CN117949851A - Battery state joint estimation method and system - Google Patents
Battery state joint estimation method and system Download PDFInfo
- Publication number
- CN117949851A CN117949851A CN202410133141.4A CN202410133141A CN117949851A CN 117949851 A CN117949851 A CN 117949851A CN 202410133141 A CN202410133141 A CN 202410133141A CN 117949851 A CN117949851 A CN 117949851A
- Authority
- CN
- China
- Prior art keywords
- battery
- health
- state
- model
- capacity
- 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
- 238000000034 method Methods 0.000 title claims abstract description 45
- 230000036541 health Effects 0.000 claims abstract description 151
- 230000032683 aging Effects 0.000 claims abstract description 32
- 230000015556 catabolic process Effects 0.000 claims abstract description 7
- 238000006731 degradation reaction Methods 0.000 claims abstract description 7
- 230000008859 change Effects 0.000 claims description 45
- 238000007600 charging Methods 0.000 claims description 31
- 230000002159 abnormal effect Effects 0.000 claims description 24
- 238000012549 training Methods 0.000 claims description 18
- 230000009467 reduction Effects 0.000 claims description 17
- 206010011906 Death Diseases 0.000 claims description 10
- 238000007599 discharging Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 3
- 230000036962 time dependent Effects 0.000 claims 1
- 230000001737 promoting effect Effects 0.000 abstract 1
- 230000007423 decrease Effects 0.000 description 26
- 230000008569 process Effects 0.000 description 12
- 238000013461 design Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000010280 constant potential charging Methods 0.000 description 3
- 238000010277 constant-current charging Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 239000003990 capacitor Substances 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 238000004146 energy storage Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000010438 heat treatment Methods 0.000 description 2
- 230000007787 long-term memory Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000000691 measurement method Methods 0.000 description 2
- 230000015654 memory Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010281 constant-current constant-voltage charging Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012983 electrochemical energy storage Methods 0.000 description 1
- 239000003792 electrolyte Substances 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000000178 monomer Substances 0.000 description 1
- 239000007774 positive electrode material Substances 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Landscapes
- Secondary Cells (AREA)
- Charge And Discharge Circuits For Batteries Or The Like (AREA)
Abstract
The invention relates to a battery state joint estimation method, in particular to the technical field of batteries. The method comprises the following steps: collecting direct health factors and indirect health factors representing the health ageing state of the battery in the battery capacity degradation data; constructing a first model, taking a direct health factor as input and taking a battery health state as output; calculating a first predicted battery capacity according to the battery state of health and the rated capacity, acquiring a corresponding OCV-SOC curve, and calculating a battery resistance-capacitance parameter for updating an initial value of a UKF algorithm to obtain a residual electric quantity state; and constructing a second model, taking the indirect health factor as input, outputting a second predicted battery capacity sequence, and obtaining a cycle period value corresponding to the battery capacity with the rated capacity of 70% to obtain the residual service life. The LSTM-RNN is used for not only realizing accurate prediction of the remaining service life and the battery health state of the battery, but also promoting accurate prediction of the remaining electric quantity state, so that the management of the battery state is more refined.
Description
Technical Field
The invention relates to the technical field of batteries, in particular to a battery state joint estimation method and system.
Background
As the population increases and the human living standard increases, the energy consumption increases rapidly. The world energy revolution has become one of the core strategic issues in today's large countries. In the context of energy structure transformation, wind power and photovoltaic are representative renewable energy sources which are environment-friendly and sustainable are rapidly developed in recent years. In recent years, with the popularization and popularization of new energy facilities, various forms of energy storage systems are increasing. But the news about major accidents of the electrochemical energy storage system at home and abroad are also endangered while the whole society is benefited, and serious life safety and economic loss are caused. Therefore, in the design of the energy storage system, the design of the BMS (Battery MANAGEMENT SYSTEM) having fault diagnosis of overcurrent, ultra-high temperature, short-circuit fault, etc. is also increasingly important.
The merits of the BMS depend on the effective detection of operating parameters such as battery voltage, current and temperature and the accurate evaluation of core parameters such as SOC (State of Charge), SOH (Stateof Health, battery State of health), RUL (REMAINING USEFUL LIFE, battery remaining life), and Power State (SOP). Wherein SOC, SOH, and RUL are key indexes for evaluating BMS. The estimation of the residual electric quantity state is widely based on a UKF algorithm (Unscented KALMAN FILTER, lossless Kalman filtering) at present, and the residual electric quantity state is predicted by constructing a state equation of the residual electric quantity state, the battery terminal voltage, the battery terminal current and the ohmic internal resistance of the battery by using an equivalent circuit model of the battery. However, as the battery ages, the electrochemical efficiency of the battery decreases, and the capacity of the battery also decreases, so that the resistance-capacitance parameter RC in the equivalent circuit model of the battery correspondingly changes, and the accuracy of the residual charge state estimation decreases. Therefore, on the basis of estimating the residual electric quantity state by the existing UKF algorithm, the influence of battery aging on the resistance-capacitance parameter is taken into consideration, so that the prediction accuracy of the residual electric quantity state is improved.
Disclosure of Invention
The invention aims at: aiming at the problem that the prediction accuracy is difficult to further improve due to incomplete consideration of influencing factors when the residual electric quantity state is predicted at present, the battery state joint estimation method and system are provided, the influence of the battery health state and the residual service life on the residual electric quantity state is comprehensively considered, and the prediction accuracy of the residual electric quantity state is improved.
