CN115656831A - Multi-step advanced prediction and fault diagnosis method for single battery voltage - Google Patents
Multi-step advanced prediction and fault diagnosis method for single battery voltage Download PDFInfo
- Publication number
- CN115656831A CN115656831A CN202211219981.XA CN202211219981A CN115656831A CN 115656831 A CN115656831 A CN 115656831A CN 202211219981 A CN202211219981 A CN 202211219981A CN 115656831 A CN115656831 A CN 115656831A
- Authority
- CN
- China
- Prior art keywords
- voltage
- cell
- battery
- fault
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 70
- 238000000034 method Methods 0.000 title claims abstract description 64
- 238000013528 artificial neural network Methods 0.000 claims abstract description 24
- 239000000178 monomer Substances 0.000 claims abstract description 23
- 238000007599 discharging Methods 0.000 claims abstract description 18
- 238000004140 cleaning Methods 0.000 claims abstract description 14
- 238000012544 monitoring process Methods 0.000 claims abstract description 12
- 238000005070 sampling Methods 0.000 claims description 36
- 238000012549 training Methods 0.000 claims description 29
- 238000012360 testing method Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 238000010219 correlation analysis Methods 0.000 claims description 3
- 238000010923 batch production Methods 0.000 claims 2
- 238000010586 diagram Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000032683 aging Effects 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 3
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 230000001681 protective effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Landscapes
- Secondary Cells (AREA)
Abstract
The invention discloses a battery monomer voltage multi-step advanced prediction and fault diagnosis method, which comprises the following steps: (1) Acquiring historical charging and discharging data of a real vehicle battery system from a cloud monitoring platform; (2) data cleaning; (3) selecting characteristic parameters; (4) constructing a data structure; (5) Establishing a battery monomer voltage prediction model based on a GRU neural network; (6) obtaining a multi-step advanced predicted value of the voltage of the single battery; (7) And obtaining the multi-step advanced diagnosis result of the voltage fault of the single battery. According to the method, a multi-step advance prediction method is used for constructing a data structure of multi-step advance prediction of the single battery voltage, the multi-step advance prediction of the single battery voltage is realized by combining an established single battery voltage prediction model based on a GRU neural network, and then diagnosis of the single battery voltage fault is realized by combining a formulated voltage fault diagnosis strategy.
Description
Technical Field
The invention relates to the field of fault diagnosis of battery management technology, in particular to a multi-step advance prediction and fault diagnosis method for single battery voltage.
Background
The accurate and advanced prediction of voltage and the diagnosis of voltage faults are beneficial to drivers and battery management systems to take protective measures, and property loss and passenger injury are reduced.
The current methods for predicting voltage and diagnosing voltage faults are mainly classified into a model-based method and a data-driven method. While these studies have achieved some success, these approaches tend to suffer from the following problems: first, the voltage prediction is real-time prediction or one-step prediction, and multi-step advanced prediction and diagnosis of battery voltage and voltage faults cannot be performed. Second, these methods cannot update the established model, but only apply it online as a fixed model after the model is trained offline. It is well known that batteries not only age during use, but also encounter unknown and complex driving conditions. The offline training model is difficult to fully consider aging and complex and variable environmental influences, and great challenges are brought to battery voltage fault detection.
Disclosure of Invention
The invention aims to provide a battery single voltage multi-step advance prediction and fault diagnosis method to solve the problems that multi-step advance prediction and diagnosis cannot be carried out on battery voltage and voltage faults and an established model cannot be updated in the prior art.
In order to solve the technical problems, the technical scheme of the invention is to provide a battery monomer voltage multi-step advance prediction and fault diagnosis method, and the innovation point is that the method comprises the following steps:
(1) Acquiring historical charging and discharging data of the real vehicle battery system from a cloud monitoring platform: acquiring historical charging and discharging data of a battery system of a real vehicle with three faults of single battery voltage overvoltage, voltage undervoltage and voltage change rate over-fast from a cloud monitoring platform;
(2) Data cleaning: performing data cleaning on the historical charging and discharging data of the real vehicle battery system obtained in the step (1);
(3) Selecting characteristic parameters: performing correlation analysis on the data cleaned in the step (2) by using a Pearson correlation coefficient method, and selecting a parameter of which the correlation coefficient with the voltage of the battery monomer is more than 0.9 as a characteristic parameter;
(4) And (3) constructing a data structure: constructing a data structure of multi-step advanced prediction of the single battery voltage by using a multi-step advanced prediction method;
(5) Establishing a battery monomer voltage prediction model based on a GRU neural network;
(6) Obtaining a multi-step advanced predicted value of the voltage of the battery monomer: training and predicting a single battery voltage prediction model based on a GRU neural network by using an incremental training principle to obtain a multi-step advanced prediction value of the single battery voltage;
(7) Obtaining a multi-step advanced diagnosis result of the voltage fault of the battery monomer: and (5) inputting the multi-step advanced prediction value of the voltage of the single battery obtained in the step (6) into a voltage fault diagnosis strategy to diagnose the voltage fault, and obtaining a multi-step advanced diagnosis result of the voltage fault of the single battery.
