A battery power correction method and system based on a drone battery management system

Abstract

The application discloses a battery power correction method and system based on a UAV battery management system, and relates to the technical field of battery quality management systems. The method comprises the following steps: synchronously collecting multiple sensor data in real time, and preprocessing the collected multiple sensor data; calculating a load fluctuation signal CKI according to the preprocessed sensor data; predicting the battery power, calculating a new information according to a current battery power prediction value and a battery power observation value, calculating a residual decay index RID according to the new information mean, and dynamically adjusting the strategy for correcting the difference between the current battery power prediction value and the battery power observation value according to the residual decay index RID; calculating a current risk scale, and dynamically adjusting the prediction compensation amount of the current battery power according to the current risk scale; and obtaining a corrected current prediction power based on the current battery power prediction value and the prediction compensation amount of the current battery power. The application can recursively correct the battery power prediction value, so that the battery power prediction value remains accurate.

Description

Technical Field

[0001] This invention relates to the field of battery performance testing, and in particular to a battery power correction method and system based on a drone battery management system. Background Technology

[0002] When drones perform highly maneuverable and tightly coupled missions, the reliability of real-time power prediction is subject to extremely stringent requirements. Within seconds of takeoff, the current can surge from tens of amperes to hundreds of amperes, then drop sharply due to deceleration or attitude adjustments. If equivalent circuit models or black-box neural networks calibrated only under steady-state conditions are used to directly provide the state of charge, the prediction curve will often exhibit sharp overestimations and underestimations, which the flight control system is completely unaware of. If incorrect remaining power is used for path planning or return-to-home decisions, it can lead to anything from premature landing and mission delays to complete power depletion and loss of control.

[0003] The current industry practice is to use Kalman filtering for power estimation and prediction. However, these algorithms all assume that the input signal changes relatively slowly, so they will have problems with inaccurate prediction under drastic load fluctuations.

[0004] Meanwhile, low-cost shunt resistors can also cause ringing and overshoot in high-current scenarios, further amplifying input noise. The core idea of ​​this technical solution is to attach a dynamic correction logic based on load fluctuation sensing to bring transient errors back to a safe range. Summary of the Invention

[0005] To improve the accuracy of battery power prediction for drones under severe load fluctuations, this invention provides a battery power correction method and system based on a drone battery management system. It designs a dynamic correction logic based on load fluctuation perception to improve the accuracy of battery power prediction values ​​when the drone's flight state (such as acceleration and deceleration) changes.

[0006] In a first aspect, the present invention provides a battery power correction method based on a drone battery management system, employing the following technical solution:

[0007] A battery power correction method based on a drone battery management system includes:

[0008] Step 1: Simultaneously collect data from multiple sensors in real time and preprocess the collected data.

[0009] Step 2: Calculate the load fluctuation signal CKI based on the preprocessed sensor data;

[0010] Step 3: Use Kalman filtering to predict battery power, calculate innovation based on the current predicted battery power and observed battery power, obtain the mean of innovation, calculate the residual decay exponent RID based on the mean of innovation, and dynamically adjust the strategy to correct the difference between the current predicted battery power and observed battery power based on the residual decay exponent RID.

[0011] Step 4: Calculate the current risk scale and dynamically adjust the predicted compensation amount of the current battery power based on the current risk scale; the risk scale, combined with the load fluctuation signal CKI and the residual decay index RID, represents the prediction error of the current battery power.

[0012] Step 5: Obtain the corrected current predicted battery level based on the current predicted battery level and the predicted compensation amount for the current battery level.

[0013] Furthermore, the preprocessing of the collected data from multiple sensors involves removing data noise using median filtering to achieve data smoothing.

[0014] Furthermore, the load fluctuation signal CKI is calculated to determine the degree of load fluctuation by calculating the kurtosis of the current signal; the kurtosis of the current signal is represented by the ratio of the fourth moment to the second moment of the current data.

[0015] Furthermore, the acquisition of the average innovation value is achieved by pre-setting a time window corresponding to the current time, acquiring the innovation value corresponding to each time of the collected data within the time window, and calculating the average innovation value.

