Battery voltage estimating method

Abstract

A battery voltage estimating method is provided. The battery voltage estimating method includes steps of: charging / discharging a battery multiple times such that an electric quantity of the battery respectively reach a plurality of specified states of charge; after the battery is charged or discharged each time, letting the battery rest until a voltage of the battery reaches a steady state voltage; sampling the steady state voltage and released voltages that are generated before the voltage of the battery reaches the steady state voltage; and training, according to the specified states of charge, the released voltages and the steady voltage and rest time of the battery, a battery voltage estimation model to estimate the released voltages and the steady voltage of a currently used battery based on current states of charge and the rest time of the currently used battery.

Description

Battery voltage prediction method This invention relates to batteries, and more particularly to a method for predicting battery voltage. With advancements in technology and a growing awareness of environmental protection, electric vehicles have become one of the mainstream green transportation options in various countries. Since the primary power source for electric vehicles is the battery, battery management has always been a crucial issue for electric vehicle designers. However, traditional battery management methods require a long resting period, such as 5 hours or even longer, after each charge or discharge to determine the voltage recovery process of the battery to its steady-state voltage during this resting period. This is extremely time-consuming, making it very inconvenient to further evaluate the battery status and application based on the voltage recovery process. To address the shortcomings of existing technologies, this invention provides a battery voltage prediction method, comprising the following steps: In a test mode, a battery is charged multiple times to charge it from an initial uncharged remaining capacity percentage to multiple specified remaining capacity percentages; in the test mode, after each charge, the battery is allowed to rest until its voltage rebounds to a steady-state voltage; in the test mode, multiple release voltages and the steady-state voltage upon reaching the steady-state voltage are sampled during the voltage rebound process after multiple charges; and in the test mode, based on the multiple specified remaining capacity percentages, multiple release voltages, and the resting time between a starting point of resting after charging and a rebound time point to each release voltage, a battery voltage prediction model is trained. In a practical prediction mode, based on any estimated resting time and a current remaining capacity percentage of any practical battery, the model predicts the release voltage and the steady-state voltage of the practical battery at any estimated time point after charging. The estimated resting time of the practical battery is the resting time between the starting point of resting after charging and any estimated rebound time point. To address the shortcomings of existing technologies, this invention provides another battery voltage prediction method, comprising the following steps: In a test mode, discharging the battery multiple times to discharge it from an initial undischarged remaining capacity percentage to multiple specified remaining capacity percentages; in the test mode, after each discharge, allowing the battery to rest until the battery voltage rebounds to a steady-state voltage; in the test mode, sampling multiple release voltages during the voltage rebound to the steady-state voltage after multiple discharges and the steady-state voltage upon reaching the steady state; and in the test mode, training a battery voltage prediction model based on the multiple specified remaining capacity percentages, multiple release voltages, and the resting time between a starting point of resting after battery discharge and a rebound time point to each release voltage, in a practical prediction mode, predicting the release voltage of the practical battery at any estimated time point after discharge and the steady-state voltage upon reaching the steady state based on an estimated resting time and a current remaining capacity percentage of any practical battery. The estimated resting time of the practical battery is the resting time between the starting point of resting after discharge and any estimated rebound time point. Please refer to Figures 1 to 4, where Figure 1 is a flowchart of the steps in the battery voltage prediction method of the present invention to establish a battery voltage prediction model based on battery charging data. The battery voltage prediction method of the present invention includes steps S101 to S107 of FIG1, which are executed by the processor 100 and charging / discharging device 200 and other hardware devices included in the battery voltage prediction system of the present invention shown in FIG2. Steps S101 to S105 are executed in a test mode, and S106 and S107 are executed in a practical prediction mode. In step S101 of Figure 1, the charging and discharging device 200 charges the battery multiple times to charge it from an initial uncharged remaining capacity percentage (e.g., 0%) to multiple specified remaining capacity percentages (State of Charge, SoC), wherein the multiple specified remaining capacity percentages are not equal to each other. In practice, the battery can be charged with different charging currents, and step S101 can be executed repeatedly. It is worth noting that the battery voltage estimation method of the present invention employs a step-by-step charging mechanism. First, the battery is discharged completely, i.e., discharged to the cutoff condition according to the cell discharge format. Next, in the first charging stage of the step-by-step charging mechanism, the battery is charged from 0% to a