Further, the historical charging and discharging data of the battery system of the real vehicle, which is acquired from the cloud monitoring platform in the step (1), includes a single battery voltage CV, a total voltage TV, an SOC, a current, a vehicle speed, a temperature and a real vehicle accumulated mileage.
Further, the data cleaning in the step (2) includes removing abnormal data with adjacent SOC variation larger than 5%, and sequentially filling and replacing the data with sampling time step missing smaller than 6 steps with the calculated mean values of the previous and subsequent data of the data with sampling time step missing smaller than 6 steps.
Further, the characteristic parameters selected in the step (3) include total voltage TV and SOC.
Further, the constructing the data structure in the step (4) specifically includes the following steps:
(41) The characteristic parameters total voltage TV and SOC selected in the step (3) and the single battery voltage CV form the input X of the GRU neural network,
selecting the single battery voltage CV as the output Y of the GRU neural network,
where t represents the total length of the data;
(42) Constructing the input X and the output Y into a data structure of the multi-step advance prediction of the single battery voltage by using a lag time step M and a multi-step advance time step N in the multi-step advance prediction method to form a new input X and an output Y which are expressed as follows:
further, the hyper-parameters of the single voltage prediction model based on the GRU neural network established in the step (5) include the number of hidden layers, the number of nodes of the hidden layers, the number of batches, the training algebra epoch, the lag time step M in the multi-step lead prediction method, and the multi-step lead time step N, and the determination values of the number of hidden layers, the number of nodes of the hidden layers, the number of batches, the training algebra epoch, the lag time step M in the multi-step lead prediction method, and the multi-step lead time step N are 1, 32, 256, 30,6, respectively.
Further, the incremental training principle in step (6) is as follows: and selecting the first 60480 charge-discharge data within the month of the interval as incremental samples by a trial-and-error method every 1 month, training the single battery voltage prediction model, and testing the trained single battery voltage prediction model by using the charge-discharge data within the month after the incremental samples as a test set to obtain the multi-step advance prediction value of the single battery voltage.
Further, the step (7) of obtaining the multi-step advanced diagnosis result of the voltage fault of the battery cell specifically includes the following steps:
(71) According to the service instruction and the operation characteristics of the battery, voltage fault diagnosis strategies are formulated for three types of faults of overvoltage, undervoltage and over-high voltage change rate, and the voltage fault diagnosis strategies specifically comprise:
when the cell voltage U cell When the voltage is more than or equal to 4.2V and less than 4.7V, diagnosing that the voltage is over-voltage and 2-level fault;
when the cell voltage U cell When the voltage is more than or equal to 4.7V, the voltage overvoltage grade 1 fault is diagnosed；
When the cell voltage U cell When the voltage is more than 2.9V and less than or equal to 3.4V, diagnosing that the voltage is undervoltage and 3-level fault;
when the cell voltage U cell When the voltage is greater than 2.4V and less than or equal to 2.9V, diagnosing that the voltage is undervoltage and 2-level fault;
when the cell voltage U cell When the voltage is less than or equal to 2.4V, diagnosing that the voltage is under-voltage and the level 1 fault;
when absolute value | Δ U of cell voltage cell When the voltage is more than or equal to 0.4V/10s, diagnosing that the voltage change rate is over-fast level 1 fault;
(72) Calculating absolute value | delta U of the change rate of the cell voltage of adjacent sampling points by using the multi-step advance predicted value of the cell voltage obtained in the step (6) cell The calculation formula is: | Δ U cell |＝|U t+Δt -U t I/Δ t, wherein U t And U t+Δt Respectively representing a front sampling point and a rear sampling point, and delta t represents a sampling interval;
(73) Using the multi-step advance predicted value of the battery cell voltage obtained in the step (6) and the absolute value | Delta U of the cell voltage change rate obtained in the step (72) cell Inputting the voltage into the voltage fault diagnosis strategy formulated in the step (71) to judge the multistep advanced prediction value of the voltage of the battery monomer and the absolute value | delta U of the voltage change rate cell And if yes, diagnosing the type and the grade of the voltage fault according to the actual threshold interval in which the voltage fault is located, and thus obtaining the single battery voltage fault multi-step advanced diagnosis result.