[0016] Furthermore, the strategy of dynamically adjusting and correcting the difference between the current predicted battery power value and the observed battery power value based on the current residual decay index RID includes obtaining the residual decay index RID for each time corresponding to the data collected within the time window corresponding to the current time; obtaining the peak value F of the residual decay index RID; calculating the ratio of the residual decay index RID to the peak value F; and normalizing the corresponding residual decay index RID through the peak value F. If the ratio of the residual decay index RID to the peak value F is larger, it indicates the beginning of the decay stage, and the prediction error of the corresponding battery power will be larger in the next time moment; conversely, it indicates the end of the decay stage, and the prediction error of the corresponding battery power will be smaller in the next time moment.

[0017] Furthermore, the risk scale obtains the processed load fluctuation signal CKI by setting the load fluctuation signal CKI to zero using preset steady-state parameters, thus correcting only for high load changes; the risk scale combines the load fluctuation signal CKI and the residual decay index RID and is represented by the ratio of the processed load fluctuation signal CKI and the residual decay index RID to the peak value F.

[0018] Furthermore, the predicted compensation amount for dynamically adjusting the current battery level based on the current risk scale... Adjust the risk scale weights and combine them with the current current value to obtain the predicted compensation amount for the current battery capacity.

[0019] Secondly, the present invention provides a battery power correction system based on a drone battery management system, which adopts the following technical solution:

[0020] A battery power correction system based on a drone battery management system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the battery power correction method based on the drone battery management system described above is implemented.

[0021] The present invention has the following technical effects:

[0022] This invention eliminates the mismatch between current / voltage and motion state caused by time differences by synchronously and in real-time acquiring data from all sensors, providing high-precision raw data for battery power prediction. This invention suppresses sensor data noise through median filtering, ensuring smooth current and voltage signals, significantly reducing the interference of outliers on battery power prediction, and enhancing system stability. This invention quantifies the degree of load fluctuation in real-time based on the kurtosis of the current signal, dynamically adjusting the battery power prediction value under high-dynamic conditions of the UAV, reducing battery power jump errors caused by sudden loads, and improving the accuracy of battery power prediction under severe load fluctuations. This invention distinguishes between fast and slow decay errors through the residual decay index (RID), employing targeted "fast correction and adjustment" or "slow correction and stable adjustment" strategies to balance correction speed and stability, reducing overshoot or hysteresis. Attached Figure Description

[0023] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description in conjunction with the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein:

[0024] Figure 1 This is a flowchart of a battery power correction method based on a drone battery management system provided in an embodiment of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] On one hand, embodiments of the present invention disclose a battery power correction method based on a drone battery management system, referring to... Figure 1 This includes the following steps:

[0027] S101: Simultaneously and in real-time acquires data from multiple sensors and preprocesses the acquired data.

[0028] Specifically, the data collected by multiple sensors on the drone should be synchronized to ensure that the time reference of the data collected by different sensors is consistent. By synchronously collecting data from all sensors in real time, it is ensured that the changes in the drone's current and voltage can accurately reflect the drone's motion state, avoiding data errors caused by time differences, and providing high-precision raw data for predicting battery power.

[0029] It should be noted that the drone's multiple sensors include at least a battery current sensor, a battery voltage sensor, a battery temperature sensor, and a drone IMU sensor. The battery current sensor measures the current of the drone's battery, the battery voltage sensor measures the voltage of the drone's battery, the battery temperature sensor measures the temperature of the drone's battery, and the drone's IMU sensor measures the drone's acceleration and angular velocity to help understand the drone's motion status.

[0030] Specifically, the preprocessing of the collected data from multiple sensors involves removing data noise using median filtering to achieve data smoothing.

[0031] Specifically, median filtering refers to taking the values ​​of the two data points before and after the current data point for each data point, and then taking the median value of these three consecutive data points as the new data point to achieve data smoothing. The following uses current data collected by a current sensor as an example to illustrate median filtering.

