Intelligent power supply method and system for fork mobile robot based on multi-source data

Abstract

The application discloses a fork type mobile robot intelligent power supply method and system based on multi-source data, relates to the technical field of robot power management, and realizes high timeliness, multi-dimension and low error dynamic analysis of the power supply operation state by constructing a multi-module cooperative structure.Under the premise of not depending on newly added hardware, the system jointly constructs a power supply data group by introducing an AFE chip and an operation characteristic sensor, combines time synchronization, dynamic interpolation and residual fusion filtering, significantly improves the synchronization accuracy and anti-interference ability of the original power supply signal, further introduces dynamic parameters such as power temperature gradient, load fluctuation rate and power change rate to construct a characteristic set, realizes high sensitivity identification of weak changes and state fluctuations in the power supply operation process, and effectively establishes a mapping channel between the sensing state and the operation parameter by using a coupling type comprehensive state function Ψsys, thereby improving the state self-sensing and parameter correction ability of the system under multiple working conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot power management, in particular to a forklift mobile robot intelligent power supply method and system based on multi-source data. BACKGROUND

[0002] As a common intelligent handling device in warehouse, logistics, manufacturing and other scenarios, the running efficiency and stability of the forklift mobile robot highly depend on the continuity and reliability of the power supply system. With the increasing complexity of the system, it is difficult to meet the real-time state judgment requirements in multi-working condition and multi-disturbance scenarios by simply relying on voltage, current and other conventional physical quantities for power supply state judgment, and there is an urgent need for an intelligent power supply sensing mechanism that integrates multi-modal sensing data.

[0003] In the existing forklift mobile robot power supply system, the battery management system mainly provides power supply state feedback, and the collected parameters include voltage, current, temperature, etc., but there is a lack of linkage feature extraction mechanism between running conditions (such as acceleration, power load fluctuation), and the time consistency and noise disturbance interference problems between multi-source data are not considered, resulting in unstable system evaluation results and low sensitivity.

[0004] The above problems mainly arise from three aspects: first, the state parameters of the power supply system have non-synchronous sampling and measurement noise, resulting in inconsistent data timing, which causes subsequent judgment distortion; second, there is currently a lack of unified feature mapping structure, which cannot establish stable state indicators from dynamic dimensions such as temperature gradient and power fluctuation; third, the system does not form an adaptive correction mechanism based on historical trends, which cannot perceive the running deviation caused by the destruction of the internal consistency of the power supply. Once the above abnormal conditions continue to accumulate, it will cause power supply response delay, sudden stop, load jump and power supply interruption of the robot in a short period of time, directly affecting the operation stability and running safety, and even causing irreversible damage to the life of the battery. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a forklift mobile robot intelligent power supply method and system based on multi-source data, which solves the problems mentioned in the background art.

[0006] To achieve the above purpose, the present application realizes the following technical scheme: a forklift mobile robot intelligent power supply system based on multi-source data, comprising a power information acquisition module, a power supply timing synchronization and noise suppression module, a dynamic feature extraction and power parameter reconstruction module, a fusion state mapping and power consistency evaluation module, a power supply health and energy state joint calculation module, and a fusion output and adaptive correction module;

[0007] The power information acquisition module acquires and pre-processes the power data through the AFE chip and the forklift mobile robot running feature sensor, and obtains a power data set DW.

[0008] The power supply timing synchronization and noise suppression module performs time synchronization and multi-scale filtering on the power supply data set DW to obtain a suppression data set SDW;

[0009] The dynamic feature extraction and power supply parameter reconstruction module extracts features from the suppression data set SDW, extracts dynamic features reflecting changes in the power supply state, and fits them into a dynamic feature set Fd;

[0010] The fusion state mapping and power supply consistency evaluation module establishes a dynamic state mapping matrix based on the dynamic feature set Fd, constructs a state mapping coefficient Φ, and performs consistency evaluation to obtain a power supply consistency deviation index ΔCvar;

[0011] The power supply health and energy state joint calculation module combines the obtained power supply consistency deviation index ΔCvar and the dynamic feature set Fd to calculate and obtain a power supply health state function Hs and an energy margin function Er;

[0012] The fusion output and adaptive correction module combines the obtained power supply health state function Hs and energy margin function Er to obtain a power supply comprehensive state function Ψsys, and judges the perception state of the power supply.

[0013] Preferably, the power supply information acquisition module includes a power supply state acquisition unit and a power supply state data processing unit;

[0014] The power supply state acquisition unit acquires power supply data, including voltage Ve, current Ie, temperature Te, acceleration av, and driving power Pw, through an AFE chip and a fork mobile robot running feature sensor;

[0015] Among them, the AFE chip is used to collect the basic running parameters of the power supply, including voltage Ve, current Ie, and temperature Te; the AFE chip is built-in with high-precision 16-bit ADC, voltage dividing resistor network, and NTC thermistor interface;

[0016] The voltage Ve is connected to the AFE input terminal through voltage dividing sampling, and is converted into a voltage ratio for ADC quantization acquisition;

[0017] The current Ie is connected in series with the main circuit through the on-chip shunt sampler sampling node, and the voltage dividing value is converted to obtain the current;

[0018] The temperature Te is collected by attaching a 10kΩ NTC patch resistor with an accuracy of 1% to the surface of the power supply and the MOSFET heat dissipation area;

[0019] The fork mobile robot running feature sensor includes an IMU inertial unit and a drive unit controller;

[0020] The acceleration av is obtained by collecting three-axis synthetic acceleration through an IMU inertia unit installed in the center of the vehicle frame;

[0021] The driving power Pw is obtained by uploading the control power output signal in real time through the feedback module in the driving unit controller.

[0022] Preferably, the power supply state data processing unit fits the obtained voltage Ve, current Ie, temperature Te, acceleration av and driving power Pw to obtain the original data set YW, and performs cleaning and normalization processing to obtain the power supply data set DW;

[0023] The cleaning includes outlier rejection, and the outlier data in the original data set YW is identified and removed by introducing a continuous point range threshold method;

[0024] The normalization processing formula is as follows:

[0025] ;

[0026] In the formula, DWo represents the oth data in the power supply data set DW, YWo represents the oth data in the original data set YW, minYWo represents the valley value of the oth data in the original data set YW, and maxYWo represents the peak value of the oth data in the original data set YW.

[0027] Preferably, the power supply timing synchronization and noise suppression module includes a time base unification and dynamic interpolation unit and a multi-window residual fusion filtering unit;

[0028] The time base unification and dynamic interpolation unit processes the time axis misalignment problem of various signals in the power supply data set DW caused by sampling frequency difference and communication delay through reference clock frame alignment and dynamic density interpolation method, and unifies the asynchronous data stream to the high-frequency main clock axis;

[0029] First, set the AFE sampling frequency as the main frequency f0, and the sampling time point set as {tk}; interpolate the missing data in the power supply data set DW by introducing a local derivative weighted interpolation method;

[0030] The sampling time point set {tk} is normalized to eliminate the dimension of time data;

[0031] The formula of the weighted interpolation method is as follows:

[0032] ;

[0033] In the formula, nDWo(t k ) represents the oth data in the power supply data set DW interpolated at time t k , λk represents a time normalization factor, , DWo(tk-1 DWo(t) represents the oth data in the power data set DW at time t k-1 DWo(t) represents the oth data in the power data set DW at time t k+1 DWo(t) represents the oth data in the power data set DW at time t k+1 DWo(t) represents the oth data in the power data set DW at time t, θ represents the slope compensation factor, the value range [0, 0.2], used to reserve the signal micro-disturbance dynamic trend, and ΔDWo represents the signal slope; .