Compared with the prior art, the battery cell voltage multi-step advanced prediction and fault diagnosis method has the following advantages:
(1) The invention uses the historical charging and discharging data of the real vehicle battery system, and overcomes the defects that the traditional method is simple in experimental test data working condition, low in complexity and insufficient in simulating the real vehicle operation scene.
(2) According to the invention, a multi-step advanced prediction method is used for constructing a multi-step advanced prediction data structure of the single battery voltage and combining the established single battery voltage prediction model based on the GRU neural network, so that the multi-step advanced prediction of the single battery voltage is realized, a guarantee is provided for the advanced diagnosis and early warning of the single battery voltage, and the safety of a battery system is improved.
(3) The invention can realize the update of the voltage prediction model by using the increment training principle, improves the capability of the model to adapt to the aging of the battery and the environmental change, ensures the fidelity of the model to be always kept at a higher level, and ensures the accuracy of the voltage prediction of the battery monomer and the reliability of fault diagnosis.
(4) The method provided by the invention uses data of the real vehicle battery system for modeling and verification, so that the method has strong adaptability and application potential, and can be directly transplanted to vehicles of the same type for real vehicle application.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the embodiments or the description of the prior art will be briefly described below.
Fig. 1 shows a flow chart of a multi-step advance prediction and fault diagnosis method for a battery cell voltage.
Fig. 2 is a diagram showing the diagnosis results of the Cell 1 Cell voltage multi-step advance prediction value and the overvoltage 2-level fault.
Fig. 3 shows a diagnosis result diagram of the Cell 2 Cell voltage multi-step advance prediction value and the undervoltage 3-level fault.
Fig. 4 is a diagram showing the absolute value of the voltage change rate of the Cell 1 Cell voltage multi-step advance prediction value and the diagnosis result of the level 1 fault with the voltage change rate being too fast.
Fig. 5 is a diagram showing the absolute value of the voltage change rate of the Cell 2 Cell voltage multi-step lead prediction value and the diagnosis result of a level 1 fault in which the voltage change rate is too fast.
Detailed Description
In order to make those skilled in the art better understand the technical solutions of the present invention, the technical solutions of the present invention will be clearly described below with reference to the accompanying drawings and embodiments.
The invention provides a battery single voltage multi-step advance prediction and fault diagnosis method, a flow chart of which is shown in figure 1, and the method comprises the following steps:
(1) Acquiring historical charging and discharging data of the real vehicle battery system from a cloud monitoring platform: historical charging and discharging data of a battery system of a real vehicle with three faults of single battery voltage overvoltage, voltage undervoltage and voltage change rate over fast are obtained from a cloud monitoring platform, and the charging and discharging data comprise single battery voltage CV, total voltage TV, SOC, current, vehicle speed, temperature and real vehicle accumulated mileage.
(2) Data cleaning: performing data cleaning on the historical charging and discharging data of the real vehicle battery system obtained in the step (1); the data cleaning comprises the steps of eliminating abnormal data with adjacent SOC changes larger than 5%, and sequentially filling and replacing the data with the sampling time step missing smaller than 6 steps by using the average values calculated by the previous data and the next data of the data with the sampling time step missing smaller than 6 steps.
(3) Selecting characteristic parameters: performing correlation analysis on the data cleaned in the step (2) by using a Pearson correlation coefficient method, and selecting a parameter with a correlation coefficient larger than 0.9 with the voltage of a battery monomer as a characteristic parameter; the selected characteristic parameters comprise total voltage TV and SOC.
(4) And (3) constructing a data structure: constructing a data structure of multi-step advanced prediction of the single battery voltage by using a multi-step advanced prediction method; wherein, constructing the data structure specifically comprises the following steps:
(41) The characteristic parameters total voltage TV and SOC selected in the step (3) and the single battery voltage CV form the input X of the GRU neural network,
selecting the single battery voltage CV as the output Y of the GRU neural network,
where t represents the total length of the data;
(42) Constructing the input X and the output Y into a data structure of the multi-step advance prediction of the single battery voltage by using a lag time step M and a multi-step advance time step N in the multi-step advance prediction method to form a new input X and an output Y which are expressed as follows:
(5) Establishing a battery monomer voltage prediction model based on a GRU neural network: the established super-parameters of the single voltage prediction model based on the GRU neural network comprise the number of hidden layers, the number of hidden layer nodes, the number of batch processing, an epoch of training algebra, a lag time step M in the multi-step lead prediction method and a multi-step lead time step N, and the determination values of the number of hidden layers, the number of hidden layer nodes, the number of batch processing, the number of training algebra epoch, the lag time step M in the multi-step lead prediction method and the multi-step lead time step N are respectively 1, 32, 256, 30 and 6.