In order to achieve the above object, the present invention provides a method for jointly estimating battery states, comprising the steps of:
collecting direct health factors and corresponding battery health states of the battery under different cycle periods, and recording the direct health factors and the corresponding battery health states as a first training set;
Constructing a first model, taking the direct health factor as input, taking the battery health state as output, and training the first model; the first model is a battery state of health aging model;
Inputting the direct health factor into the trained first model to obtain the battery health state; calculating the battery health state and rated battery capacity to obtain a first predicted battery capacity; acquiring an OCV-SOC curve graph according to the first predicted battery capacity; substituting the OCV value and the SOC value which are in one-to-one correspondence in the graph into a battery equivalent circuit equation, updating the resistance-capacitance parameter in the UKF algorithm, and calculating the state of remaining battery charge.
As a preferred technical scheme of the present application, the method further comprises the step of estimating the remaining service life of the battery, and the step comprises:
Collecting charging or discharging data under different cycle periods in the battery capacity degradation data, and extracting indirect health factors related to the battery capacity; the cycle period corresponds to the indirect health factor one by one, and the indirect health factor is used for representing the service life aging state of the battery;
Modeling according to the direct proportion relation between the indirect health factor and the battery capacity, and constructing a second model; taking the indirect health factor as input, taking second predicted battery capacity as output, and training the second model, wherein the second model is a battery residual service life aging model;
and collecting the indirect health factor input trained second model under the previous cycle period, and continuously outputting the corresponding second predicted battery capacity under all future battery cycle periods until reaching an end-of-life threshold, wherein the cycle period value corresponding to the end-of-life threshold is the residual service life of the battery.
As a preferable technical scheme of the application, the indirect health factor comprises equal voltage drop discharge time, average discharge voltage, constant current charge time and constant voltage charge time.
As a preferable technical scheme of the application, the battery health state and the residual service life adopt a slow scale mode, and estimation is carried out once when the battery health state and the residual service life reach integer multiples of a prediction period threshold value.
As a preferable technical scheme of the application, the estimation of the residual electric quantity state is carried out continuously in the process of charging or discharging the battery in a fast scale mode.
As a preferable technical scheme of the application, under the condition that the preset condition of a laboratory is constant, taking the time that the resistance-capacitance parameter in the equivalent circuit of the battery can be regarded as constant as a period, collecting different actual battery capacities of the battery under different periods; and meanwhile, intermittent constant-current discharge is carried out on the battery, and the discharge voltage and the residual electric quantity state are recorded in real time, so that the OCV-SOC curve graphs corresponding to a plurality of different actual battery capacities are obtained.
The method comprises the steps of calculating a first difference value of the residual service life predicted values of two adjacent times, and obtaining a service life reduction rate according to the change relation of the first difference value along with time; calculating a second difference value of the life reduction rate of two adjacent times, and obtaining a life change rate according to the change relation of the second difference value along with time; and adjusting the prediction period threshold according to the change proportion of the service life change rate of two adjacent times.
The method comprises the steps of obtaining a battery with the lowest voltage when the starting time difference value of equalization control is smaller than a preset deviation threshold value, and recording the battery as a first abnormal battery; collecting the historical battery health state of the first abnormal battery, and calculating a first variance; collecting the current battery health state of the first abnormal battery, and calculating a second variance; calculating variance variation amounts of the first variance and the second variance; and when the variance variation is larger than a variance variation threshold, the first abnormal battery is disconnected.
A battery state joint estimation system, comprising:
The data acquisition module is used for acquiring direct health factors of the battery health ageing state and corresponding battery health states under different cycle periods and recording the direct health factors and the corresponding battery health states as a first training set;
The model training module is used for constructing a first model, inputting the direct health factor into the first model and outputting the health state of the battery; the first model is a battery state of health aging model;
The data processing module is used for inputting the direct health factor into the trained first model to obtain the battery health state; calculating the battery health state and rated battery capacity to obtain a first predicted battery capacity; acquiring an OCV-SOC curve graph according to the first predicted battery capacity; substituting the OCV value and the SOC value which are in one-to-one correspondence in the graph into a battery equivalent circuit equation, updating the resistance-capacitance parameter in the UKF algorithm, and calculating the state of remaining battery charge.
Compared with the prior art, the invention has the beneficial effects that:
1. According to the invention, the battery state of health and the residual service life can be used for evaluating the battery aging state, and the battery aging state can influence the voltage, current and the change degree of the internal resistance of the battery, so that the factors need to be recalibrated when the residual electric quantity state is evaluated, and the accuracy and the reliability of the residual electric quantity state are improved. Meanwhile, the battery state of health and the residual service life are accurately predicted, estimation errors caused by independently estimating the state of the residual electric quantity can be reduced, and the cumulative effect of the errors in the propagation process is reduced.
2. In the invention, a plurality of indirect health factors which are strongly related to the health state of the battery are comprehensively collected as the input of the aging model, so that the reliability of the system is enhanced. In actual use, the estimation of the state of charge is continued throughout the charging or discharging process. Meanwhile, when current passes through the internal resistance of the battery and the internal resistance of the line, heat loss generally occurs to cause the temperature of the battery to rise, so that the estimation of the state of residual electric charge is affected. According to the aging model, an LSTM-RNN long-short-term memory cyclic convolution network is adopted, and a hidden layer at the previous moment is used as a part of input at the current moment in calculation, so that the current layer can fuse information characteristics extracted at the previous moment, the relevance of the state of the residual electric quantity between continuous moments is maintained, and the accuracy of prediction can be improved. Meanwhile, the method can adapt to more complex environments, and better processes uncertainty and interference by comprehensively considering a plurality of factors, so as to provide a more stable estimation result.