[0034] The multi-window residual fusion filtering unit is based on a sliding window structure of different time scales, combines high-frequency detection residual and statistical anomaly analysis, performs multi-level noise suppression and disturbance identification separation processing, and obtains a suppressed data set SDW;

[0035] S1, constructing a multi-scale sliding window: setting two time scale windows, a micro window Ws with a length of 5 points for capturing high-frequency disturbances, and a macro window Wl with a length of 20 points; and obtaining micro window mean μs and macro window mean μl;

[0036] S2, constructing a residual ratio factor as a dynamic correction quantity: at time t, the micro window mean μs and the macro window mean μl are subtracted to obtain a deviation, and then the absolute value of the deviation is taken; then the absolute deviation value is divided by the absolute value of the macro window mean μl to obtain a disturbance residual ratio XEr;

[0037] S3, fusion filtering based on residual weight, final smoothing output residual adaptive fusion strategy;

[0038] The formula is as follows: SDWo(t) = ω(t) × μl(t) + [1- ω(t)] × μs(t);

[0039] In the formula, ω(t) represents the smoothing weight coefficient at time t, μs(t) represents the micro window mean at time t, μl(t) represents the macro window mean at time t, SDWo(t) represents the oth data in the suppressed data set SDW at time t, and specifically represents the first data channel in the suppressed data set SDW; and all data channels are independently calculated, wherein the smoothing weight coefficient ω is obtained by the ratio of 1 to 1 plus the disturbance residual ratio XEr;

[0040] Preferably, the dynamic feature extraction and power parameter reconstruction module includes a power supply trend feature extraction unit and a feature normalization and reconstruction unit;

[0041] The power supply trend feature extraction unit extracts features from the suppressed data set SDW, extracts change rate features and fluctuation features that reflect the change of the power supply state over time, including power supply temperature change gradient ΔTg, load fluctuation rate Δa and power change rate ΔPr;

[0042] The power temperature change gradient ΔTg is obtained in the following manner: first, the temperature change rate of the power is obtained by subtracting the previous temperature from the current temperature of the power and dividing by the time interval; then, the absolute value of the temperature change rate of each power is taken, and the temperature change rates of all powers are averaged to obtain the power temperature change gradient ΔTg;

[0043] The load fluctuation rate Δa is obtained in the following manner: first, the acceleration data in a period of time is continuously obtained from the IMU, and the average value of all acceleration data in the period of time is calculated; the difference between the acceleration value of each sampling point in the period of time and the average acceleration obtained just now is calculated; then, the difference is squared to obtain the average value; finally, the average value is squared to obtain the load fluctuation rate Δa;

[0044] The power change rate ΔPr is obtained in the following manner: first, the difference between the driving power Pw(t) at time t and the driving power Pw(t-Δt) at time t-Δt is calculated; then, the difference is divided by the driving power Pw(t-Δt) at time t-Δt to obtain the relative change ratio, and finally, the absolute value of the ratio is taken to obtain the power change rate ΔPr;

[0045] The feature normalization and reconstruction unit normalizes and standardizes the obtained power temperature change gradient ΔTg, load fluctuation rate Δa and power change rate ΔPr, and fits them into a dynamic feature set Fd.

[0046] Preferably, the fusion state mapping and power consistency evaluation module includes a state mapping construction unit and a power consistency evaluation unit.

[0047] The state mapping construction unit combines the dynamic feature set Fd with the voltage Ve and temperature Te of the power to establish a mapping relationship and construct a state mapping coefficient Φ.

[0048] The state mapping coefficient Φ is obtained by the following formula:

[0049] ;

[0050] In the formula, Φi(t) represents the state mapping coefficient of the i-th power at time t, Tei(t) represents the temperature of the i-th power at time t, pTe(t) represents the average temperature of all powers at time t, Vei(t) represents the voltage of the i-th power at time t, pVe(t) represents the average voltage of all powers at time t, ΔTg(t) represents the power temperature change gradient at time t, Δa(t) represents the load fluctuation rate at time t, and ΔPr(t) represents the power change rate at time t.

[0051] The acquisition formula of the state mapping coefficient Φ is derived from the standardized residual in mathematics and the disturbance suppression in signal analysis.

[0052] The standardized residual in mathematics: similar to the Z-score commonly used in statistics, the difference between the individual value of a variable and the overall mean is divided by the fluctuation scale;

[0053] In the formula, the difference between the current power state (temperature, voltage) and the system average is divided by the dynamic disturbance factor (such as temperature change rate, power change rate) to achieve dynamic adjustment; this structure is essentially a "difference relativization process", the purpose is to strip off the overall jitter caused by environmental or system disturbance, and make the individual deviation more significant;

[0054] The disturbance suppression in signal analysis: the denominator adds ΔTg(t)+1, 1+ΔPr(t), which is similar to the dynamic gain adjustment commonly used in signal processing; by comparing the real-time disturbance index value, the evaluation error caused by the overall system change is suppressed; it belongs to the typical fractional dynamic suppression formula structure, commonly used in robust system modeling.

[0055] Preferably, the power consistency evaluation unit statistically evaluates the state mapping coefficients Φ of all power supplies, calculates the power consistency deviation index ΔCvar, and quantifies the internal balance of the entire power supply group;

[0056] The acquisition method of the power consistency deviation index ΔCvar is as follows: first, extract the state mapping coefficient of each power supply at time t, calculate the average value of the state mapping coefficients of all power supplies at time t; then calculate the deviation between the state mapping coefficient of the power supply and the average value, and square the deviation value; sum all the deviation squared values of the power supplies and average them; finally, take the square root to obtain the power consistency deviation index ΔCvar;

[0057] Compare the obtained power consistency deviation index ΔCvar with the preset deviation index ratio Tcvar to determine the health status of the power supply;

[0058] The health status of the power supply is matched and obtained by the following method:

[0059] When the power consistency deviation index ΔCvar≤deviation index ratio Tcvar, it indicates that the power supply state is uniform and in a healthy state;

[0060] When the power consistency deviation index ΔCvar>deviation index ratio Tcvar, it indicates that the power supply is abnormal and deviates from the normal behavior, triggering active balancing adjustment to adjust the state of the power supply.

[0061] Preferably, the power supply health and energy state joint calculation module comprises a state function construction unit and a state index output unit;

[0062] The state function construction unit combines the acquired power consistency deviation index ΔCvar and the dynamic feature set Fd to construct a power supply health state function Hs;

[0063] The power supply health state function Hs is acquired by the following formula:

[0064] ;

[0065] In the formula, ΔCvar(t) represents the power consistency deviation index at time t, c1 represents a temperature gradient adjustment coefficient, and the value range is [0.1, 1.0], c2 represents a load fluctuation adjustment coefficient, and the value range is [0.1, 1.0];

[0066] The state index output unit further combines the current power supply output power and the driving power demand to construct an energy margin function Er;

[0067] The constructed energy margin function Er is acquired by the following formula:

[0068] ;

[0069] In the formula, Ve(t) represents the voltage at time t, Ie(t) represents the current at time t, p1 represents a balance adjustment factor, and the value range is [0.05, 1.0], EP represents a non-zero constant, and the value range is [10 -6 , 10 -2 ].