[0032] The median filter expression is as follows:

[0033]

[0034] These three data points are the raw current signal data that need to undergo median filtering. This represents the current value at the previous time point of the current data point; This represents the current value at the current moment; This represents the current value at the next time step after the current data point.

[0035] Median filtering can effectively remove noise and outliers from the data, ensuring a smoother signal and thus making battery power prediction more reliable.

[0036] S102: Calculate the load fluctuation signal CKI based on the preprocessed sensor data.

[0037] In fact, the battery power of a drone is affected by changes in flight status (such as acceleration and deceleration). For example, when a drone accelerates, the load fluctuates drastically, and it is necessary to detect the degree of load fluctuation in order to adjust the predicted battery power value in a timely manner.

[0038] The response time of current during load abrupt changes is shorter than that of voltage and temperature. Therefore, the degree of load fluctuation can be detected by detecting the degree of current fluctuation. Specifically, the load fluctuation signal CKI can be calculated by determining the degree of load fluctuation by calculating the kurtosis of the current signal; the kurtosis of the current signal is represented by the ratio of the fourth moment to the second moment of the current data.

[0039] Specifically, if the current signal fluctuates very drastically, the kurtosis value of the corresponding data curve of the current time series data will be very large. Therefore, the kurtosis of the current signal can be used to determine whether there is load fluctuation and the degree of load fluctuation.

[0040] The kurtosis of the current signal satisfies the following polynomial:

[0041]

[0042] The fourth moment (i.e., the average of the fourth power of the data) is used to measure the degree of kurtosis in the data, especially whether abnormally large fluctuations (i.e., load fluctuations) have occurred.

[0043] The calculation of the fourth moment involves all current data points, and the specific expression is as follows:

[0044]

[0045] in, It is the average value of the current data. It is the number of data points. It is the current value at the current moment.

[0046] The quadratic moment (the average of the squares of the data) is used to measure the range of fluctuation in the data.

[0047] The expression for calculating the second moment is:

[0048]

[0049] The kurtosis of a current signal is measured by the ratio of its fourth moment to its second moment, which indicates the signal's "sharpness." When the load fluctuates significantly, the kurtosis value of the current signal will be high, indicating severe fluctuations. By calculating the kurtosis of the current signal, the degree of load fluctuation in the drone can be represented, allowing for adjustments to the battery level at these times and ensuring more accurate battery level predictions.

[0050] S103: Use Kalman filtering to predict battery power, calculate innovation based on the current predicted battery power value and the observed battery power value, obtain the mean of innovation, calculate the residual decay index RID based on the mean of innovation, and dynamically adjust and correct the difference between the current predicted battery power value and the observed battery power value based on the residual decay index RID.

[0051] In fact, battery power predictions sometimes contain errors. It's necessary to determine whether these errors have been reduced to avoid excessive correction of already decayed errors. Therefore, we need to obtain the latest observed battery power value at time t and the predicted battery power value at time t, and calculate the error between the two. The larger the error, the greater the potential for a significant error, which could be due to algorithmic errors or decay errors caused by intense activity. However, the algorithm should be relatively stable. Furthermore, if a large kurtosis is also present, it indicates that the current error is very likely due to a short period of high load.

[0052] Specifically, the new information is calculated as the difference between the current predicted battery level and the observed battery level, expressed as follows:

[0053]

[0054] Among them, The Kalman filter predicts the battery charge at time t. This represents the observed battery charge level at time t.

[0055] It should be noted that the battery charge prediction value obtained by the Kalman filter at time t refers to the battery charge prediction value obtained using the conventional Kalman filter. Specifically, the conventional Kalman filter prediction of battery charge is based on the battery state equation (physical model) and the state estimate from the previous time step, and is obtained by recursively applying the Kalman filter to obtain the current battery charge prediction value. The process of predicting battery charge using the conventional Kalman filter is existing well-known technology and will not be elaborated here.