3. When the key features are updated, different time strategies are adopted according to different change degrees of the key features. The method for updating the residual electric quantity state in a fast scale achieves the real-time performance of prediction, the method for updating the battery state of health and the residual service life in a slow scale reduces the operation burden of a system, reflects the change of the actual state of the battery under the actual working condition, and improves the precision and accuracy of battery state estimation. The combination of the two time scale modes realizes the rolling optimization of the battery state estimation and enhances the adaptability of the invention.
4. The accuracy of the prediction model not only needs to build a reasonable algorithm model, but also needs to select direct or indirect health factors capable of reasonably representing the aging state of the battery, and the health factors need to accurately describe the change trend of the health state of the battery. The reasonable choice of health factors plays a vital role in the accuracy of data-driven state estimation. In the invention, not only the direct health factor which directly reflects the battery state change is collected by a direct measurement method, but also the indirect health factor related to the battery capacity is extracted by an indirect analysis method, and a plurality of battery state elements are synthesized to be used as the input of an aging model, so that the consideration of system variables is wider. The prediction precision and the accuracy of the battery capacity are improved, so that the accuracy of battery state estimation is improved, and the reliability of the system is enhanced.
5. Because the fast charge mode and the slow charge mode can cause the service life reduction rate of the battery to change, the resistance-capacitance parameters under different service life states of the battery also change, and further cause the interval change in which the resistance-capacitance parameters can be regarded as unchanged, the prediction period of the slow scale needs to be adapted to the change of the service life reduction rate of the battery. The life change rate is positive, which means that the battery life decline trend is slowed down and the prediction period is increased; the negative life change rate indicates an increase in the battery life decrease trend and a decrease in the prediction period. The prediction of the battery state of health and the residual service life is adapted to different charging modes, the prediction accuracy of the battery state and the application range of the invention are improved, and meanwhile, the influence of different charging modes on the battery can be intuitively seen by a user through displaying the service life change rate to the user, so that the good charging habit of the user is promoted.
Description of the drawings:
FIG. 1 is a flow chart of a method for joint estimation of battery state;
FIG. 2 is a circuit diagram of a second-order equivalent model of a battery;
FIG. 3 is a graph showing B0005 versus 71 threshold at the onset of prediction;
FIG. 4 is a graph showing B0005 with a threshold of 81 as the starting point of prediction;
FIG. 5 is a graph showing B0005 with a threshold of 91 at the onset of prediction;
FIG. 6 is a graph of superparameter design values for the LSTM-RNN model;
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It will be apparent that the described embodiments are some, but not all, embodiments of the invention.
Thus, the following detailed description of the embodiments of the invention is not intended to limit the scope of the invention, as claimed, but is merely representative of some embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that, under the condition of no conflict, the embodiments of the present invention and the features and technical solutions in the embodiments may be combined with each other.
It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures.
In the description of the present invention, it should be noted that, the terms "upper", "lower", and the like indicate an azimuth or a positional relationship based on the azimuth or the positional relationship shown in the drawings, or an azimuth or a positional relationship conventionally put in use of the inventive product, or an azimuth or a positional relationship conventionally understood by those skilled in the art, such terms are merely for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the apparatus or element to be referred must have a specific azimuth, be constructed and operated in a specific azimuth, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and the like, are used merely to distinguish between descriptions and should not be construed as indicating or implying relative importance.
Embodiment one: referring to figures 1-5 of the drawings,
Collecting direct health factors representing the health aging state of a battery and corresponding battery health states in battery capacity degradation data, and recording the direct health factors and the corresponding battery health states as a first training set; the battery state of health is the ratio of the current battery capacity to the rated battery capacity, and represents the capacity of the battery in the current state relative to the capacity of the new battery for storing electric quantity. The direct health factor includes battery voltage, current, temperature, and self-discharge rate. The direct health factors are in one-to-one correspondence with the battery cycle periods; and the corresponding battery state of health needs to be obtained.
Constructing a first model, taking the direct health factor as input, taking the battery health state as output, and training the first model; the first model is a battery state of health aging model;
Inputting the direct health factor into the trained first model to obtain the battery health state; calculating the battery health state and rated battery capacity to obtain a first predicted battery capacity; acquiring an OCV-SOC curve graph according to the first predicted battery capacity; substituting the OCV value and the SOC value which are in one-to-one correspondence in the graph into a battery equivalent circuit equation, updating the resistance-capacitance parameter in the UKF algorithm, and calculating the state of remaining battery charge.
In this embodiment, as shown in fig. 2, the battery adopts a second-order RC equivalent circuit model, which includes a voltage source, an internal resistor, and a resistor-capacitor network (RC) related to the remaining battery state, so as to describe the electrical relationship between the remaining battery state and the terminal voltage. Typically, in an equivalent circuit model, the resistance represents the self-discharge effect and the high value capacitor or voltage source represents the battery open circuit voltage. The mathematical relation formula of the equivalent circuit model in the figure is as follows:
；
Wherein U oc is the open circuit voltage of the battery, R 0 is ohmic internal resistance, R 1 is electrochemical polarized internal resistance, C 1 is electrochemical polarized capacitance, U c1 is the terminal voltage of the electrochemical polarized internal resistance of the battery, R 2 is concentration polarized internal resistance, C 2 is concentration polarized capacitance, U c2 is the terminal voltage of the concentration polarized internal resistance of the battery, U is terminal voltage, and I is current; the relationship of f (remaining state of charge (t))=u oc is the OCV-SOC curve.