[0070] Preferably, the fusion output and adaptive correction module comprises a comprehensive state function construction unit and a perceived state judgment correction unit;

[0071] The comprehensive state function construction unit constructs a power supply comprehensive state function Ψsys of a nonlinear coupling structure through the power supply health state function Hs and the energy margin function Er;

[0072] The power supply comprehensive state function Ψsys is acquired by the following formula:

[0073] ;

[0074] In the formula, Ψsys represents the power supply comprehensive state function at time t;

[0075] The perceived state judgment correction unit compares the acquired power supply comprehensive state function Ψsys and the state threshold Tsys in a fixed period, and performs hierarchical judgment, and different running suggestions and parameter correction mechanisms are automatically triggered in different state intervals;

[0076] The state threshold Tsys is obtained in the following manner: at time t, first, the integrated state function value in a historical time window is counted, the average value μΨ of the power supply integrated state function is calculated; at the same time, the standard deviation of the power supply integrated state function is calculated; then, the state threshold at the current time is obtained by subtracting the product of the standard deviation and the sensitivity coefficient from the average value;

[0077] The state classification judgment method is as follows:

[0078] When the power supply integrated state function Ψsys is greater than the average value μΨ of the power supply integrated state function, it indicates a stable operation state; the current control parameter is maintained, and no intervention is needed.

[0079] When the state threshold Tsys is less than the power supply integrated state function Ψsys and is less than or equal to the average value μΨ of the power supply integrated state function, it indicates a mild attention state; a slight amplitude parameter correction is triggered, including adjusting the cooling strategy or reducing the charging rate.

[0080] When the power supply integrated state function Ψsys is less than or equal to the state threshold Tsys, it indicates an abnormal risk state; it is suggested to switch to a protection mode or stop discharging, and logs are recorded for subsequent tracking.

[0081] The intelligent power supply method for the forked mobile robot based on multi-source data comprises the following steps:

[0082] Step one, the power supply information acquisition module collects power supply data through the AFE chip and the forked mobile robot running feature sensor, and performs preprocessing to obtain a power supply data set DW.

[0083] Step two, the power supply timing synchronization and noise suppression module performs time synchronization and multi-scale filtering on the power supply data set DW to obtain a suppressed data set SDW.

[0084] Step three, the dynamic feature extraction and power supply parameter reconstruction module extracts features from the suppressed data set SDW to extract dynamic features reflecting the state change of the power supply, and fits them into a dynamic feature set Fd.

[0085] Step four, the fusion state mapping and power consistency evaluation module establishes a dynamic state mapping matrix based on the dynamic feature set Fd, constructs a state mapping coefficient Φ, and performs consistency evaluation to obtain a power consistency deviation index ΔCvar.

[0086] Step five, the power supply health and energy state joint calculation module combines the power consistency deviation index ΔCvar and the dynamic feature set Fd to calculate and obtain a power supply health state function Hs and an energy margin function Er.

[0087] Step six, the fusion output and adaptive correction module combines the obtained power health state function Hs and energy margin function Er, obtains a power supply comprehensive state function Ψsys, and judges the perception state of the power supply.

[0088] The application provides a forked mobile robot intelligent power supply method and system based on multi-source data, which has the following beneficial effects:

[0089] (1) When the system is running, the high timeliness, multi-dimension and low error dynamic analysis of the power supply running state is realized by constructing a multi-module collaborative structure. On the premise of not relying on newly added hardware, the system significantly improves the synchronization accuracy and anti-interference ability of the original power supply signal by introducing an AFE chip and running characteristic sensor to jointly construct a power supply data group, combining time synchronization, dynamic interpolation and residual fusion filtering; further introducing power temperature gradient, load fluctuation rate and power change rate and other dynamic parameters to construct a feature set, realizing high sensitivity identification of weak changes and state fluctuations in the power supply running process.

[0090] By constructing a standardized residual mapping coefficient Φ and a consistency deviation index ΔCvar, the internal running balance of the power supply is quantitatively analyzed; the coupling comprehensive state function Ψsys is introduced, which effectively establishes the mapping channel between the perception state and the running parameter, and improves the state self-perception and parameter correction ability of the system under multiple working conditions. Compared with the existing method, the application enhances the robust regulation and control ability of the power supply system and improves the response and processing accuracy of the system to running fluctuations, energy abnormalities and balance deviations on the basis of maintaining the structural universality and interface compatibility.

[0091] (2) By constructing the power supply timing synchronization and noise suppression module and the dynamic feature extraction and power supply parameter reconstruction module, the time axis unification, data missing repair and multi-level noise suppression collaborative mechanism are realized in the multi-source asynchronous power supply data processing, which further improves the continuity, stability and time sequence correlation expression ability of the power supply data. Specifically, the use of reference clock frame alignment and dynamic interpolation strategy solves the time misalignment problem caused by the difference in sampling frequency or communication delay of different sensing channels, so that various power supply parameters can be fused and calculated under the unified master clock axis, enhancing the data comparability and processing accuracy in the time dimension.

[0092] (3) By combining the current power supply voltage, current data and real-time power demand signal of the robot driving unit, an energy margin function of the power supply is constructed, which is used to reflect whether the power supply has sufficient electrical output capacity to meet the operation demand under the current state. This method complements the deficiency of traditional BMS in "supply-demand matching accuracy", not only can realize the accurate calculation of residual energy, but also provides an important reference for subsequent power scheduling, peak limitation and load distribution, and improves the power regulation flexibility and response efficiency of the whole system.

[0093] (4) The nonlinear coupling mechanism is introduced by constructing the unit through the comprehensive state function, which organically integrates the power supply health state function and the energy margin function, so that the power supply comprehensive state function not only contains the consistency information between the battery packs, but also reflects the boundary state of the actual power supply capacity. The problem of "single index isolated judgment" in traditional health evaluation is solved. The coupling structure can integrate and describe the nonlinear interaction of different state factors, so as to more comprehensively and dynamically reflect the comprehensive bearing capacity of the power supply system in the running process.

[0094] The perception state judgment correction unit establishes an adaptive state threshold generation method based on time-varying window, which breaks through the limitation of traditional fixed threshold mechanism on the response lag of running state fluctuation. Through historical window statistical analysis combined with sensitive factor dynamic correction of current judgment threshold, the state discrimination process has time sequence sensitivity and steady state drift adaptability, which can maintain the flexible response ability to slight abnormality and early degradation signal in system state fluctuation. BRIEF DESCRIPTION OF DRAWINGS

[0095] Figure 1 The figure is a flowchart of the intelligent power supply system of the forked mobile robot based on multi-source data of the application;

[0096] Figure 2 The figure is a flowchart of the intelligent power supply method of the forked mobile robot based on multi-source data of the application;

[0097] Figure 3 The figure is a flowchart of the power supply comprehensive state function of the application;

[0098] Figure 4 The figure is a trend chart of the power supply comprehensive state function of the application. DETAILED DESCRIPTION

[0099] The technical solutions in the embodiments of the application will be described clearly and completely in combination with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0100] Embodiment 1

[0101] The application provides a fork mobile robot intelligent power supply system based on multi-source data, please refer to Figures 1 to 4 , comprising a power information acquisition module, a power supply time sequence synchronization and noise suppression module, a dynamic feature extraction and power parameter reconstruction module, a fusion state mapping and power consistency evaluation module, a power supply health and energy state joint calculation module, and a fusion output and adaptive correction module.