(6) Obtaining a multi-step advanced predicted value of the voltage of the battery monomer: training and predicting a battery monomer voltage prediction model based on a GRU neural network by using an incremental training principle to obtain a multi-step advanced prediction value of the battery monomer voltage; the principle of incremental training of the invention is as follows: every 1 month, selecting the first 60480 charge-discharge data within the month of the interval as an incremental sample by a trial-and-error method, training the single battery voltage prediction model, and testing the trained single battery voltage prediction model by using the charge-discharge data one month after the incremental sample as a test set to obtain the multi-step advance prediction value of the single battery voltage.
(7) Obtaining a multi-step advanced diagnosis result of the voltage fault of the single battery: inputting the multi-step advanced prediction value of the voltage of the single battery obtained in the step (6) into a voltage fault diagnosis strategy to diagnose the voltage fault, and obtaining a multi-step advanced diagnosis result of the voltage fault of the single battery, wherein the specific process is as follows:
(71) According to the use description and the operation characteristics of the battery, voltage fault diagnosis strategies are formulated for three types of faults of overvoltage, undervoltage and over-high voltage change rate, and the voltage fault diagnosis strategies specifically comprise:
when the cell voltage U cell 4.2V or more andwhen the voltage is less than 4.7V, the voltage overvoltage 2-level fault is diagnosed;
when the cell voltage U cell When the voltage is more than or equal to 4.7V, diagnosing the fault as a voltage overvoltage level 1 fault;
when the cell voltage U cell When the voltage is more than 2.9V and less than or equal to 3.4V, diagnosing that the voltage is undervoltage and 3-level fault;
when the cell voltage U cell When the voltage is greater than 2.4V and less than or equal to 2.9V, diagnosing that the voltage is undervoltage and 2-level fault;
when the cell voltage U cell When the voltage is less than or equal to 2.4V, diagnosing that the voltage is under-voltage and the level 1 fault;
when absolute value | Δ U of cell voltage cell When | is greater than or equal to 0.4V/10s, diagnosing that the voltage change rate is over-fast 1-level fault;
(72) Calculating absolute value | delta U of the change rate of the cell voltage of adjacent sampling points by using the multi-step advance predicted value of the cell voltage obtained in the step (6) cell And the calculation formula is as follows: | Δ U cell |＝|U t+Δt -U t L/Δ t, wherein U t And U t+Δt Respectively representing a front sampling point and a rear sampling point, and delta t represents a sampling interval;
(73) Using the multi-step advance predicted value of the battery cell voltage obtained in the step (6) and the absolute value | Delta U of the cell voltage change rate obtained in the step (72) cell I, inputting the voltage into the voltage fault diagnosis strategy formulated in the step (71) to judge the multistep advanced prediction value of the voltage of the battery monomer and the absolute value | delta U of the voltage change rate cell And if yes, diagnosing the type and the grade of the voltage fault according to the actual threshold interval in which the voltage fault is located, and thus obtaining the single battery voltage fault multi-step advanced diagnosis result.
The foregoing is a specific technical solution of the present invention, and the following is a specific implementation manner and implementation process performed on the basis of the above technical solution.
The implementation method comprises the steps of screening out a single battery Cell 1 with two types of faults including a single voltage overvoltage 2-level fault and a voltage change rate over-fast 1-level fault from a cloud monitoring platform, and a single battery Cell 2 with two types of faults including a single voltage undervoltage 3-level fault and a voltage change rate over-fast 1-level fault as an implementation object, wherein the data lengths of two batteries are the same, and the implementation method comprises the following specific implementation processes:
(1) Acquiring historical charging and discharging data of the real vehicle battery system from a cloud monitoring platform: historical charging and discharging data of the battery Cell 1 and the battery Cell 2 are obtained from the cloud monitoring platform and comprise battery Cell voltage CV, total voltage TV, SOC, current, vehicle speed, temperature and actual vehicle accumulated mileage.
(2) Data cleaning: and performing data cleaning on historical charging and discharging data of the battery Cell 1 and the battery Cell 2 with voltage faults, wherein the data cleaning comprises the steps of eliminating abnormal data with adjacent SOC change larger than 5%, and sequentially filling and replacing the data with the sampling time step missing smaller than 6 steps by using the front and back data calculation mean values of the data with the sampling time step missing smaller than 6 steps.