first specified remaining charge percentage, for example, 10%. Next, in the second charging stage of the step-by-step charging mechanism, the battery is charged from 0% to a second specified remaining charge percentage, for example, 20%. Next, other charging stages for specified remaining charge percentages are performed in the same manner. Finally, in the Nth charging stage of the step-by-step charging mechanism, the battery is charged from 0% to an Nth specified remaining charge percentage, for example, 90%. The above example illustrates charging the battery from 0% to a first specified remaining charge percentage through a ninth specified remaining charge percentage, for example, "10%", "20%", "30%", "40%", "50%", "60%", "70%", "80%", and "90%" respectively, but the present invention is not limited to this. In step S102, after each battery charge, the charging / discharging device 200 allows the battery to rest for a period of time (this period is called the resting time). During the resting time, charging of the battery is stopped, and the battery voltage is allowed to recover to a steady-state voltage. Once the battery voltage has recovered to a steady-state voltage, the battery can then be fully charged according to the full-charge conditions. In step S103, the processor 100 samples multiple release voltages during the process of the battery voltage rebounding to a steady-state voltage after multiple charging, and a resting time between a starting time point after the battery is charged and a rebound time point to each of the multiple release voltages. In step S104, the processor 100 samples the steady-state voltage of the battery after it has been charged for a period of time and then rests for a period of time. The steady-state voltage is a voltage that remains unchanged and is maintained at a constant value after a long period of time. In step S105, the processor 100 trains a battery voltage prediction model based on multiple specified remaining charge percentages of the battery after multiple charges, multiple release voltages, and a resting time between a starting time point after the battery enters a resting state and a rebound time point to each of the multiple release voltages. The model predicts a release voltage at a predicted time point and a steady-state voltage at a steady state for any practical battery based on any predicted resting time and a current remaining charge percentage. The predicted resting time of a practical battery as described herein refers to the expected resting time for a battery in actual use. The practical battery described in this article refers to a battery used in actual applications; this terminology is used to distinguish it from batteries tested in test mode. The estimated resting time of the practical battery described in this article is the resting time between the starting point of the practical battery entering resting after charging or discharging and any estimated rebound time point. The battery voltage prediction method of the present invention may further include the following steps for training a battery voltage prediction model: the processor 100 analyzes the correlation between multiple release voltages at multiple rebound time points after each battery charge and the steady-state voltage to generate multiple correlation parameters; and the processor 100 establishes a battery voltage prediction model based on the multiple correlation parameters from multiple charges. For example, the processor 100 may establish an equivalent model circuit, which includes multiple sets of test circuits. Each set of test circuits may include a resistor and a capacitor connected in parallel, such as, but not limited to, the first test circuit TES1 and the second test circuit TES2 shown in Figure 3. The first test circuit TES1 includes a first capacitor C1 and a first resistor R1, and the second test circuit TES2 includes a second capacitor C2 and a second resistor R2. Furthermore, the processor 100 can establish second-order training model equations: , in This represents the battery's discharge voltage. This represents the battery's resting time. Representing the steady-state voltage of the battery, V1, V2, , For multiple model training parameters, V1 represents the first voltage difference included in the multiple model training parameters (this is the voltage difference generated by the first capacitor C1 or the first resistor R1 of the first test circuit TES1 in Figure 3 when the battery is charged by current I), and V2 represents the second voltage difference (this is the voltage difference generated by the second capacitor C2 or the second resistor R2 of the second test circuit TES2 in Figure 3 when the battery is charged by current I, where current I can be the charging current or discharging current of the battery). This represents the first time constant (which is the product of the capacitance value of the first capacitor C1 and the resistance value of the first resistor R1 in the first test circuit TES1 in Figure 3). This represents the second time constant (which is the product of the capacitance value of the second capacitor C2 and the resistance value of the second resistor R2 in the second test circuit TES2 in Figure 3). The processor 100 can substitute multiple discharge voltages, a steady-state voltage, and multiple resting times generated after each charge or discharge of the battery into the aforementioned second-order training model equations to calculate multiple model training parameters. Specifically, the processor 100 can calculate multiple model training parameters V1, V2, ..., V3 generated after each charge or discharge based on the multiple sets of model training parameters V1, V2, ... , To select or calculate a set of multiple model training parameters V1, V2, ... that best represent or closely approximate the actual state of the battery. , Substitute these equations into the second-order training model equation to form a second-order battery prediction equation. For