[0102] The power information acquisition module collects power data through an AFE chip and a fork mobile robot running feature sensor, and performs preprocessing to obtain a power data set DW.

[0103] The power supply time sequence synchronization and noise suppression module performs time synchronization and multi-scale filtering on the power data set DW to obtain a suppression data set SDW.

[0104] The dynamic feature extraction and power parameter reconstruction module extracts features from the suppression data set SDW to extract dynamic features reflecting changes in the power state, and fits them into a dynamic feature set Fd.

[0105] The fusion state mapping and power consistency evaluation module establishes a dynamic state mapping matrix based on the dynamic feature set Fd, constructs a state mapping coefficient Φ, and performs consistency evaluation to obtain a power consistency deviation index ΔCvar.

[0106] The power supply health and energy state joint calculation module combines the obtained power consistency deviation index ΔCvar and the dynamic feature set Fd to calculate and obtain a power health state function Hs and an energy margin function Er.

[0107] The fusion output and adaptive correction module combines the obtained power health state function Hs and energy margin function Er to obtain a power supply comprehensive state function Ψsys, and judges the perception state of the power supply.

[0108] In this embodiment, by constructing a multi-module linked fork mobile robot intelligent power supply system, the perception accuracy and dynamic response capability of the power supply state are improved while the system hardware structure remains unchanged. It obtains a power data set covering multiple physical quantities through the power information acquisition module, and introduces an AFE chip and a running feature sensor to realize integrated collection and structured preprocessing of multi-source parameters such as current, voltage, temperature, vibration, etc., providing high-credibility data support for subsequent modeling; the power supply time sequence synchronization and noise suppression module uses time alignment mechanism and multi-scale filtering strategy to enhance the alignment consistency of different channel data, and effectively suppresses the multi-frequency interference and sampling error in the signal.

[0109] The dynamic feature extraction and parameter reconstruction module can extract dynamic features reflecting changes in power supply aging, load disturbance, power supply fluctuation and other working conditions, realize dynamic modeling of nonlinear and multidimensional parameters, and establish a function mapping relationship between multiple power supply states based on state mapping and consistency evaluation module, and identify the stability difference between power supply paths based on consistency deviation index, which helps to identify abnormal paths in advance.

[0110] The power supply health and energy state joint calculation module establishes a dual evaluation mechanism of health degree and energy margin, which can comprehensively reflect the current remaining available capacity and potential risk level of the power supply system. Finally, through the fusion output and adaptive correction module, real-time perception, hierarchical judgment and operation instruction self-adjustment mechanism of comprehensive state are realized, which improves the resolution and control feedback efficiency of the system to abnormal state. Overall, the system has the characteristics of high fusion degree, strong dynamic response and excellent abnormal fault tolerance, and is suitable for intelligent power supply management needs of the fork mobile robot in complex tasks and variable environments.

[0111] Embodiment 2

[0112] This embodiment is an explanation and description in embodiment 1, please refer to Figure 3 , in particular: the power supply information acquisition module includes a power supply state acquisition unit and a power supply state data processing unit;

[0113] The power supply state acquisition unit collects power supply data through AFE chips and fork mobile robot running feature sensors, including voltage Ve, current Ie, temperature Te, acceleration av and driving power Pw;

[0114] Among them, the AFE chip is used to collect the basic running parameters of the power supply level, including voltage Ve, current Ie and temperature Te; the AFE chip is built-in with high-precision 16-bit ADC, voltage dividing resistor network and NTC thermistor interface;

[0115] The voltage Ve is connected to the AFE input terminal through voltage dividing sampling, and is converted into voltage ratio for ADC quantization acquisition;

[0116] The current Ie is connected in series with the main loop through the on-chip shunt sampler sampling node, and the voltage dividing value is converted into current for acquisition;

[0117] The temperature Te is collected by attaching a 10kΩ NTC patch resistor with an accuracy of 1% to the surface of the power supply and the MOSFET heat dissipation area;

[0118] The fork mobile robot running feature sensor includes an IMU inertial unit and a drive unit controller;

[0119] The acceleration av is collected by the IMU inertial unit installed in the central frame to obtain the three-axis combined acceleration;

[0120] The driving power Pw is uploaded in real time by the feedback module in the driving unit controller to control the power output signal collection.

[0121] The power state data processing unit fits the acquired voltage Ve, current Ie, temperature Te, acceleration av and driving power Pw to obtain an original data set YW, and performs cleaning and normalization processing to obtain a power data set DW.

[0122] The cleaning includes abnormal point elimination, and the abnormal data in the original data set YW is identified and removed by introducing a consecutive point range threshold method.

[0123] The normalization processing formula is as follows:

[0124] ;

[0125] In the formula, DWo represents the oth data in the power data set DW, YWo represents the oth data in the original data set YW, minYWo represents the valley value of the oth data in the original data set YW, and maxYWo represents the peak value of the oth data in the original data set YW.

[0126] In this embodiment, by refining the power information collection module into a power state collection unit and a power state data processing unit, the running state of the power supply system of the fork mobile robot is realized. Fine, multi-dimensional and high-precision dynamic monitoring.

[0127] The power state collection unit integrates AFE chips and running characteristic sensors to complete the synchronous collection of voltage, current, temperature, acceleration and power and other multi-source parameters. Among them, the AFE chip internally integrates a high-precision analog-to-digital converter, a voltage dividing network and a thermal sensitive interface, which ensures high-fidelity sampling of the core parameters of the power supply. The coupling layout of NTC thermistor and MOSFET region makes the temperature acquisition closer to the actual heat source characteristics, which helps to identify the risk of thermal runaway early. At the same time, the introduction of running characteristic parameters such as vehicle body acceleration and driving power provides dynamic context information such as load disturbance and running intensity for subsequent feature extraction, improving the scene perception ability of the data.

[0128] The power state data processing unit introduces a continuous point range threshold method to identify and remove fluctuation abnormal points on the basis of completing the fitting of the original data set, which can effectively filter non-physical fluctuations caused by interference, electromagnetic noise or mechanical impact, and improve the stability and robustness of subsequent data calculation. On this basis, unified normalization processing is performed on each item of data, and parameters of different dimensions and different dimensions are uniformly mapped to a relative quantization interval, which facilitates subsequent multivariate model fusion and state evaluation. Overall, the construction of this module not only expands the state perception dimension of the power supply system, but also enhances the reliability, fusion adaptability and dynamic expression ability of the data through processing strategies, providing a high-quality data basis for subsequent power supply state modeling and health evaluation.