[0056] Meanwhile, the battery power observation value at time t refers to the battery power value displayed by the UAV's built-in battery management system at time t. The method for obtaining the battery power value displayed by the UAV's built-in battery management system can be any of the existing methods (except Kalman filtering). This invention does not impose any restrictions, as long as the battery power value can be displayed in real time. For example, the ampere-hour integration method can be used.

[0057] Specifically, the residual decay exponent RID is calculated based on the mean of the new information, as shown in the following expression:

[0058]

[0059] in, The residual decay exponent, The smoothing factor (which controls the impact of the current time error on the final result, and takes a value between 0 and 1) can be taken as an empirical value; Let be the average value of the information at time t, and let represent the average error to prevent errors caused by short-term disturbances.

[0060] As one possible implementation, by setting a time window of 2 milliseconds from time t, the average value of the innovation value corresponding to each time point of each collected data point within the time window is calculated to obtain the innovation mean. 2 milliseconds is a hyperparameter. After obtaining the residual decay exponent at time t, to indicate whether the current stage is the beginning of decay or the decay recovery stage, since the changes in the battery charge measurement results at the next time point differ in different stages, the residual decay exponent RID value corresponding to each time point within 2 milliseconds from time t is obtained. The peak value F(t) of the residual decay exponent RID value is obtained, and the ratio of the residual decay exponent RID to the peak value F is calculated. The peak value F is used to normalize the corresponding residual decay exponent RID; if The larger the value, the more it indicates the beginning of the decay phase, and the corresponding error will be larger in the next moment. The smaller the value, the closer it is to the end of the decay phase, and the corresponding error will decrease in the next moment.

[0061] Specifically, by calculating the residual decay index RID, the system can dynamically adjust the error correction strategy, quickly correcting errors with rapid decay and extending the correction time for errors with slow decay, thereby improving the accuracy of battery power prediction.

[0062] S104: Calculate the current risk scale and dynamically adjust the predicted compensation amount of the current battery power based on the current risk scale.

[0063] Specifically, the risk scale combines the load fluctuation signal CKI and the residual decay index RID to represent the prediction error of the current battery charge.

[0064] Specifically, the risk scale obtains the processed load fluctuation signal CKI by setting the load fluctuation signal CKI to zero using preset steady-state parameters, thus correcting only for high load variations. The risk scale combines the load fluctuation signal CKI and the residual decay index RID, expressed as the ratio of the processed load fluctuation signal CKI and the residual decay index RID to the peak value F. Therefore, in steady state, the load fluctuation signal CKI is set to zero using steady-state parameters, eliminating the need for step division of the load.

[0065] Specifically, the risk scale expression of this invention is as follows:

[0066]

[0067] in, Load fluctuation signal Z is the residual decay exponent, and Z is a steady-state parameter, taken as an empirical value. First, subtract Z to ensure that the contribution is zero in steady state; It is the ratio of the fourth moment to the second moment of the current data, therefore It is a second-order current characteristic, from which a comprehensive risk scale is obtained.

[0068] Specifically, the predicted compensation amount that dynamically adjusts the current battery level based on the current risk scale. Adjust the risk scale weights and combine them with the current current value to obtain the predicted compensation amount for the current battery capacity.

[0069] Specifically, based on the magnitude of the aforementioned risk scale, the system determines whether the predicted battery capacity needs to be adjusted. If the error is large, the compensation amount increases; if the error is small, the compensation amount decreases.

[0070] The expression for the predicted compensation amount for the current battery level is as follows:

[0071]

[0072] in, This is the predicted compensation amount for the current battery level. The value of the current at the current moment. The () function and θ and τ are only used to adjust the risk scale. The weights of these parameters serve to normalize the data; θ≈0.5 and τ≈0.25 are hyperparameters.

[0073] In fact, by calculating the risk scale, the system can dynamically adjust the compensation amount to ensure that the predicted battery power value is always within a reasonable range, thereby improving accuracy.

[0074] S105: Obtain the corrected current predicted battery level based on the current predicted battery level and the predicted compensation amount for the current battery level.