(3) Selecting characteristic parameters: and selecting parameters with high voltage correlation with the battery cells from the data of the battery Cell 1 and the battery Cell 2 after data cleaning by using a Pearson correlation coefficient method as characteristic parameters, wherein the characteristic parameters selected by the battery Cell 1 and the battery Cell 2 comprise total voltage TV and SOC.
(4) And (3) constructing a data structure: the method for constructing the multi-step advanced prediction data structure of the Cell voltages of the Cell 1 and the Cell 2 by using the multi-step advanced prediction method specifically comprises the following steps:
(41) The characteristic parameters total voltage TV and SOC of the battery Cell Cell 1 and the battery Cell Cell 2 selected in the step (3) and the battery Cell voltage CV form an input X of a GRU neural network,
selecting the single battery voltage CV as the output Y of the GRU neural network,
wherein t represents the total length of the data;
(42) Constructing the input X and the output Y into a battery cell voltage multi-step advance prediction data structure by using a lag time step M and a multi-step advance time step N in the multi-step advance prediction method, and forming new input X and output Y which are expressed as follows:
(5) The method comprises the steps of establishing a single voltage prediction model based on a GRU neural network, wherein the over-parameters of the established single voltage prediction model based on the GRU neural network comprise the number of hidden layers, the number of nodes of the hidden layers, the number of batch processing, a training algebra epoch, a lag time step M in a multi-step advance prediction method and a multi-step advance time step N, and the determination values of the number of the hidden layers, the number of nodes of the hidden layers, the number of batch processing, the training algebra epoch, the lag time step M in the multi-step advance prediction method and the multi-step advance time step N are respectively 1, 32, 256, 30 and 6.
(6) Obtaining a multi-step advance predicted value of the voltage of the battery monomer: training and predicting a single voltage prediction model based on a GRU neural network by using an increment training principle to obtain a multi-step advanced prediction value of the single battery voltage, wherein the increment training principle is as follows: every 1 month, selecting the first 60480 charge-discharge data within the month of the interval as an incremental sample by a trial-and-error method, training the single battery voltage prediction model, and testing the trained single battery voltage prediction model by using the charge-discharge data one month after the incremental sample as a test set to obtain the multi-step advance prediction value of the single battery voltage. In this embodiment, each of the Cell cells Cell 1 and Cell 2 has 4 months of charge and discharge data, and the Cell cells Cell 1 and Cell 2 can respectively train and test the model for 3 times according to the incremental training principle. As shown in fig. 2 and 3, the model is subjected to 3 incremental training times for each of the Cell 1 and the Cell 2, and then the Cell voltage multi-step advanced prediction value and the true value are compared by using the charging and discharging data of the Cell 1 and the Cell 2 in the 4 th month for testing, so that the prediction value and the true value are basically overlapped, and the method is proved to be capable of accurately realizing the multi-step advanced prediction of the Cell voltage.
(7) Obtaining a multi-step advanced diagnosis result of the voltage fault of the battery monomer: inputting the multi-step advanced predicted values of the Cell voltages of the Cell 1 and the Cell 2 obtained in the step (6) into a voltage fault diagnosis strategy to diagnose the voltage fault, so as to obtain a multi-step advanced diagnosis result of the Cell voltage fault, which specifically comprises the following steps:
(71) According to the service instruction and the operation characteristics of the battery, voltage fault diagnosis strategies are formulated for three types of faults of overvoltage, undervoltage and over-high voltage change rate, and the voltage fault diagnosis strategies specifically comprise:
when the cell voltage U cell When the voltage is greater than or equal to 4.2V and less than 4.7V, diagnosing that the voltage is over-voltage and 2-level fault;
when the cell voltage U cell When the voltage is greater than or equal to 4.7V, the voltage overvoltage level 1 fault is diagnosed;
when the cell voltage U cell When the voltage is more than 2.9V and less than or equal to 3.4V, diagnosing that the voltage is under-voltage and 3-level fault;
when the cell voltage U cell When the voltage is greater than 2.4V and less than or equal to 2.9V, diagnosing that the voltage is undervoltage and 2-level fault;
when the cell voltage U cell When the voltage is less than or equal to 2.4V, diagnosing that the voltage is under-voltage and the level 1 fault;
when absolute value of cell voltage | Δ U cell When the voltage is more than or equal to 0.4V/10s, diagnosing that the voltage change rate is over-fast level 1 fault;
(72) Calculating absolute value | delta U of Cell voltage change rate of adjacent sampling points by using the multi-step advanced predicted values of the Cell voltages of the Cell 1 and the Cell 2 obtained in the step (6) cell The calculation formula is: | Δ U cell |＝|U t+Δt -U t I/Δ t, wherein U t And U t+Δt Respectively representing a front sampling point and a rear sampling point, and delta t represents a sampling interval;
(73) Using the battery obtained in the step (6)The multi-step advance prediction values of the Cell voltages of the Cell 1 and the Cell 2 and the absolute value | Δ U of the Cell voltage change rates of the Cell 1 and the Cell 2 obtained in the step (72) cell I, inputting the voltage into the voltage fault diagnosis strategy formulated in the step (71) to judge the multistep advance predicted values and the absolute values | delta U of the voltage change rates of the battery Cell voltages of the battery Cell 1 and the battery Cell 2 cell And if yes, diagnosing the type and the grade of the voltage fault according to the actual threshold interval in which the voltage fault is located, and thus obtaining the single battery voltage fault multi-step advanced diagnosis result.