example, as shown in Figure 4, the second-order battery prediction equation describes the state of a voltage difference changing over time. This equation is included in the battery voltage prediction model described above and is used to predict the release voltage of any practical battery at a predicted time point and the steady-state voltage when it reaches a steady state, based on any predicted resting time and a current remaining charge percentage. The actual state of the battery described above refers to the state of the voltage difference between the release voltage and the steady-state voltage as time changes during the process of the battery's release voltage rebounding to the steady-state voltage after the battery is charged or discharged with current I to reach a specified remaining percentage of charge and then left to stand. The example above uses a second-order model, but in practice, more orders can be used. The order can depend on the battery's composition materials. For example, the processor 100 can establish multi-order training model equations: … , in This represents the battery's discharge voltage. This represents the battery's resting time. Represents the steady-state voltage of the battery, V1~VN. ~ For training parameters of multiple models, V1~VN represent multiple voltage differences (the voltage differences generated by multiple capacitors or resistors in multiple test circuits when the battery is charged by current I). ~ This represents N time constants (which are multiple products of the capacitance values ​​of multiple capacitors in multiple test circuits and the resistance values ​​of multiple resistors in multiple test circuits). The processor 100 can substitute multiple discharge voltages, a steady-state voltage, and multiple resting times after each charge or discharge of the battery into the multi-order training model equations to calculate multiple sets of model training parameters V1~VN. ~ The processor 100 can train parameters V1~VN based on multiple sets of models generated after multiple charging or discharging cycles. ~ To select or calculate a model training parameter V1~VN that best matches or closely approximates the actual state of the battery. ~ Substitute the equations into the multi-order training model to form a multi-order equation for battery prediction, which is included in the battery voltage prediction model mentioned above. Please refer to Figures 5 to 7, where Figure 5 is a flowchart of the steps in the battery voltage prediction method of this invention to establish a battery voltage prediction model based on battery discharge data. The battery voltage prediction method of the present invention includes steps S201 to S207 as shown in FIG5, which can be executed by the processor 100 and other hardware devices such as the charging and discharging device 200 as shown in FIG2. In step S201, in test mode, the charging / discharging device 200 discharges the battery multiple times to discharge the battery from an initial undischarged remaining capacity percentage (e.g., 100%) to multiple specified remaining capacity percentages, wherein the multiple specified remaining capacity percentages are not equal to each other. In practice, step S201 can be repeatedly executed by discharging the battery multiple times with different discharge currents. It is worth noting that the battery voltage estimation method of the present invention employs a step-by-step discharge mechanism. First, the battery is fully charged, i.e., charged to the cutoff condition according to the cell charging format. Next, in the first discharge stage of the step-by-step discharge mechanism, the battery is discharged from an initial undischarged remaining charge percentage, for example, 100%, to a first specified remaining charge percentage, for example, 90%. Next, in the second discharge stage of the step-by-step discharge mechanism, the battery is discharged from 100% to a second specified remaining charge percentage, for example, 80%. Then, other discharge stages for specified remaining charge percentages are performed in the same manner. Finally, in the Nth discharge stage of the step-by-step discharge mechanism, the battery is discharged from 100% to an Nth specified remaining charge percentage, for example, 10%. The above example illustrates discharging the battery from 100% to the first to ninth specified remaining charge percentages, for example, "90%", "80%", "70%", "60%", "50%", "40%", "30%", "20%" and "10%", but the present invention is not limited thereto. In step S202, in test mode, after each battery discharge, the charging / discharging device 200 allows the battery to rest for a period of time (this period is called the resting time). During the resting time, the battery discharge is stopped, as shown in Figure 6, waiting for the battery voltage to rebound to a steady state. When the battery voltage rebounds to the steady-state voltage, the battery can be discharged completely according to the discharge cutoff conditions. In step S203, in test mode, the processor 100 samples multiple release voltages during the voltage rebound to steady-state voltage process after multiple discharges of the battery, and samples a resting time between a resting start time point after the battery discharges and a rebound time point to each of the multiple release voltages. For example, it samples the battery data between a resting start time point and a steady-state time point on a voltage discharge rebound curve marked by the solid line shown in Figure 7. In step S204, in test mode, the processor 100 samples the steady-state voltage of the battery after it has been discharged for a period of time and then rebounds to a steady-state voltage within a rest period. In step S205, in test mode, processor 100 trains a battery voltage prediction model based on multiple specified remaining charge percentages of the battery after multiple discharges, multiple release voltages, and a resting time between the battery discharge and each