As the preferable technical scheme of the application, the prediction of the residual electric quantity state is related to the resistance-capacitance parameter in the battery equivalent circuit according to the mathematical relation formula of the equivalent circuit model. The UKF algorithm can be used to predict the state of charge of the battery by determining the value of the resistance-capacitance parameter. And the resistance-capacitance parameter is obtained by substituting an OCV-SOC curve calibrated in advance into the mathematical relationship formula of the equivalent circuit model.
Under the condition that the preset condition of a laboratory is constant, collecting different actual battery capacities of the battery under different periods by taking the time that the resistance-capacitance parameter in the equivalent circuit of the battery can be regarded as constant as a period; meanwhile, intermittent constant-current discharge is carried out on the battery, the discharge voltage is recorded in real time, and a plurality of OCV-SOC curves are calibrated under different actual battery capacities.
The inventors have found in practice that the resistance-capacitance parameters in the second-order equivalent circuit model of the battery are not fixed, different states of charge remaining correspond to different resistance-capacitance parameters. In order to identify the resistance-capacitance parameters in different residual electric quantity states, the open circuit voltage in the different residual electric quantity states needs to be obtained first. In different states of health of the battery, the open circuit voltage corresponding to the same state of charge value will decrease with the decrease of the actual battery capacity, so that it is necessary to recalibrate the OCV-SOC curve under different capacities of the battery. Because the calibration of the OCV-SOC curve is obtained through actual experiments, the OCV-SOC curve in the embodiment needs to be calibrated in advance and stored in the system.
The inventors have found in practice that the state of health of a battery is affected by the capacity of the battery, and that the battery capacity data is affected by a large number of factors, such as higher temperatures may enhance the activity of electrochemical substances inside the battery, the chemical reaction is accelerated, and the concentration of the electrochemical substances decreases with the change of the chemical reaction process, resulting in a decrease in the capacity of the battery. The resistance-capacitance parameter of the battery slowly decreases as the battery service time increases. At the same time, the decrease in battery capacity further leads to a decrease in battery open circuit voltage, which in turn may affect the estimation of the state of charge remaining. The estimation of the state of charge of the remaining battery is thus adjusted to different battery states of health.
The inventors have found in practice that the battery state of health changes very slowly over time, so that the battery state of health is updated in a slow scale manner, when the current cycle period of the battery reaches an integer multiple of the prediction period threshold, the battery state of health is predicted using the first model. And multiplying the battery state of health by the battery rated capacity to obtain the first predicted battery capacity corresponding to the current cycle period. The prediction period threshold is determined according to the system operation capability.
The inventors have found in practice that it is difficult to monitor the battery capacity on-line in practical applications. For example, in the ampere-hour integration method, if the current measurement is inaccurate in practice, the calculation error of the residual electric quantity state is caused, and the error is larger and larger due to long-term accumulation. For example, an open circuit voltage method used in calibrating an OCV-SOC curve requires a long standing time and as many measurements as possible to obtain a relatively accurate OCV-SOC curve. The test environment of the open circuit voltage method is used for ensuring constant control of the ambient temperature, but the ambient temperature of the battery in practical application is changed at any time, so that the OCV-SOC curve cannot accurately estimate the state of charge value.
Because the difference between the calibration environment and the actual use environment is large, the OCV-SOC curve cannot accurately estimate the state of charge value. However, in the current cycle, the state of health of the battery is regarded as unchanged, so that the resistance-capacitance parameter of the battery is regarded as unchanged, and the resistance-capacitance parameter can be used as a constant value to calculate the state of charge value under the current cycle, and the resistance-capacitance parameter under a certain capacity is calculated by substituting the open circuit voltage value and the state of charge value into the mathematical relation formula of the equivalent circuit model.
The inventors have therefore chosen to predict the state of health of the battery by means of said first model. And then multiplying the battery state of health and the battery rated capacity to obtain a first predicted battery capacity, acquiring an OCV-SOC curve under the corresponding capacity, further calculating the resistance-capacitance parameter under the first predicted battery capacity, updating the resistance-capacitance parameter into a state equation of a UKF algorithm, solving the problem that different capacities influence the estimation of the state of the residual electric quantity, and improving the estimation accuracy of the state of the residual electric quantity. The resistance-capacitance parameter is considered unchanged at the current stage of the first predicted battery capacity until the first predicted battery capacity is recalculated.
Since the state of charge is continuously changing throughout the charge or discharge cycle, a fast-scale time approach is employed, i.e., the kf algorithm is used to continuously predict the state of charge during the charge or discharge. And when the battery cycle period reaches integer times of the prediction period threshold value, updating the resistance-capacitance parameter used by the state equation in the UKF algorithm. The method not only realizes the real-time update of the state of the residual electric quantity, but also improves the stability of the prediction accuracy.
In this embodiment, the first model uses an LSTM-RNN long-term memory recurrent neural network, and its super-parametric design is shown in FIG. 6.
As a preferred technical scheme of the present application, the method further comprises the step of estimating the remaining service life of the battery, and the step comprises:
Collecting charging or discharging data under different cycle periods in the battery capacity degradation data, and extracting indirect health factors related to the battery capacity; the cycle period corresponds to the indirect health factor one by one, and the indirect health factor is used for representing the service life aging state of the battery; the charge or discharge data includes battery voltage, current, temperature, capacity, and charge or discharge time.