[0129] Embodiment 3

[0130] This embodiment is an explanation and description in embodiment 2, please refer to Figure 1 and Figure 3 , specifically: the power timing synchronization and noise suppression module includes a time base unification and dynamic interpolation unit and a multi-window residual fusion filtering unit;

[0131] The time base unification and dynamic interpolation unit processes the time axis misalignment problem of various signals in the power data set DW caused by sampling frequency difference and communication delay through reference clock frame alignment and dynamic density interpolation method, and uniformly maps the asynchronous data stream to the high-frequency master clock axis;

[0132] First, set the AFE sampling frequency as the main frequency f0, and the sampling time point set as {tk}; introduce a local derivative weighted interpolation method to interpolate the missing data in the power data set DW;

[0133] The formula of the weighted interpolation method is as follows:

[0134] ;

[0135] In the formula, nDWo(t k ) represents the oth item of data in the power data set DW interpolated at time t k , λk represents a time normalization factor, DWo(t k-1 ) represents the oth item of data in the power data set DW at time t k-1 , DWo(t k+1 ) represents the oth item of data in the power data set DW at time t k+1 , and θ represents a slope compensation factor, and ΔDWo represents the signal slope;

[0136] The multi-window residual fusion filtering unit is based on a sliding window structure of different time scales, combines high-frequency detection residual and statistical anomaly analysis, performs multi-level noise suppression and disturbance identification and separation processing, and obtains a suppression data set SDW;

[0137] S1, constructing a multi-scale sliding window: setting two time scale windows, a micro window Ws with a length of 5 points, and a macro window Wl with a length of 20 points, and obtaining a micro window mean value μs and a macro window mean value μl;

[0138] S2, constructing a residual ratio factor as a dynamic correction amount: at time t, the micro window mean value μs and the macro window mean value μl are subtracted to obtain a deviation, and then the absolute value of the deviation is taken; then the absolute deviation value is divided by the absolute value of the long-time window mean value to obtain a disturbance residual ratio XEr;

[0139] S3, fusion filtering based on a residual weight, and finally smoothing an output residual adaptive fusion strategy;

[0140] The formula is as follows: SDWo(t) = ω(t) x μl(t) + [1-ω(t)] x μs(t);

[0141] In the formula, ω(t) represents a smoothing weight coefficient at time t, μs(t) represents a micro window mean value at time t, μl(t) represents a macro window mean value at time t, and SDWo(t) represents the oth data in the suppression data group SDW at time t.

[0142] The dynamic feature extraction and power parameter reconstruction module includes a power supply trend feature extraction unit and a feature normalization and reconstruction unit;

[0143] The power supply trend feature extraction unit extracts features from the suppression data group SDW, extracts change rate features and fluctuation features reflecting the change of the power supply state over time, including power supply temperature change gradient ΔTg, load fluctuation rate Δa and power change rate ΔPr;

[0144] The power supply temperature change gradient ΔTg is obtained in the following manner: first, the current temperature of the power supply is subtracted from the previous temperature, and then divided by the time interval to obtain the temperature change rate of the power supply; then, the absolute value of the temperature change rate of each power supply is taken, and the temperature change rates of all power supplies are averaged to obtain the power supply temperature change gradient ΔTg;

[0145] The load fluctuation rate Δa is obtained in the following manner: first, the acceleration data within a period of time is continuously obtained from the IMU, and the average value of all acceleration data within the period of time is calculated; the difference between the acceleration value of each sampling point within the period of time and the average acceleration obtained just now is calculated; then the difference values are squared to obtain the average value; finally, the average value is squared to obtain the load fluctuation rate Δa;

[0146] The power change rate ΔPr is obtained in the following manner: first, the difference between the driving power Pw(t) at time t and the driving power Pw(t-Δt) at time t-Δt is calculated; then the difference is divided by the driving power Pw(t-Δt) at time t-Δt to obtain a ratio of the relative change; and finally, the absolute value of the ratio is taken to obtain the power change rate ΔPr.

[0147] The feature normalization and reconstruction unit normalizes and standardizes the obtained power temperature change gradient ΔTg, the load fluctuation rate Δa, and the power change rate ΔPr, and fits them into a dynamic feature set Fd.

[0148] The fusion state mapping and power consistency evaluation module includes a state mapping construction unit and a power consistency evaluation unit.

[0149] The state mapping construction unit, based on the dynamic feature set Fd, combines the voltage Ve and the temperature Te of the power supply to establish a mapping relationship and construct a state mapping coefficient Φ.

[0150] The state mapping coefficient Φ is obtained by the following formula:

[0151] ;

[0152] In the formula, Φi(t) represents the state mapping coefficient of the i-th power supply at time t, Tei(t) represents the temperature of the i-th power supply at time t, pTe(t) represents the average temperature of all power supplies at time t, Vei(t) represents the voltage of the i-th power supply at time t, pVe(t) represents the average voltage of all power supplies at time t, ΔTg(t) represents the power temperature change gradient at time t, Δa(t) represents the load fluctuation rate at time t, and ΔPr(t) represents the power change rate at time t.

[0153] In this embodiment, by constructing the "power supply timing synchronization and noise suppression module" and the "dynamic feature extraction and power parameter reconstruction module", a collaborative mechanism of time axis unification, data missing repair and multi-level noise suppression is realized in multi-source asynchronous power supply data processing, further improving the continuity, stability and time sequence correlation expression ability of the power supply data. Specifically, the use of reference clock frame alignment and dynamic interpolation strategy solves the time misalignment problem caused by differences in sampling frequency or communication delay in different sensing channels, enabling various power supply parameters to be fused and calculated under a unified master clock axis, enhancing the data comparability and processing accuracy in the time dimension. At the same time, the introduction of a multi-scale sliding window structure and residual ratio analysis realizes the comprehensive modeling of short-term disturbances and long-term trends, while maintaining the change sensitivity and suppressing random noise, improving the reliability of subsequent feature recognition.

[0154] The dynamic feature extraction module constructs time variation features around "power temperature variation gradient", "load fluctuation rate" and "power change rate"; these features are from the change rate and fluctuation degree of the sensing data, which can reflect the state evolution trend of the power supply under complex operating environment, and are different from the traditional static mean or extreme value judgment, and have stronger time sensitivity and dynamic adaptability. Finally, the state mapping coefficient is constructed by mapping and combining the extracted features and the basic voltage temperature parameters, and the structure expression reflecting the difference and consistency deviation degree of the power supply is established, which provides parameter basis for subsequent power supply health state evaluation and collaborative power supply regulation.

[0155] Embodiment 4

[0156] This embodiment is an explanation and description in embodiment 3, please refer to Figure 1 , specifically: the power consistency evaluation unit statistically evaluates the state mapping coefficient Φ of all power supplies, calculates the power consistency deviation index ΔCvar, and quantifies the internal balance of the entire power supply group;

[0157] The acquisition method of the power consistency deviation index ΔCvar is: first, extract the state mapping coefficient of each power supply at time t, calculate the average value of the state mapping coefficient of all power supplies at time t; then calculate the deviation between the state mapping coefficient of the power supply and the average value, and square the deviation value; sum all the deviation square values of the power supplies and average them; finally, take the square root to obtain the power consistency deviation index ΔCvar;

[0158] Compare the obtained power consistency deviation index ΔCvar with the preset deviation index ratio Tcvar to judge the health state of the power supply;

[0159] The health state of the power supply is matched and obtained by the following way:

[0160] When the power consistency deviation index ΔCvar≤deviation index ratio Tcvar, it indicates that the power supply state is uniform and in a healthy state;

[0161] When the power consistency deviation index ΔCvar>deviation index ratio Tcvar, it indicates that the power supply is abnormal and the behavior is out of control, triggering active balancing adjustment to adjust the state of the power supply.