release voltage rebound. In practical prediction mode, the model predicts a release voltage of the practical battery at a predicted time point and a steady-state voltage at a steady state point based on a predicted resting time of the practical battery and a current remaining charge percentage. In test mode, processor 100 can substitute the discharge voltage, steady-state voltage, and resting time of the battery during each discharge into the aforementioned second-order or multi-order training model equations to calculate a set of multiple model training parameters V1~V2. , Or V1~VN and ~ The processor 100 can train parameters V1~V2 based on multiple sets of models generated for each discharge. , Or V1~VN and ~ To select or calculate a set of multiple model training parameters V1~V2 that best match or closely approximate the actual state of the battery. , Or V1~VN and ~ Substitute the equations into the second-order or multi-order training model to form a battery prediction multi-order second-order or multi-order equation, which is included in the above battery voltage prediction model. In steps S106-S107 and S206-S207, before the practical battery enters a practical prediction mode and reaches a steady state, the processor 100 determines whether the current remaining charge percentage of the practical battery is equal to any one of a plurality of specified remaining charge percentages. If the processor 100 determines that the current remaining charge percentage of the practical battery is equal to any one of the plurality of specified remaining charge percentages, the processor 100 executes a battery voltage prediction model (including the above-mentioned battery prediction second-order or multi-order equations), and based on a set of multiple model training parameters corresponding to a specified remaining charge percentage, predicts the release voltage or steady-state voltage of the practical battery at a certain rebound time point when it has a current remaining charge percentage. If the processor 100 determines that the current remaining charge percentage of the utility battery is not equal to any of the multiple specified remaining charge percentages, in a utility prediction mode, the processor 100 executes a battery voltage prediction model (including the above-mentioned battery prediction second-order or multi-order equations). Based on the fact that the utility battery has one or more of the multiple specified remaining charge percentages that are close to the current remaining charge percentage, the processor 100 can estimate the release voltage or steady-state voltage of the utility battery at a certain rebound time point when it has the current remaining charge percentage, according to one or more sets of multiple model training parameters. The processor 100 can use interpolation to perform the above estimation. For example, when the current remaining charge percentage of the practical battery is 35%, which is not equal to any of a plurality of specified remaining charge percentages of 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, and 90%, the processor 100 can execute a battery voltage prediction model. Based on a plurality of model training parameters when the practical battery has a specified remaining charge percentage of 30% that is close to the current remaining charge percentage of 35%, and a plurality of model training parameters when it has a specified remaining charge percentage of 40% that is close to the current remaining charge percentage of 35%, the processor 100 estimates the plurality of model training parameters when the practical battery has a current remaining charge percentage of 35%, so as to predict the release voltage or steady-state voltage of the practical battery at any rebound time point when the current remaining charge percentage is 35%. In practice, in test mode, the battery can be tested multiple times with different charging / discharging currents. For example, the current values ​​used in multiple tests may include a specified current value of 10 amps and a specified current value of 20 amps. In practical prediction mode, the processor 100 can execute a battery voltage prediction model based on the different charging or discharging current values ​​before resting. The processor 100 can estimate multiple model training parameters for a specified current value of 15 amps before resting based on multiple model training parameters for a specified current value of 10 amps close to 15 amps before resting, and multiple model training parameters for a specified current value of 20 amps close to 15 amps before resting, to predict the release voltage or steady-state voltage of the practical battery at any rebound time point during resting. It is worth noting that after establishing a battery voltage prediction model, it is not necessary to wait 5 hours or longer after discharging or charging to obtain the steady-state voltage of the practical battery, as shown in Figure 6. Instead, based solely on the current remaining charge percentage of the practical battery and the estimated resting time, the battery voltage prediction model can be used to quickly predict the release voltage at any rebound time point during the resting time after charging or discharging, as well as the steady-state voltage when the practical battery reaches steady state, before the practical battery reaches steady state. Please refer to Figures 8 to 10, which are graphs of the battery voltage prediction method according to an embodiment of the present invention. As shown in Figure 8, the battery voltage rebound curve measured after the battery has been charged and left to stand for a period of time is almost identical to the battery voltage (including the discharge voltage before reaching steady state and the steady-state voltage when reaching steady state) predicted by the battery voltage prediction method of the present invention as a function of standing time. As shown in Figure 9, the battery voltage prediction method of