Modeling according to the direct proportion relation between the indirect health factor and the battery capacity, and constructing a second model; taking the indirect health factor as input, taking the residual service life as output, and training the second model, wherein the second model is a battery residual service life aging model;
And collecting the indirect health factor input trained second model under the previous cycle period, and continuously predicting the corresponding second predicted battery capacity under all future battery cycle periods until reaching an end-of-life threshold, wherein the cycle period value corresponding to the end-of-life threshold is the residual service life of the battery.
In this embodiment, the remaining service life of the battery is used to represent the time when the battery is expected to be normally used in the current state, and the battery is generally estimated by the current performance characteristics and historical data statistics of the battery, and is generally represented by a cycle period; one cycle period indicates that the battery performs a complete charge and discharge process. Typically 70% of the rated capacity of the battery is used as the end-of-life threshold. In view of the problems that the battery is difficult to monitor the capacity on line and difficult to estimate accurately in actual operation, in the embodiment, an indirect analysis method is adopted, parameters which are strongly related to the battery capacity are selected or calculated from data monitored in real time to serve as indirect health factors, and key characteristics are extracted through a neural network in the second model to operate, so that the estimation of the residual service life of the battery is realized. As the state of health of the battery decreases, indicating a decrease in the actual battery capacity, the remaining life of the battery also decreases, and battery data at different cycle periods may change accordingly.
The inventor finds that the change of the residual service life of the battery is very slow in practice, so that the battery is also predicted in a slow scale mode, when the current cycle of the battery reaches an integral multiple of the threshold value of the predicted cycle, the neural network of the second model is used for predicting the second predicted battery capacity in different future cycle until the second predicted battery capacity reaches the end-of-life threshold value, and the value of the cycle corresponding to the end-of-life threshold value is the residual service life of the battery corresponding to the current cycle.
In this embodiment, the second model also uses an LSTM-RNN long-term memory recurrent neural network, and its super-parametric design is the same as that of the first model, as shown in fig. 6. Training data for the second model is collected from a B0005 battery; the acquisition cycle is from an initial state of the battery to an end-of-life threshold. And setting a prediction starting point threshold, taking an indirect health factor from the initial state to the prediction starting point threshold as a training set, and taking an indirect health factor from the prediction starting point threshold to the life end threshold as a test set. In order to enable the second model to train by using more indirect health factors, the comprehensive evaluation capacity of the second model is improved, and a prediction starting point threshold is set in the middle stage of battery capacity degradation. And 3 predictions are made for each round of testing, the prediction start threshold for 3 predictions being set to 71, 81 and 91, respectively. After multiple training, setting different prediction starting point thresholds to obtain optimal residual service life prediction results as shown in fig. 3-5. According to the graph, under the condition that different prediction starting point thresholds are set, the similarity of a prediction curve and a real curve is high, and the overall prediction effect of the second model on the residual service life is excellent.
The battery state joint estimation method provided by the application can be used for estimating the aging state of the battery by utilizing the battery state of health and the residual service life, and the aging state of the battery can influence the voltage, current and the change degree of the internal resistance of the battery, so that the factors are required to be recalibrated when the residual electric quantity state is estimated, and the accuracy and the reliability of the residual electric quantity state estimation are improved. Meanwhile, the battery state of health and the residual service life are accurately predicted, estimation errors caused by independently estimating the residual electric quantity state can be reduced, the cumulative effect of the errors in the propagation process is reduced, and the accuracy of predicting the residual electric quantity state is improved.
In actual use, the estimation of the state of charge is continued throughout the charging or discharging process. Meanwhile, when current passes through the internal resistance of the battery and the internal resistance of the line, heat loss generally occurs to cause the temperature of the battery to rise, so that the estimation of the state of residual electric charge is affected. The aging model in the invention adopts the LSTM-RNN long-short-term memory cyclic convolution network, and the hidden layer at the previous moment is used as a part of the input of the current moment in the calculation process, so that the current layer can fuse the information characteristics extracted at the previous moment, the relevance of the state of the residual electric quantity between continuous moments is maintained, and the accuracy of prediction is improved. Meanwhile, the method can adapt to more complex environments, and better processes uncertainty and interference by comprehensively considering a plurality of factors, so as to provide a more stable estimation result.
When the key features are updated, different time strategies are adopted according to different change degrees of the key features. The method for updating the residual electric quantity state in a fast scale achieves the real-time performance of prediction, the method for updating the battery state of health and the residual service life in a slow scale reduces the operation burden of a system, reflects the change of the actual state of the battery under the actual working condition, and improves the precision and accuracy of battery state estimation. The combination of the two time scale modes realizes the rolling optimization of the battery state estimation and enhances the adaptability of the invention.
As a preferable technical scheme of the application, the indirect health factor comprises equal voltage drop discharge time, average discharge voltage, constant current charge time and constant voltage charge time.
In this embodiment, the isobaric drop discharge time is obtained by calculating a difference between a time when the battery is at a high voltage and a time when the battery is at a low voltage in a single discharge cycle. The average discharging voltage is the average voltage of the battery in unit time in the discharging process. As the battery service time increases, battery performance and battery capacity decrease continuously as the battery cycle increases, and the output capacity of the battery decreases, so that the voltage decreases accordingly, thereby decreasing the isobaric discharge time and the average discharge voltage.