[0075] Specifically, the revised formula for calculating the current predicted power consumption is as follows:

[0076]

[0077] in, To correct the predicted battery charge at time t, To correct the Kalman filter's prediction of the battery charge at time t, This represents the predicted compensation amount for the battery charge at time t.

[0078] Specifically, this invention quantifies the load fluctuation level in real time based on the kurtosis of the current signal, and dynamically adjusts the predicted battery level value as the UAV's flight state (such as acceleration and deceleration) changes. This reduces the battery level jump error caused by sudden load changes and improves the accuracy of battery level prediction under severe load fluctuations. The battery level correction method described in this invention combines a Kalman filter to correct the predicted battery level value when load fluctuations occur, reducing the error in the battery level measurement results caused by load fluctuations under high loads, ensuring that the predicted battery level value remains accurate.

[0079] On the other hand, embodiments of the present invention also disclose a battery power correction system based on a drone battery management system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the battery power correction method based on a drone battery management system according to the present invention is implemented.

[0080] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0081] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as resistive random access memory (DRAM), dynamic random access memory (DRAM), static random access memory (SRAM), etc., or any other medium that can be used to store desired information and can be accessed by an application program, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device.

[0082] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A battery power correction method based on a UAV battery management system, characterized in that, include: Step 1: Synchronously collect data from multiple sensors in real time and preprocess the collected data. Step 2: Calculate the load fluctuation signal CKI based on the preprocessed sensor data; Step 3: Use Kalman filtering to predict battery power, calculate innovation based on the current predicted battery power and observed battery power, obtain the mean of innovation, calculate the residual decay exponent RID based on the mean of innovation, and dynamically adjust the strategy to correct the difference between the current predicted battery power and observed battery power based on the residual decay exponent RID. Step 4: Calculate the current risk scale and dynamically adjust the predicted compensation amount of the current battery capacity based on the current risk scale; the risk scale combines the load fluctuation signal CKI and the residual decay index RID to represent the prediction error of the current battery capacity; the risk scale expression is: ;in, Load fluctuation signal Z is the residual decay exponent, and Z is the steady-state parameter. Indicates the risk scale. It is a second-order current characteristic. The peak value of the residual decay exponent RID; Step 5: Obtain the corrected current predicted battery level based on the current predicted battery level and the predicted compensation amount; the expression for the predicted compensation amount for the current battery level is as follows: ; wherein, is a predicted compensation of the current battery power, is a current value of the current at the current time, and are hyperparameters.

2. The battery power correction method based on the unmanned aerial vehicle battery management system according to claim 1, wherein The preprocessing of the collected data from multiple sensors involves removing data noise using median filtering to achieve data smoothing.

3. The battery power correction method based on the unmanned aerial vehicle battery management system according to claim 1, wherein The calculated load fluctuation signal CKI determines the degree of load fluctuation by calculating the kurtosis of the current signal; the kurtosis of the current signal is represented by the ratio of the fourth moment to the second moment of the current data.

4. The battery power correction method based on the unmanned aerial vehicle battery management system according to claim 1, wherein The process of obtaining the average value of the new information involves using a preset time window corresponding to the current time to obtain the new information value corresponding to each time of the collected data within the time window and calculating the average value of the new information.

5. The battery power correction method based on the unmanned aerial vehicle battery management system according to claim 4, characterized in that, The strategy of dynamically adjusting and correcting the difference between the current predicted battery power and the observed battery power based on the current residual decay index (RID) includes obtaining the residual decay index (RID) for each data collection moment within the current time window; obtaining the peak value (F) of the residual decay index (RID); calculating the ratio of the residual decay index (RID) to the peak value (F); and normalizing the corresponding residual decay index (RID) through the peak value (F). If the ratio of the residual decay index (RID) to the peak value (F) is larger, it indicates the beginning of the decay stage, and the prediction error of the corresponding battery power will be larger in the next moment; conversely, it indicates the end of the decay stage, and the prediction error of the corresponding battery power will be smaller in the next moment.

6. A battery power correction system based on a drone battery management system, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the battery power correction method based on a UAV battery management system according to any one of claims 1-5.

Images