this embodiment of the invention measures the voltage difference between the battery's release voltage and the measured steady-state voltage at multiple rebound time points during a resting period after the battery is charged. This curve almost matches the voltage difference between the battery's release voltage and the measured steady-state voltage at multiple rebound time points predicted by the battery voltage prediction method of this invention. As shown in Figure 10, the error between the battery voltage predicted by the battery voltage prediction method of this embodiment and the measured battery voltage is extremely small. Accordingly, the battery voltage prediction method of this embodiment can accurately predict the battery voltage (including the release voltage at multiple rebound time points and the steady-state voltage when the battery reaches steady state) during a rest period after charging before the battery reaches steady state. In summary, this invention provides a battery voltage prediction method. This method can train a battery voltage prediction model for batteries with various material compositions to quickly predict the battery's discharge voltage and steady-state voltage during a resting period after charging or discharging. Therefore, unlike traditional battery management methods that require a long resting period after each charge or discharge cycle to determine the battery's discharge voltage or even steady-state voltage, this method eliminates the need for such a method. S101~S107, S201~S207: Step 200: Charging / Discharging Device 100: Processor TES1: First Test Circuit C1: First Capacitor R1: First Resistor TES2: Second Test Circuit C2: Second Capacitor R2: Second Resistor I: Current V1: First Voltage Difference V2: Second Voltage Difference Figure 1 is a flowchart of the steps in the battery voltage prediction method according to an embodiment of the present invention to establish a battery voltage prediction model based on battery charging data. Figure 2 is a block diagram of the battery voltage prediction system according to an embodiment of the present invention. Figure 3 is a circuit diagram of the battery data simulation circuit established by the battery voltage prediction method of this invention. Figure 4 is a graph showing the difference between the battery discharge voltage measured over time and the measured steady-state voltage after the battery has been discharged and left to stand for a period of time, according to the battery voltage prediction method of this embodiment of the invention. Figure 5 is a flowchart of the steps in the battery voltage prediction method of this invention to establish a battery voltage prediction model based on battery discharge data. Figure 6 is a curve showing the battery rebounding to steady-state voltage after discharge. Figure 7 is a graph showing the battery voltage rebound over time after the battery has been discharged and left to stand for a period of time, as measured by the battery voltage prediction method of this embodiment of the invention. Figure 8 is a graph showing the battery voltage rebound over time and the predicted voltage after the battery has been charged and left to stand for a period of time, according to the battery voltage prediction method of this embodiment of the invention. Figure 9 is a graph showing the difference between the battery's discharge voltage and the estimated voltage measured after the battery has been charged and left to stand for a period of time, and the measured steady-state voltage, according to the battery voltage prediction method of this embodiment of the invention. Figure 10 is a graph showing the error between the measured battery voltage and the estimated voltage after the battery has been charged and left to stand for a period of time, according to the battery voltage estimation method of this embodiment of the invention. S101~S107: Steps

Claims

1. A battery voltage prediction method includes the following steps: In a test mode, a battery is charged multiple times to charge the battery from an initial uncharged remaining capacity percentage to multiple specified remaining capacity percentages; In the test mode, after each charge, the battery is allowed to rest until the battery voltage rebounds to a steady-state voltage; In the test mode, multiple release voltages during the voltage rebound to the steady-state voltage after multiple charges and the steady-state voltage upon reaching the steady state are sampled; and In the test mode, based on the multiple specified remaining capacity percentages, the multiple release voltages, and a resting time between a starting point of resting after charging and a rebound time point to each release voltage, a battery voltage prediction model is trained in a practical prediction mode to predict the release voltage of the practical battery at any estimated time point after charging and the steady-state voltage upon reaching the steady state based on any estimated resting time and a current remaining capacity percentage of any practical battery. The estimated resting time of the practical battery is the resting time between the starting time point when the practical battery enters resting after charging and any estimated rebound time point. The battery voltage prediction method as described in claim 1 further includes the following steps for training the battery voltage prediction model: In the test mode, analyzing the correlation between the release voltage and the steady-state voltage at multiple rebound time points after each charge of the battery to generate multiple correlation parameters; and establishing the battery voltage prediction model based on the multiple correlation parameters of the battery after multiple charges. The battery voltage prediction method as described in claim 1 further includes the following steps for training the battery voltage prediction model: In the test mode, establishing first- and second-order training model equations: , in This