In this embodiment, the current charging mode of the battery adopts a mode of combining constant current charging and constant voltage charging, so as to balance charging time and charging safety. The constant current charging time refers to time t 1 required by a stage that the battery is charged by constant current I 1 in the charging process; the constant voltage charging time refers to the time t required by the battery in the charging process in the stage of charging by adopting constant voltage, and the current at the time is gradually reduced and is denoted as I t, t>t1. In the initial stage of battery charging, the internal resistance is smaller, and the battery can be rapidly charged by adopting larger charging current until the voltage rises to a preset constant voltage threshold value; at the end of charging, according to the OCV-SOC curve of the battery, the voltage of the battery is greatly changed, which is easy to cause battery overshoot and influence the service life and safety of the battery, so that a constant-voltage charging mode is often adopted until the charging current is reduced to a preset charging cut-off current threshold value. Therefore, the battery capacity can be expressed as the sum of the constant-current charge amount and the constant-voltage charge amount:
；
Wherein Q is the total electric quantity, and T is the time when the charge cut-off current threshold is reached. The formula shows that the constant-current charging time and the constant-voltage charging time are in direct proportion to the electric quantity of the battery.
The accuracy of the prediction model not only needs to build a reasonable algorithm model, but also needs to select direct or indirect health factors capable of reasonably representing the aging state of the battery, and the health factors need to accurately describe the change trend of the health state of the battery. The reasonable choice of health factors plays a vital role in the accuracy of data-driven state estimation. In the invention, not only the direct health factor which directly reflects the battery state change is collected by a direct measurement method, but also the indirect health factor related to the battery capacity is extracted by an indirect analysis method, and a plurality of battery state elements are synthesized to be used as the input of an aging model, so that the range of considering the system variables is wider. The prediction accuracy and the accuracy of the battery capacity are improved, the accuracy of battery state estimation is improved, and the reliability of the system is enhanced.
In practice, it is found that the remaining service life of the battery continuously decreases with the increase of the service time, so that the rate of decrease of the remaining service life of the battery is reflected by the rate of decrease of the remaining service life. Current battery charging modes include fast and slow charging. In order to meet the rapid power consumption requirements in social life, the voltage and power of rapid charging are higher and higher. However, in the fast charge mode, the charging voltage is high, the current is high, the heating value of the battery is also increased, and the decomposition of the positive electrode active material and the electrolyte is caused, so that the reduction rate of the residual service life of the battery is increased; the more rapid charging is used, the more the remaining service life reduction rate is increased, so that the resistance-capacitance parameter in the battery equivalent circuit model can be regarded as a constant interval reduction, and the prediction period threshold value is correspondingly reduced. When the slow charge mode is used, the charging current and the power are relatively smaller, the influence on the service life of the battery is smaller, the reduction rate of the residual service life of the battery is reduced, the resistance-capacitance parameter in the equivalent circuit model of the battery can be regarded as an unchanged interval to be increased, and therefore the prediction period threshold value is correspondingly increased.
The method comprises the steps of obtaining a first difference value by subtracting the residual service life predicted value of a current predicted period from the residual service life predicted value of a previous predicted period; and dividing the first difference value by the twice predicted elapsed time to obtain a current life reduction rate, wherein the life reduction rate is the change amount of the residual service life in unit time. Subtracting the life reduction rate of the previous prediction period from the life reduction rate of the current prediction period to obtain a second difference; dividing the second difference by the twice predicted elapsed time to obtain a life change rate, wherein the life change rate is the change amount of the life reduction rate in unit time; the life change rate is positive, and the reduction of the life of the battery is reduced; the rate of change of lifetime is negative, indicating an increase in the magnitude of the decrease in lifetime of the battery. And taking the ratio of the life change rate of the current prediction period to the life change rate of the previous prediction period as a scale factor for adjusting the threshold value of the next prediction period. And if the life change rate of the last prediction period is 0, the battery is in a factory state, and the threshold value of the next prediction period is kept unchanged.
Because the fast charge mode and the slow charge mode can cause the service life reduction rate of the battery to change, the resistance-capacitance parameters under different service life states of the battery also change, and further cause the interval change in which the resistance-capacitance parameters can be regarded as unchanged, the prediction period of the slow scale needs to be adapted to the change of the service life reduction rate of the battery. The life change rate is positive, which means that the battery life decline trend is slowed down and the prediction period is increased; the negative life change rate indicates an increase in the battery life decrease trend and a decrease in the prediction period. The prediction of the battery state of health and the residual service life is adapted to different charging modes, the prediction accuracy of the battery state and the application range of the invention are improved, and meanwhile, the influence of different charging modes on the battery can be intuitively seen by a user through displaying the service life change rate to the user, so that the good charging habit of the user is promoted.
In practice, it has been found that the inconsistent battery capacity can cause the battery pack to have inconsistent depths of charge or discharge for each cell. The battery with smaller capacity and poorer performance can reach the full-charge state in advance, so that the battery with large capacity and good performance can not reach the full-charge state. The BMS module is therefore generally required to initiate equalization control to reduce the overshoot or overdischarge of the smaller capacity battery, but repeated charging or discharging of the battery accelerates the aging of the battery. If the capacity of a certain monomer is abnormally reduced, the voltage deviation between the battery pack and the highest voltage in the battery pack is increased, so that the threshold value of balance control is easily reached, the charge or discharge times of the battery are further increased, and the aging speed of the battery is accelerated. Many factors that cause abnormal changes in the capacity of the battery, including abnormal performance or abnormal temperature of the battery, are present. The battery capacity caused by abnormal temperature can be recovered after the temperature is normal. The battery capacity caused by abnormal performance is difficult to recover, and the prediction of the whole battery cell health state, the residual service life and the residual electric quantity state can be influenced. Therefore, it is necessary to discriminate a battery having an abnormal capacity.
