data loss scenario, thereby ensuring the accuracy of battery cloud diagnosis in this special scenario, improving the detection rate of battery health and safety hazards, and ensuring vehicle safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 Schematic diagram of the solution algorithm of the present invention.

[0029] Figure 2 A schematic diagram of the SOC denominator array obtained by using the solution algorithm of the present invention. DETAILED DESCRIPTION

[0030] Refer to the following Figure 1 The specific implementation of the present invention is described. In order to fully understand the present invention, many details are described below, but for those skilled in the art, the present invention can be implemented without these details. For well-known components, methods and processes, they are not described in detail below.

[0031] The embodiment of the present invention provides an algorithm for solving the SOC denominator array in the data loss mode, referring to Figure 1 , including the following steps:

[0032] 1. Collect vehicle current, time, and state of charge SOC i Data, the current and time data of the vehicle discharge process are integrated by ampere-hour to obtain the capacity array Q corresponding to each moment 1*N , and obtain SOC i The time Tc corresponding to the changed position 1*M , where N is the current array length, M is SOCi The number of times changes occurred.

[0033] 2. Calibrate the capacity values ​​Q(Tc(k+1)) and Q(Tc(k)) corresponding to the time Tc(k+1) and Tc(k). Subtract the capacity value Q(Tc(k+1)) at the time Tc(k) from the capacity value Q(Tc(k)) at the time Tc(k+1) to get qi(k). Calculate the q corresponding to all Tc times. i , that is, forming the initial SOC denominator array VD 0 .

[0034] 3. VD obtained in step 2 0 Clustering is performed, and the numerical value of the clustering result is marked as a L , a L Assign to VD L , where VD L All elements are a L .

[0035] 4. Using VD L With VD 0 The capacity value is corrected based on the difference between them, and the corresponding virtual one-dimensional array distribution state is calculated based on the corrected capacity value.

[0036] Specifically, in step 4 of this embodiment, VD is used L With VD 0 The difference between the virtual The formula is as follows:

[0037]

[0038] detaQ(k)=VD L (k)-q i (k)

[0039]

[0040] Where I is the current, t is the time, i is the time serial number, k is the serial number of the number of times the SOC changes, and SOC i (i) SOC i (i-1) represents the SOC at time i and i-1 respectively i Value, q i is the capacity value corresponding to the two SOC changes, detaQ is the compensation value of the lost capacity between the two SOC changes; Q L The updated capacity value is calculated for the Lth iteration, and detaQ is compensated only once when the SOC remains unchanged during the update.

[0041] 5. Computing Virtualization With real SOCi The mean square error J between them. Specifically, the calculation formula adopted by the mean square error J in this embodiment is as follows:

[0042] 6. When the mean square error J is greater than the threshold m1, search The position where the difference is greater than the threshold m2 is marked as k1; for VD L The elements from k1 to M in the cluster are re-clustered to obtain a L+1 , and a L+1 Assign to VD L+1 , where VD L+1 The element at the corresponding position from k1 to the last position is a L+1 , other position elements remain the same as VD L Same, L represents the number of iterations.

[0043] The threshold m2 is greater than the threshold m1. The threshold m1 can be set to 0.01-0.05 according to different application scenarios; and the threshold m2 can be set to 0.02-0.05 according to different application scenarios.

[0044] 7. Repeat steps 4 to 7 until the mean square error J is less than the threshold m1, and output the SOC denominator one-dimensional array VD G , where G is the final number of iterations.

[0045] Figure 2 The effect diagram of a specific embodiment of the SOC denominator array using the solution algorithm of the present invention is shown in FIG. In the figure, the "*" represents the lost SOC denominator, and the blue "·" represents the one-dimensional array of the SOC denominator obtained after 4 iterations.

[0046] The clustering in the above step 3 can select the average value of the class with the largest average value in the clustering result as the clustering result; or select the class with the largest average value in the clustering result as the target array, perform secondary outlier filtering, and obtain the maximum value after filtering as the clustering result; or select the class with the largest average value in the clustering result as the target array, perform probability density distribution calculation, and obtain the element with the largest probability density distribution value as the clustering result.

[0047] The objective function of the mean square error J in step 5 above may also be replaced by the sum of absolute values ​​of errors or the average value of absolute values ​​of errors.

[0048] The above is only a specific implementation of the present invention, but the design concept of the present invention is not limited to this. Any non-substantial changes to the present invention using this concept shall be deemed as an infringement of the protection scope of the present invention.

Claims

1. An algorithm for solving the SOC denominator array in the data loss mode, characterized in that: The following steps are involved: Step 1: Collect vehicle current, time, and state of charge (SOC) i Data, the current and time data of the vehicle discharge process are integrated by ampere-hour to obtain the capacity array Q corresponding to each moment 1*N , and obtain SOC i The time Tc corresponding to the changed position 1*M , where N is the current array length, M is SOC i The number of changes that occurred; Step 2: Calculate each SOC i The original capacity change corresponding to the changed position is used to obtain the initial denominator array VD 0 ; Step 3: VD obtained in step 2 0 Clustering is performed, and the numerical value of the clustering result is marked as a L , a L Assign value to VD L , where VD L All elements are a L ; Step 4: Use VD L With VD 0 The capacity value is corrected based on the difference between them, and the corresponding virtual Step 5: Calculate virtual With real SOC i The mean square error J between Step 6: When the mean square error J is greater than the threshold m1, search The position where the difference is greater than the threshold m2 is marked as k1; for VD L The elements from k1 to M in the cluster are re-clustered to obtain a L+1 , and a L+1 Assign value to VD L+1 , where VD L+1 The element at the corresponding position from k1 to the last position is a L+1 , other position elements remain the same as VD L Same, L represents the number of iterations; Step 7: Repeat steps 4 to 7 until the mean square error J is less than the threshold m1, and output the SOC denominator one-dimensional array VD G , where G is the final number of iterations.

2. The algorithm for solving the SOC denominator array in the data loss mode according to claim 1, characterized in that: The initial denominator array VD in step 2 0 The method of obtaining is as follows: calibrate the capacity values ​​Q(Tc(k+1)) and Q(Tc(k)) corresponding to the time Tc(k+1) and Tc(k), Q(Tc(k+1))-Q(Tc(k)) is q i (k), find the q corresponding to all Tc moments i , that is, forming the initial SOC denominator array VD 0 .

3. The algorithm for solving the SOC denominator array in the data loss mode according to claim 1, characterized in that: The clustering in step three selects the average value of the class with the largest average value in the clustering result as the clustering result; or selects the class with the largest average value in the clustering result as the target number array, performs secondary outlier filtering, and obtains the maximum value after filtering as the clustering result; or selects the class with the largest average value in the clustering result as the target array, performs probability density distribution calculation, and obtains the element with the largest probability density distribution value as the clustering result.

4. The algorithm for solving the SOC denominator array in the data loss mode according to claim 1, characterized in that: In step 4, VD is used L With VD 0 The difference between the virtual The formula is as follows: detaQ(k)=VD L (k)-q i (k) Where I is the current, t is the time, i is the time serial number, k is the serial number of the number of times the SOC changes, and SOC i (i) SOC i (i-1) represents the SOC at time i and i-1 respectively i value, qi is the capacity value corresponding to the two SOC changes, detaQ is the compensation value of the lost capacity between the two SOC changes; Q L The updated capacity value is calculated for the Lth iteration, and detaQ is compensated only once when the SOC remains unchanged during the update.

5. The algorithm for solving the SOC denominator array in the data loss mode according to claim 1, characterized in that: The objective function of the mean square error J in step 5 is as follows:

6. The algorithm for solving the SOC denominator array in the data loss mode according to claim 1, characterized in that: The objective function of the mean square error J in step 5 adopts the sum of absolute values ​​of errors or the average value of absolute values ​​of errors.

7. The algorithm for solving the SOC denominator array in the data loss mode according to claim 1, characterized in that: The threshold m2 is greater than the threshold m1, and the threshold m1 is set to 0.01-0.05, and the threshold m2 is set to 0.02-0.

05.

8. The algorithm for solving the SOC denominator array in the data loss mode according to claim 1, characterized in that: The solution algorithm is applied to new energy vehicle battery systems or energy storage battery systems.

Images