represents the release voltage. This represents the settling time. Representing the steady-state voltages, V1, V2, 、 For training parameters of multiple models, V1 represents a first voltage difference, V2 represents a second voltage difference, Represents a first time constant. Represents a second time constant; In the test mode, the release voltage, steady-state voltage, and resting time of each battery charge are substituted into the second-order training model equation to calculate a set of multiple model training parameters; and In the test mode, based on the multiple sets of multiple model training parameters generated for each battery charge, a set of multiple model training parameters is selected or calculated and substituted into the second-order training model equation to form a battery prediction second-order equation, which is included in the battery voltage prediction model. The battery voltage prediction method as described in claim 1 further includes the following steps for training the battery voltage prediction model: In the test mode, establishing a multi-order training model equation: … , in This represents the release voltage. This represents the settling time. Representing the steady-state voltage, V1~VN, ~ For training parameters of multiple models, V1~VN represent N voltage differences. ~ Representing N time constants; In the test mode, the release voltage, steady-state voltage, and resting time of each battery charge are substituted into the multi-order training model equation to calculate a set of multiple model training parameters; and In the test mode, based on the multiple sets of multiple model training parameters generated for each battery charge, a set of multiple model training parameters is selected or calculated and substituted into the multi-order training model equation to form a battery prediction multi-order equation, which is included in the battery voltage prediction model. A battery voltage prediction method includes the following steps: In a test mode, discharging a battery multiple times to discharge the battery from an initial undischarged remaining capacity percentage to multiple specified remaining capacity percentages; In the test mode, after each discharge, allowing the battery to rest until the battery voltage rebounds to a steady-state voltage; In the test mode, sampling multiple release voltages during the voltage rebound to the steady-state voltage after multiple discharges and the steady-state voltage upon reaching the steady state; and In the test mode, training a battery voltage prediction model in a practical prediction mode based on multiple specified remaining capacity percentages, multiple release voltages, and a resting time between a starting time point after the battery discharges and a rebound time point to each release voltage, and based on an estimated resting time and a current remaining capacity percentage of any practical battery, predicting the release voltage and the steady-state voltage of the practical battery at any estimated time point after discharge in a practical prediction mode. The estimated resting time of the practical battery is the resting time between the starting time point when the practical battery enters resting after discharge and any estimated rebound time point. The battery voltage prediction method as described in claim 5 further includes the following steps for training the battery voltage prediction model: In the test mode, analyzing the correlation between the release voltage and the steady-state voltage at multiple rebound time points after each discharge of the battery, so as to generate multiple correlation parameters respectively; and establishing the battery voltage prediction model based on the multiple correlation parameters of the battery after multiple discharges. The battery voltage prediction method as described in claim 5 further includes the following steps for training the battery voltage prediction model: In the test mode, establishing first- and second-order training model equations: , in This represents the release voltage. This represents the settling time. Representing the steady-state voltages, V1, V2, 、 For training parameters of multiple models, V1 represents a first voltage difference, V2 represents a second voltage difference, Represents a first time constant. Each represents a second time constant; in the test mode, the discharge voltage, the steady-state voltage, and the resting time of each battery discharge are substituted into the second-order training model equation to calculate a set of multiple model training parameters; and in the test mode, a set of multiple model training parameters are selected or calculated from the multiple sets of multiple model training parameters generated for each battery discharge and substituted into the second-order training model equation to form a battery prediction second-order equation, which is included in the battery voltage prediction model. The battery voltage prediction method as described in claim 5 further includes the following steps for training the battery voltage prediction model: In the test mode, a multi-order training model equation is established: … , in This represents the release voltage. This represents the settling time. Representing the steady-state voltage, V1~VN, ~ For training parameters of multiple models, V1~VN represent N voltage differences. ~ Represents N time constants; In the test mode, the discharge voltage, steady-state voltage, and resting time of each battery discharge are substituted into the multi-order training model equation to calculate a set of multiple model training parameters. ； In the test mode, based on the multiple sets of model training parameters generated for each discharge of the battery, a set of multiple model training parameters is selected or calculated and substituted into the multi-order training model equation to form a battery prediction multi-order equation, which is included in the battery voltage prediction model.

Images