[0162] The power supply health and energy state joint calculation module includes a state function construction unit and a state index output unit;

[0163] The state function construction unit combines the obtained power consistency deviation index ΔCvar and the dynamic feature set Fd to construct the power supply health state function Hs;

[0164] The power supply health state function Hs is obtained by the following formula:

[0165] ;

[0166] In the formula, ΔCvar(t) represents the power consistency deviation index at time t, c1 represents a temperature gradient adjustment coefficient, and c2 represents a load fluctuation adjustment coefficient;

[0167] The state index output unit further constructs an energy margin function Er in combination with the current power supply output power and the driving power demand;

[0168] The energy margin function Er is obtained by the following formula:

[0169] ;

[0170] In the formula, Ve(t) represents the voltage at time t, Ie(t) represents the current at time t, p1 represents a balance adjustment factor, and EP represents a non-zero constant.

[0171] In the embodiment, by setting the "power consistency evaluation unit" and the "power supply health and energy state joint calculation module", a fusion processing mechanism from micro power state mapping to macro health evaluation is realized in the fork type mobile robot power supply system. By statistically analyzing the state mapping coefficients of all power supplies, a "consistency deviation index" reflecting the balanced state between the power supply groups is constructed. Not only can the difference degree of each power unit in the running behavior be quantitatively described, but also the behavior deviation trend can be identified at the initial stage of data fluctuation, thereby providing time buffer and early warning basis for subsequent risk intervention. Compared with the traditional threshold alarm mode, the structure realizes the forward perception and hierarchical response of abnormal state, and enhances the identification ability of the system to local imbalance in the parallel power supply scene of multiple power supplies.

[0172] In the embodiment, the energy margin function of the power supply is constructed by combining the current power supply voltage, current data and the real-time power demand signal of the robot driving unit, which is used to reflect whether the power supply has sufficient electrical output capacity to meet the operation demand under the current state. This method complements the deficiency of the traditional BMS in "supply-demand matching accuracy". It not only realizes the accurate calculation of the remaining energy, but also provides an important reference for subsequent power scheduling, peak limitation and load distribution, and improves the power regulation flexibility and response efficiency of the whole system.

[0173] Embodiment 5

[0174] This embodiment is an explanation and description in embodiment 4. Please refer to Figure 3 and Figure 4 Specifically, the fusion output and adaptive correction module includes a comprehensive state function construction unit and a perception state judgment correction unit.

[0175] The comprehensive state function construction unit constructs a power supply comprehensive state function Ψsys of a nonlinear coupling structure through the power supply health state function Hs and the energy margin function Er.

[0176] The power supply comprehensive state function Ψsys is obtained through the following formula:

[0177] ;

[0178] Ψsys represents the power supply comprehensive state function at time t.

[0179] The perception state determination correction unit compares the obtained power supply comprehensive state function Ψsys and the state threshold Tsys in a fixed period, and performs hierarchical judgment, and automatically triggers different operation suggestions and parameter correction mechanisms in different state intervals.

[0180] The state threshold Tsys is obtained as follows: at time t, first, the comprehensive state function values in a historical time window are counted to calculate the average value μΨ of the power supply comprehensive state function; at the same time, the standard deviation of the power supply comprehensive state function is calculated; then, the state threshold at the current moment is obtained by subtracting the result of the standard deviation multiplied by the sensitivity coefficient from the mean value.

[0181] The state hierarchical judgment method is as follows:

[0182] When the power supply comprehensive state function Ψsys is greater than the average value μΨ of the power supply comprehensive state function, it represents a stable operation state.

[0183] When the state threshold Tsys is less than the power supply comprehensive state function Ψsys and is less than or equal to the average value μΨ of the power supply comprehensive state function, it represents a mild attention state.

[0184] When the power supply comprehensive state function Ψsys is less than or equal to the state threshold Tsys, it represents an abnormal risk state.

[0185] In this embodiment, the comprehensive state function construction unit introduces a nonlinear coupling mechanism to organically integrate the power supply health state function and the energy margin function, so that the power supply comprehensive state function not only contains the consistency information between the battery groups, but also reflects the boundary state of the actual power supply capacity. The problem of "single index isolated judgment" in traditional health evaluation is solved. The coupling structure can integrate and describe the nonlinear interaction of different state factors, so as to more comprehensively and dynamically reflect the comprehensive bearing capacity of the power supply system in the running process.

[0186] The perception state judgment correction unit establishes an adaptive state threshold generation method based on a time-varying window, breaking through the limitation of the response lag of the traditional fixed threshold mechanism to the running state fluctuation. Through historical window statistical analysis combined with dynamic correction of the current judgment threshold by a sensitive factor, the state discrimination process has time sequence sensitivity and steady state drift adaptability, and can maintain the flexible response ability to slight abnormalities and early degradation signals in the system state fluctuation.

[0187] The embodiment not only outputs the quantitative state function, but also sets three state intervals of stable, attention and abnormal through the perception state judgment correction unit. Under different intervals, corresponding running instructions and parameter adjustment mechanisms can be automatically triggered, such as prompting early warning strategies in slight attention, and activating active balancing or load reduction control means in abnormal state.

[0188] Embodiment 6

[0189] The intelligent power supply method of the fork mobile robot based on multi-source data, please refer to Figure 2 , Specifically: comprising the following steps:

[0190] Step one, the power information acquisition module collects power data through AFE chip and fork mobile robot running feature sensor, and pre-processes to obtain power data set DW;

[0191] Step two, the power supply time sequence synchronization and noise suppression module performs time synchronization and multi-scale filtering on the power data set DW to obtain the suppression data set SDW;

[0192] Step three, the dynamic feature extraction and power parameter reconstruction module extracts features from the suppression data set SDW to extract dynamic features reflecting the power state change, and fits them into a dynamic feature set Fd;

[0193] Step four, the fusion state mapping and power consistency evaluation module establishes a dynamic state mapping matrix based on the dynamic feature set Fd, constructs a state mapping coefficient Φ, and performs consistency evaluation to obtain a power consistency deviation index ΔCvar;

[0194] Step five, the power supply health and energy state joint calculation module combines the power consistency deviation index ΔCvar and the dynamic feature set Fd to calculate and obtain the power health state function Hs and the energy margin function Er;

[0195] Step six, the fusion output and adaptive correction module combines the power health state function Hs and the energy margin function Er to obtain the power supply comprehensive state function Ψsys, and judges the perception state of the power supply.

[0196] In this embodiment, the intelligent power supply method of the forked mobile robot based on multi-source data breaks through the problem of relying on single voltage or current signal of the traditional battery management system by constructing a power supply state perception chain covering six consecutive steps of collection, processing, extraction, mapping, evaluation and feedback. In step one, the joint collection mechanism of AFE chip and running characteristic sensor is introduced, which can not only obtain the basic electrical parameters of the power supply, but also synchronously obtain the running condition information such as acceleration, temperature and driving power, realizing the expansion of battery data dimension; in step two, by setting a unified master clock reference and combining multi-scale residual filtering, the data misplacement and noise pollution problems caused by inconsistent sampling frequency of multiple signal sources and communication delay are effectively solved, and the consistency and analyzability of data time sequence are improved.