As shown in fig. 2, when the multi-step advance prediction value of the Cell voltage of the Cell 1 is input into the voltage failure diagnosis policy formulated in step (71), it can be determined that the multi-step advance prediction value of the voltage of the Cell 1 from the sampling point 2602 to the sampling point 2604 is in the threshold interval where the voltage overvoltage 2-level failure is located, so that the voltage overvoltage 2-level failure is diagnosed from the sampling point 2602 to the sampling point 2604, and the diagnosis result is the same as the true voltage overvoltage 2-level failure of the Cell 1.
As shown in fig. 3, when the multi-step advance prediction value of the Cell voltage of the Cell 2 is input into the voltage failure diagnosis policy formulated in step (71), it can be determined that the multi-step advance prediction value of the voltage of the Cell 2 from the sampling point 2547 to the sampling point 2604 is in the threshold interval where the voltage under-voltage 3-level failure is located, so that the voltage under-voltage 3-level failure is diagnosed from the sampling point 2547 to the sampling point 2604, and the diagnosis result is the same as the real voltage under-voltage 3-level failure of the Cell 2.
As shown in fig. 4, the absolute value | Δ U of the voltage change rate of the battery Cell 1 cell If | is input into the voltage fault diagnosis strategy formulated in step (71), the absolute value | Δ U of the voltage change rate of the Cell 1 at the sampling point 2062 and the sampling point 2605 can be determined cell I is in the threshold interval where the level 1 fault with too fast voltage change rate is located, so sample point 2062 and sample point 2605 are diagnosed as voltage changesAnd the diagnosis result is the same as the real voltage change rate of the Cell 1, namely, the 1-level fault with the over-fast transformation rate.
As shown in fig. 5, the absolute value | Δ U of the voltage change rate of the battery Cell 2 cell | input into the voltage fault diagnosis strategy formulated in step (71), the absolute value | Δ U of the voltage change rate of the Cell 2 sampling point 2605 can be determined cell I is in the threshold interval where the voltage change rate is too fast for the level 1 fault, so the sampling point 2605 is diagnosed as the level 1 fault with too fast voltage change rate, and the diagnosis result is the same as the true level 1 fault with too fast voltage change rate of Cell 2.
In summary, the multi-step advance prediction and fault diagnosis method for the single battery voltage provided by the invention adopts a multi-step advance prediction method to construct a data structure of the multi-step advance prediction of the single battery voltage and combines the established single battery voltage prediction model based on the GRU neural network to realize the multi-step advance prediction of the single battery voltage. In addition, the invention can realize the update of the voltage prediction model by using the increment training principle, improves the capability of the model to adapt to the aging of the battery and the environmental change, ensures the fidelity of the model to be always kept at a higher level, and ensures the accuracy of the multi-step advanced prediction of the single battery voltage. Finally, the three faults of voltage overvoltage, voltage undervoltage and voltage change rate over-high of the single battery are accurately diagnosed by combining the formulated fault diagnosis strategy. The method provided by the invention provides a new idea for the advanced diagnosis of the voltage fault.
Claims (8)
1. A battery cell voltage multi-step advance prediction and fault diagnosis method is characterized by comprising the following steps:
(1) Acquiring historical charging and discharging data of the real vehicle battery system from a cloud monitoring platform: acquiring historical charging and discharging data of a battery system of a real vehicle with three faults of single battery voltage overvoltage, voltage undervoltage and voltage change rate over-fast from a cloud monitoring platform;
(2) Data cleaning: performing data cleaning on the historical charging and discharging data of the real vehicle battery system obtained in the step (1);
(3) Selecting characteristic parameters: performing correlation analysis on the data cleaned in the step (2) by using a Pearson correlation coefficient method, and selecting a parameter of which the correlation coefficient with the voltage of the battery monomer is more than 0.9 as a characteristic parameter;
(4) And (3) constructing a data structure: constructing a data structure of multi-step advanced prediction of the single battery voltage by using a multi-step advanced prediction method;
(5) Establishing a battery monomer voltage prediction model based on a GRU neural network;
(6) Obtaining a multi-step advance predicted value of the voltage of the battery monomer: training and predicting a battery monomer voltage prediction model based on a GRU neural network by using an incremental training principle to obtain a multi-step advanced prediction value of the battery monomer voltage;
(7) Obtaining a multi-step advanced diagnosis result of the voltage fault of the single battery: and (5) inputting the multi-step advanced prediction value of the voltage of the single battery obtained in the step (6) into a voltage fault diagnosis strategy to diagnose the voltage fault, and obtaining a multi-step advanced diagnosis result of the voltage fault of the single battery.