As a preferable technical scheme of the application, the time of the battery management system for starting the equalization control of the battery pack is obtained and is recorded as the first time; acquiring the time of the last equalization control start, and recording the time as second time; calculating a difference of the first time minus the second time; and when the difference value is smaller than a preset deviation threshold value, the starting time of the equalization control is too early, the voltage deviation of the battery pack is too large, and the battery is marked as a first abnormal battery.
And detecting the real-time temperature of the first abnormal battery, if the real-time temperature exceeds a temperature alarm threshold value given by a battery manufacturer, sending alarm information by the BMS module, and immediately disconnecting the battery from the battery pack, so that the current stops flowing in the first abnormal battery, and the influence of self ohmic internal resistance heating of the battery on the temperature is reduced. If the real-time temperature continuously drops to be consistent with the ambient temperature after the disconnection, outputting a first battery health state of the first abnormal battery by using the first model; acquiring all historical battery health states and calculating a first variance; calculating a second variance of the historical battery state of health and the first battery state of health; the second variance is subtracted from the first variance to obtain a variance variation, the variance variation is larger than a variance variation threshold, abnormal degradation of the battery health state occurs in the first abnormal battery, the battery performance is reduced, the disconnection state of the battery is maintained, and abnormal information of the battery is output through the BMS.
And the abnormal battery is disconnected from the battery pack, so that the resistance-capacitance parameter of the abnormal battery is difficult to influence the total resistance-capacitance parameter of the equivalent circuit model of the battery pack, and the prediction accuracy of the health state, the residual service life and the residual electric quantity state of the battery pack is maintained. Meanwhile, the situation that the BMS performs balance control too early due to abnormal batteries is reduced, the charge and discharge times among the single batteries in the battery pack are reduced, the aging speed of the single batteries is reduced, and the service time of the battery pack is further prolonged.
A battery state joint estimation system, comprising:
The data acquisition module is used for acquiring direct health factors of the battery health ageing state and corresponding battery health states under different cycle periods and recording the direct health factors and the corresponding battery health states as a first training set;
The model training module is used for constructing a first model, inputting the direct health factor into the first model and outputting the health state of the battery; the first model is a battery state of health aging model;
The data processing module is used for inputting the direct health factor into the trained first model to obtain the battery health state; calculating the battery health state and rated battery capacity to obtain a first predicted battery capacity; acquiring an OCV-SOC curve graph according to the first predicted battery capacity; substituting the OCV value and the SOC value which are in one-to-one correspondence in the graph into a battery equivalent circuit equation, updating the resistance-capacitance parameter in the UKF algorithm, and calculating the state of remaining battery charge.
The above embodiments are only for illustrating the present invention and not for limiting the technical solutions described in the present invention, and although the present invention has been described in detail in the present specification with reference to the above embodiments, the present invention is not limited to the above specific embodiments, and thus any modifications or equivalent substitutions are made to the present invention; all technical solutions and modifications thereof that do not depart from the spirit and scope of the invention are intended to be included in the scope of the appended claims.
Claims (10)
1. A battery state joint estimation method, comprising the steps of:
collecting direct health factors and corresponding battery health states of the battery under different cycle periods, and recording the direct health factors and the corresponding battery health states as a first training set;
Constructing a first model, taking the direct health factor as input, taking the battery health state as output, and training the first model; the first model is a battery state of health aging model;
Inputting the direct health factor into the trained first model to obtain the battery health state; calculating the battery health state and rated battery capacity to obtain a first predicted battery capacity; acquiring an OCV-SOC curve graph according to the first predicted battery capacity; substituting the OCV value and the SOC value which are in one-to-one correspondence in the graph into a battery equivalent circuit equation, updating the resistance-capacitance parameter in the UKF algorithm, and calculating the state of remaining battery charge.
2. The method of claim 1, further comprising estimating remaining useful life of the battery, the step comprising:
Collecting charging or discharging data under different cycle periods, and extracting indirect health factors related to battery capacity; the cycle period corresponds to the indirect health factor one by one, and the indirect health factor is used for representing the service life aging state of the battery;
Modeling according to the direct proportion relation between the indirect health factor and the battery capacity, and constructing a second model; taking the indirect health factor as input, taking second predicted battery capacity as output, and training the second model, wherein the second model is a battery residual service life aging model;
and collecting the indirect health factor input trained second model under the previous cycle period, and continuously outputting the corresponding second predicted battery capacity under all future battery cycle periods until reaching an end-of-life threshold, wherein the cycle period value corresponding to the end-of-life threshold is the residual service life of the battery.
3. The method of claim 2, wherein the indirect health factor comprises an equal voltage drop discharge time, an average discharge voltage, a constant current charge time, and a constant voltage charge time.
4. A battery state joint estimation method according to claim 2, wherein the battery state of health and the remaining life are estimated once when the time reaches an integer multiple of a prediction period threshold in a slow scale manner.
5. A battery state joint estimation method according to any one of claims 1 or 2, wherein the estimation of the state of charge is performed continuously during the charging or discharging of the battery in a fast scale manner.