[0197] Further, in step three, the feature extraction method reflecting state trend such as dynamic change rate and fluctuation rate is used to obtain feature parameters with time-varying sensitivity, which provides more accurate basic data for subsequent state evolution judgment; in step four, by constructing a dynamic state mapping matrix and a state mapping coefficient, the comparative analysis between different power supply monomers is effectively realized, which can quantitatively describe the consistency degree of the power supply group and overcome the problem that the traditional method cannot dynamically reflect the accumulation of group deviation.

[0198] In step five, the consistency deviation index and dynamic feature set are combined to construct a health state function and an energy margin function, so that the power supply system state evaluation is not limited to single-point abnormal detection, but also has the comprehensive judgment ability of the overall balance and residual energy capacity; finally in step six, a unified power supply comprehensive state function is constructed by using a nonlinear coupling structure to fuse various state indexes, and a hierarchical judgment mechanism with dynamic updating threshold is introduced, which can automatically perceive the risk level within the running cycle, and then match the operation suggestion or start parameter correction strategy, thereby enhancing the response ability and running stability of the system under complex tasks.

[0199] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and deformations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A forked mobile robot intelligent power supply system based on multi-source data, characterized in that: The power supply information acquisition module, the power supply time sequence synchronization and noise suppression module, the dynamic characteristic extraction and power supply parameter reconstruction module, the fusion state mapping and power supply consistency evaluation module, the power supply health and energy state joint calculation module, and the fusion output and adaptive correction module are included. The power supply information acquisition module collects power supply data through an AFE chip and a fork mobile robot running characteristic sensor, and performs preprocessing to obtain a power supply data set DW; The power supply information acquisition module includes a power supply state acquisition unit and a power supply state data processing unit; The power supply state acquisition unit collects power supply data through an AFE chip and a fork mobile robot running characteristic sensor, including voltage Ve, current Ie, temperature Te, acceleration av, and driving power Pw; The power supply time sequence synchronization and noise suppression module performs time synchronization and multi-scale filtering on the power supply data set DW to obtain a suppression data set SDW; The dynamic characteristic extraction and power supply parameter reconstruction module extracts features from the suppression data set SDW to extract dynamic characteristics reflecting changes in the power supply state, and fits them into a dynamic feature set Fd; The fusion state mapping and power supply consistency evaluation module establishes a dynamic state mapping matrix based on the dynamic feature set Fd, constructs a state mapping coefficient Φ, and performs consistency evaluation to obtain a power supply consistency deviation index ΔCvar; The fusion state mapping and power supply consistency evaluation module includes a state mapping construction unit and a power supply consistency evaluation unit; The state mapping construction unit combines the dynamic feature set Fd with the voltage Ve and temperature Te of the power supply to establish a mapping relationship and construct a state mapping coefficient Φ; The state mapping coefficient Φ is obtained by the following formula: ； In the formula, Φi(t) represents the state mapping coefficient of the i-th power supply at time t, Tei(t) represents the temperature of the i-th power supply at time t, pTe(t) represents the average temperature of all power supplies at time t, Vei(t) represents the voltage of the i-th power supply at time t, pVe(t) represents the average voltage of all power supplies at time t, ΔTg(t) represents the power supply temperature change gradient at time t, Δa(t) represents the load fluctuation rate at time t, and ΔPr(t) represents the power change rate at time t; The power supply health and energy state joint calculation module jointly calculates the power supply consistency deviation index ΔCvar and the dynamic feature set Fd to obtain a power supply health state function Hs and an energy margin function Er; The fusion output and adaptive correction module combines the power supply health state function Hs and the energy margin function Er to obtain a power supply comprehensive state function Ψsys, and judges the perception state of the power supply.

2. The multi-source data based intelligent power supply system for a forked mobile robot according to claim 1, wherein: The power supply state acquisition unit collects the voltage Ve, current Ie, and temperature Te of the power supply level basic running parameters through the AFE chip; the AFE chip has a built-in high-precision 16-bit ADC, a voltage dividing resistor network, and an NTC thermistor interface; The voltage Ve is connected to the AFE input end through voltage dividing sampling, converted to a voltage ratio, and quantified by ADC to obtain; The current Ie is connected in series with the main circuit through the on-chip shunt sampler sampling node, and the voltage dividing value is converted to obtain the current. The temperature Te is acquired by a 10kΩ NTC patch resistor with an accuracy of 1% attached to the surface of the power supply and the heat dissipation area of the MOSFET; The fork mobile robot operation characteristic sensor comprises an IMU inertial unit and a driving unit controller; The acceleration av is acquired by an IMU inertial unit installed in the center of the vehicle frame to acquire the three-axis combined acceleration; The driving power Pw is acquired by the real-time uploading of the control power output signal through the feedback module in the driving unit controller.

3. The multi-source data based forklift mobile robot intelligent power supply system according to claim 1, characterized in that: The power supply state data processing unit fits the acquired voltage Ve, current Ie, temperature Te, acceleration av and driving power Pw to obtain an original data set YW, and performs cleaning and normalization processing to obtain a power supply data set DW; The cleaning includes abnormal point elimination, and the abnormal data in the original data set YW is identified and removed by introducing a continuous point range threshold method; The normalization processing formula is as follows: ； In the formula, DWo represents the oth data in the power supply data set DW, YWo represents the oth data in the original data set YW, minYWo represents the valley value of the oth data in the original data set YW, and maxYWo represents the peak value of the oth data in the original data set YW.

4. The multi-source data based forklift mobile robot intelligent power supply system according to claim 3, characterized in that: The power supply timing synchronization and noise suppression module comprises a time base unification and dynamic interpolation unit and a multi-window residual fusion filtering unit; The time base unification and dynamic interpolation unit processes the time axis misalignment problems of various signals in the power supply data set DW caused by sampling frequency differences and communication delays through reference clock frame alignment and dynamic density interpolation method, and unifies the asynchronous data stream to the high-frequency main clock axis; Firstly, the AFE sampling frequency is set as the main frequency f0, and the sampling time point set is recorded as {tk}; the missing data in the power supply data set DW is interpolated by introducing a local derivative weighted interpolation method; The formula of the weighted interpolation method is as follows: ； In the formula, nDWo(t) k ) represents time t k The o-th data item in the interpolated power data set DW, where λk represents the time normalization factor, DWo(t) k-1 ) represents time t k-1 The oth data item in the power data group DW, DWo(t) k+1 ) represents time t k+1 The o-th data in the power data group DW at that time, where θ represents the slope compensation factor and ΔDWo represents the signal slope; The multi-window residual fusion filtering unit is based on a sliding window structure of different time scales, combines high-frequency detection residual and statistical abnormality analysis, performs multi-level noise suppression and disturbance identification and separation processing, and obtains a suppression data set SDW; S1, construct a multi-scale sliding window: set two time scale windows, a micro window Ws with a length of 5 points, and a macro window Wl with a length of 20 points; and obtain the micro window mean μs and the macro window mean μl; S2, construct a residual ratio factor as a dynamic correction quantity: at time t, the micro window mean μs and the macro window mean μl are subtracted to obtain a deviation, and then the absolute value of the deviation is taken; then the absolute deviation value is divided by the absolute value of the long-time window mean to obtain the disturbance residual ratio XEr; S3, residual weight-based fusion filtering, finally adaptive fusion strategy of smoothing output residual; The formula is as follows: SDWo(t)=ω(t)×μl(t)+[1-ω(t)]×μs(t); In the formula, ω(t) represents the smoothing weight coefficient at time t, μs(t) represents the micro window mean at time t, μl(t) represents the macro window mean at time t, and SDWo(t) represents the oth data in the suppression data set SDW at time t.