2. The battery cell voltage multi-step advanced prediction and fault diagnosis method according to claim 1, wherein the historical charging and discharging data of the battery system of the real vehicle obtained from the cloud monitoring platform in the step (1) comprises battery cell voltage CV, total voltage TV, SOC, current, vehicle speed, temperature and accumulated mileage of the real vehicle.
3. The multi-step advanced prediction and fault diagnosis method for the battery cell voltage as claimed in claim 1, wherein the data cleaning in the step (2) comprises eliminating abnormal data with adjacent SOC variation larger than 5%, and sequentially filling and replacing the data with sampling time step missing smaller than 6 steps by using the calculated mean values of the previous and subsequent data for the data with sampling time step missing smaller than 6 steps.
4. The battery cell voltage multi-step advanced prediction and fault diagnosis method according to claim 1, wherein the characteristic parameters selected in the step (3) comprise total voltage TV and SOC.
5. The battery cell voltage multi-step advance prediction and fault diagnosis method according to claim 1, wherein the step (4) of constructing a data structure specifically comprises the following steps:
(41) The characteristic parameters total voltage TV and SOC selected in the step (3) and the single battery voltage CV form the input X of the GRU neural network,
selecting the single battery voltage CV as the output Y of the GRU neural network,
where t represents the total length of the data;
(42) Constructing the input X and the output Y into a data structure of the multi-step advance prediction of the single battery voltage by using a lag time step M and a multi-step advance time step N in the multi-step advance prediction method to form a new input X and an output Y which are expressed as follows:
6. the method as claimed in claim 1, wherein the super-parameters of the GRU neural network-based cell voltage prediction model established in the step (5) include the number of hidden layers, the number of hidden layer nodes, the number of batch processes, a training algebra epoch, a lag time step M in the multi-step lead prediction method, and a multi-step lead time step N, and the determination values of the number of hidden layers, the number of hidden layer nodes, the number of batch processes, the training algebra epoch, the lag time step M in the multi-step lead prediction method, and the multi-step lead time step N are 1, 32, 256, 30,6, respectively.
7. The battery cell voltage multi-step advance prediction and fault diagnosis method according to claim 1, wherein the incremental training principle in the step (6) is as follows: every 1 month, selecting the first 60480 charge-discharge data within the month of the interval as an incremental sample by a trial-and-error method, training the single battery voltage prediction model, and testing the trained single battery voltage prediction model by using the charge-discharge data one month after the incremental sample as a test set to obtain the multi-step advance prediction value of the single battery voltage.