6. The method for joint estimation of battery states according to any one of claims 1 or 2, wherein in the case where the laboratory preset condition is constant, different actual battery capacities of the battery under different periods are collected with the time in which the resistance-capacitance parameter in the battery equivalent circuit can be regarded as constant as a period; and meanwhile, intermittent constant-current discharge is carried out on the battery, and the discharge voltage and the residual electric quantity state are recorded in real time, so that the OCV-SOC curve graphs corresponding to a plurality of different actual battery capacities are obtained.
7. The method of claim 4, further comprising calculating a first difference between the remaining service life predictions of two adjacent times, and obtaining a life degradation rate according to a time-dependent relationship of the first difference; and calculating a second difference value of the service life reduction rate of two adjacent times, and obtaining the service life change rate according to the change relation of the second difference value along with time.
8. The battery state joint estimation method according to claim 7, wherein the prediction period threshold is adjusted according to a change ratio of the life change rate of two adjacent times.
9. The method of claim 7, further comprising obtaining a battery with a lowest voltage when the equalization control start time difference is less than a preset deviation threshold, and recording the battery as a first abnormal battery; collecting the historical battery health state of the first abnormal battery, and calculating a first variance; collecting the current battery health state of the first abnormal battery, and calculating a second variance; calculating variance variation amounts of the first variance and the second variance; and when the variance variation is larger than a variance variation threshold, the first abnormal battery is disconnected.
10. A battery state joint estimation system, comprising:
The data acquisition module is used for acquiring direct health factors of the battery health ageing state and corresponding battery health states under different cycle periods and recording the direct health factors and the corresponding battery health states as a first training set;
The model training module is used for constructing a first model, inputting the direct health factor into the first model and outputting the health state of the battery; the first model is a battery state of health aging model;
The data processing module is used for inputting the direct health factor into the trained first model to obtain the battery health state; calculating the battery health state and rated battery capacity to obtain a first predicted battery capacity; acquiring an OCV-SOC curve graph according to the first predicted battery capacity; substituting the OCV value and the SOC value which are in one-to-one correspondence in the graph into a battery equivalent circuit equation, updating the resistance-capacitance parameter in the UKF algorithm, and calculating the state of remaining battery charge.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410133141.4A CN117949851A (en) | 2024-01-31 | 2024-01-31 | Battery state joint estimation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410133141.4A CN117949851A (en) | 2024-01-31 | 2024-01-31 | Battery state joint estimation method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117949851A true CN117949851A (en) | 2024-04-30 |
Family
ID=90803480
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410133141.4A Pending CN117949851A (en) | 2024-01-31 | 2024-01-31 | Battery state joint estimation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117949851A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118169600A (en) * | 2024-05-14 | 2024-06-11 | 四川赛柏智晶科技有限公司 | Method and system for monitoring and maintaining service life state of storage battery pack of energy storage system |
-
2024
- 2024-01-31 CN CN202410133141.4A patent/CN117949851A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118169600A (en) * | 2024-05-14 | 2024-06-11 | 四川赛柏智晶科技有限公司 | Method and system for monitoring and maintaining service life state of storage battery pack of energy storage system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108254696B (en) | Battery health state evaluation method and system | |
| JP5936711B2 (en) | Storage device life prediction apparatus and storage device life prediction method | |
| EP3018753B1 (en) | Battery control method based on ageing-adaptive operation window | |
| JP5442583B2 (en) | State detection device for power supply and power supply device | |
| CN110320474A (en) | A kind of life-span prediction method of lithium ion battery Ageing Model | |
| WO2020198118A1 (en) | Methods, systems, and devices for estimating and predicting a remaining time to charge and a remaining time to discharge of a battery | |
| CN111463513B (en) | Method and device for estimating full charge capacity of lithium battery | |
| CN110376536B (en) | SOH detection method and device for battery system, computer equipment and storage medium | |
| CN117949851A (en) | Battery state joint estimation method and system | |
| KR101882287B1 (en) | Battery life estimation method and device of it | |
| CN113075554A (en) | Lithium ion battery pack inconsistency identification method based on operation data | |
| CN112305426B (en) | Lithium ion battery power state estimation system under multi-constraint condition | |
| Lu et al. | Modeling discharge characteristics for predicting battery remaining life | |
| CN114839538A (en) | Method for extracting degradation characteristics of lithium ion battery for estimating residual life | |
| US11656293B2 (en) | Method and apparatus for parameterizing an electrochemical battery model for a battery based on data from a plurality of batteries | |
| Wu et al. | State-of-charge and state-of-health estimating method for lithium-ion batteries | |
| US20220341997A1 (en) | Apparatus and Method for Diagnosing a Battery | |
| CN113759258A (en) | Power battery SOC estimation method and device and pure electric vehicle | |
| KR100911315B1 (en) | Apparatus and method for estimating battery's resistance characteristics based on open circuit voltage estimated by battery voltage variation pattern | |
| CN116893366A (en) | Method and device for operating a system to detect anomalies in an electrical energy store of a device | |
| CN115616434A (en) | Degradation model calibration-based lithium battery SOC and SOH real-time estimation method | |
| KR102205318B1 (en) | Method for estimating state of charge(soc) | |
| US20240213553A1 (en) | Battery Management System and Operation Method Thereof | |
| US20230213587A1 (en) | Method and System for Efficiently Monitoring Battery Cells of a Device Battery in an External Central Processing Unit Using a Digital Twin | |
| CN117031340A (en) | Method and device for calculating remaining life value of battery and storage 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 |