5. The multi-source data based forklift mobile robot intelligent power supply system according to claim 4, characterized in that: The dynamic characteristic extraction and power supply parameter reconstruction module comprises a power supply trend characteristic extraction unit and a characteristic normalization and reconstruction unit; The power supply trend feature extraction unit extracts features from the suppression data set SDW, extracts the change rate feature and the fluctuation feature reflecting the change of the power supply state over time, including the power supply temperature change gradient ΔTg, the load fluctuation rate Δa and the power change rate ΔPr; The acquisition method of the power supply temperature change gradient ΔTg is: first, the current temperature of the power supply is subtracted from the previous temperature, and then divided by the time interval to obtain the temperature change rate of the power supply; then the absolute value of the temperature change rate of each power supply is taken, and the temperature change rates of all power supplies are averaged to obtain the power supply temperature change gradient ΔTg; The acquisition method of the load fluctuation rate Δa is: first, the acceleration data in a period of time is continuously acquired from the IMU, and the average value of all acceleration data in this period of time is calculated; the acceleration value of each sampling point in this period of time is compared with the average acceleration obtained just now, and the difference between them is calculated; then the difference is squared to obtain the average value; finally, the average value is squared to obtain the load fluctuation rate Δa; The acquisition method of the power change rate ΔPr is: first, the difference between the driving power Pw(t) at time t and the driving power Pw(t-Δt) at time t-Δt is calculated; Then divide the difference by the driving power Pw(t-Δt) at time t-Δt to obtain the relative change ratio, and finally take the absolute value of the ratio to obtain the power change rate ΔPr; The feature normalization and reconstruction unit normalizes and standardizes the obtained power supply temperature change gradient ΔTg, load fluctuation rate Δa and power change rate ΔPr, and fits them into a dynamic feature set Fd.

6. The multi-source data based forklift mobile robot intelligent power supply system according to claim 5, characterized in that: The power consistency evaluation unit statistically evaluates the state mapping coefficients Φ of all power supplies, calculates the power consistency deviation index ΔCvar, and quantifies the internal balance of the entire power supply group; The acquisition method of the power consistency deviation index ΔCvar is: first, the state mapping coefficient of each power supply at time t is extracted, and the average value of the state mapping coefficients of all power supplies at time t is calculated; then the deviation between the state mapping coefficient of the power supply and the average value is calculated, and the deviation value is squared; the sum of the squared deviation values of all power supplies is averaged; finally, the square root is taken to obtain the power consistency deviation index ΔCvar; The obtained power consistency deviation index ΔCvar is compared with the preset deviation index ratio Tcvar to determine the health status of the power supply; The value range of the deviation index ratio Tcvar is [0.01, 0.5]; The health status of the power supply is matched and obtained in the following way: When the power consistency deviation index ΔCvar is less than or equal to the deviation index ratio Tcvar, it indicates that the power supply state is uniform and in a healthy state; When the power consistency deviation index ΔCvar is greater than the deviation index ratio Tcvar, it indicates that the power supply is abnormal and the behavior is out of control, triggering active balancing adjustment to adjust the state of the power supply.

7. The multi-source data based, forked mobile robot intelligent power supply system according to claim 6, characterized in that: The power supply health and energy state joint calculation module includes a state function construction unit and a state index output unit; The state function construction unit combines the acquired power consistency deviation index ΔCvar and the dynamic characteristic set Fd to construct a power health state function Hs; The power health state function Hs is acquired through the following formula: ； In the formula, ΔCvar(t) represents the power consistency deviation index at time t, c1 represents a temperature gradient adjustment coefficient, and c2 represents a load fluctuation adjustment coefficient; The state index output unit further combines the current power output and the driving power demand to construct an energy margin function Er; The energy margin function Er is acquired through the following formula: ； In the formula, Ve(t) represents the voltage at time t, Ie(t) represents the current at time t, p1 represents a balance adjustment factor, and EP represents a non-zero constant.

8. The multi-source data based, forked mobile robot intelligent power supply system according to claim 7, characterized in that: The fusion output and adaptive correction module includes a comprehensive state function construction unit and a perceived state judgment correction unit; The comprehensive state function construction unit constructs a power supply comprehensive state function Ψsys of a nonlinear coupling structure through the power health state function Hs and the energy margin function Er; The power supply comprehensive state function Ψsys is acquired through the following formula: ； In the formula, Ψsys represents the power supply comprehensive state function at time t; The perceived state judgment correction unit compares the acquired power supply comprehensive state function Ψsys and a state threshold Tsys in a fixed period and performs hierarchical judgment, and automatically triggers different operation suggestions and parameter correction mechanisms in different state intervals; The state threshold Tsys is acquired in the following manner: at time t, first, the comprehensive state function values in a historical time window are counted to calculate the average value μΨ of the power supply comprehensive state function; at the same time, the standard deviation of the power supply comprehensive state function is calculated; then, the state threshold at the current moment is obtained by subtracting the product of the standard deviation and a sensitivity coefficient from the average value; The state hierarchical judgment manner is as follows: When the power supply comprehensive state function Ψsys is greater than the average value μΨ of the power supply comprehensive state function, it represents a stable operation state; When the state threshold Tsys is less than the power supply comprehensive state function Ψsys and is less than or equal to the average value μΨ of the power supply comprehensive state function, it represents a mild attention state; When the power supply comprehensive state function Ψsys is less than or equal to the state threshold Tsys, it represents an abnormal risk state.

9. The intelligent power supply method for the forked mobile robot based on multi-source data, applied to the intelligent power supply system for the forked mobile robot based on multi-source data according to any one of claims 1-8, characterized in that: The method comprises the following steps: Step one, the power information acquisition module acquires and pre-processes power data through an AFE chip and a fork mobile robot operation characteristic sensor to obtain a power data set DW; Step two, the power supply timing synchronization and noise suppression module performs time synchronization and multi-scale filtering on the power data set DW to obtain a suppressed data set SDW; Step three, the dynamic characteristic extraction and power parameter reconstruction module extracts features from the suppressed data set SDW to extract dynamic characteristics reflecting power state changes and fit them into a dynamic characteristic set Fd; Step four, the fusion state mapping and power consistency evaluation module establishes a dynamic state mapping matrix based on the dynamic characteristic set Fd, constructs a state mapping coefficient Φ, and performs consistency evaluation to acquire a power consistency deviation index ΔCvar; Step five, the power supply health and energy state joint calculation module combines the acquired power consistency deviation index ΔCvar and the dynamic feature set Fd to calculate the power supply health state function Hs and the energy margin function Er; Step six, the fusion output and adaptive correction module combines the acquired power supply health state function Hs and the energy margin function Er to obtain the power supply comprehensive state function Ψsys, and judges the perception state of the power supply.

Images