8. The battery cell voltage multi-step advance prediction and fault diagnosis method according to claim 1, wherein the obtaining of the battery cell voltage fault multi-step advance diagnosis result in the step (7) specifically comprises the following steps:
(71) According to the service instruction and the operation characteristics of the battery, voltage fault diagnosis strategies are formulated for three types of faults of overvoltage, undervoltage and over-high voltage change rate, and the voltage fault diagnosis strategies specifically comprise:
when the cell voltage U cell When the voltage is greater than or equal to 4.2V and less than 4.7V, diagnosing that the voltage is over-voltage and 2-level fault;
when the cell voltage U cell When the voltage is greater than or equal to 4.7V, the voltage overvoltage level 1 fault is diagnosed;
when the cell voltage U cell When the voltage is more than 2.9V and less than or equal to 3.4V, diagnosing that the voltage is under-voltage and 3-level fault;
when the cell voltage U cell When the voltage is greater than 2.4V and less than or equal to 2.9V, diagnosing that the voltage is undervoltage and 2-level fault;
when the cell voltage U cell When the voltage is less than or equal to 2.4V, diagnosing that the voltage is under-voltage and the level 1 fault;
when absolute value | Δ U of cell voltage cell When | is greater than or equal to 0.4V/10s, diagnosing that the voltage change rate is over-fast 1-level fault;
(72) Calculating absolute value | delta U of the change rate of the cell voltage of adjacent sampling points by using the multi-step advance predicted value of the cell voltage obtained in the step (6) cell And the calculation formula is as follows: | Δ U cell |＝|U t+Δt -U t I/Δ t, wherein U t And U t+Δt Respectively representing a front sampling point and a rear sampling point, and delta t represents a sampling interval;
(73) Using the multi-step advance predicted value of the battery cell voltage obtained in the step (6) and the absolute value | Delta U of the cell voltage change rate obtained in the step (72) cell I, inputting the voltage into the voltage fault diagnosis strategy formulated in the step (71) to judge the multistep advanced prediction value of the voltage of the battery monomer and the absolute value | delta U of the voltage change rate cell And if yes, diagnosing the type and the grade of the voltage fault according to the actual threshold interval in which the voltage fault is located, and thus obtaining a multi-step advanced diagnosis result of the voltage fault of the single battery.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211219981.XA CN115656831A (en) | 2022-10-08 | 2022-10-08 | Multi-step advanced prediction and fault diagnosis method for single battery voltage |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211219981.XA CN115656831A (en) | 2022-10-08 | 2022-10-08 | Multi-step advanced prediction and fault diagnosis method for single battery voltage |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115656831A true CN115656831A (en) | 2023-01-31 |
Family
ID=84984623
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211219981.XA Pending CN115656831A (en) | 2022-10-08 | 2022-10-08 | Multi-step advanced prediction and fault diagnosis method for single battery voltage |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115656831A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117630683A (en) * | 2024-01-25 | 2024-03-01 | 北京科技大学 | Multi-scale fusion GRU network-based automobile battery SOC multi-step prediction method and system |
-
2022
- 2022-10-08 CN CN202211219981.XA patent/CN115656831A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117630683A (en) * | 2024-01-25 | 2024-03-01 | 北京科技大学 | Multi-scale fusion GRU network-based automobile battery SOC multi-step prediction method and system |
| CN117630683B (en) * | 2024-01-25 | 2024-03-29 | 北京科技大学 | Multi-scale fusion GRU network-based automobile battery SOC multi-step prediction method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110133508B (en) | Safety early warning method for power battery of electric automobile | |
| Hong et al. | Online accurate state of health estimation for battery systems on real-world electric vehicles with variable driving conditions considered | |
| CN111624494B (en) | Battery analysis method and system based on electrochemical parameters | |
| Farmann et al. | A comprehensive review of on-board State-of-Available-Power prediction techniques for lithium-ion batteries in electric vehicles | |
| CN112904219B (en) | Big data-based power battery health state prediction method | |
| Kim et al. | Estimation of Li-ion battery state of health based on multilayer perceptron: As an EV application | |
| CN110667434A (en) | Working condition-adaptive pure electric vehicle driving mileage estimation method and system | |
| CN110412470B (en) | SOC estimation method for power battery of electric vehicle | |
| CN112098874B (en) | Lithium ion battery electric quantity prediction method considering aging condition | |
| CN112782594B (en) | Method for estimating SOC (state of charge) of lithium battery by data-driven algorithm considering internal resistance | |
| CN116609676B (en) | Method and system for monitoring state of hybrid energy storage battery based on big data processing | |
| CN112611976A (en) | Power battery state of health estimation method based on double differential curves | |
| CN115656831A (en) | Multi-step advanced prediction and fault diagnosis method for single battery voltage | |
| CN114167301A (en) | Power battery evaluation method based on real vehicle data of electric vehicle | |
| CN114462259A (en) | SOC-based battery remaining life prediction method and system, automobile and medium | |
| CN112860782A (en) | Pure electric vehicle driving range estimation method based on big data analysis | |
| WO2021052540A1 (en) | Condition value for rechargeable batteries | |
| CN114036647A (en) | Power battery safety risk assessment method based on real vehicle data | |
| CN113791351B (en) | Lithium battery life prediction method based on transfer learning and difference probability distribution | |
| CN115097312A (en) | Lithium ion battery fusion life prediction model combining data driving model and empirical model | |
| Li et al. | A novel method for lithium-ion battery fault diagnosis of electric vehicle based on real-time voltage | |
| CN117272783A (en) | Power battery life prediction method based on cloud edge collaborative multi-model fusion | |
| CN115469226A (en) | Real-time safety early warning method for electric vehicle power battery based on operation big data | |
| CN114881395A (en) | Electric vehicle charging and discharging abnormity detection method based on multivariate Gaussian distribution model | |
| CN116080470A (en) | Power battery monitoring method and device for electric automobile, server and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |