Agricultural machinery working condition prediction and adaptive control method based on multi-sensor fusion

Abstract

The application provides a kind of agricultural machinery working condition prediction and adaptive control method based on multi-sensor fusion, it is related to agricultural machinery working condition prediction technical field, comprising: constructing working condition data stack, according to time sequence buffer multi-source sensor collection mechanical operation data and surrounding environment data;After the working condition data is preprocessed, a standardized feature matrix is obtained, which is input into the working condition prediction model, and a predicted working condition parameter set is output;Calculate the prediction working condition adaptation degree coefficient to represent the matching degree of the predicted working condition and the ideal working state;The predicted working condition parameter set and the adaptation degree coefficient are input into the multi-objective optimization controller, combined with the operation safety requirement, the adaptive control parameter set is iteratively solved by dynamic inertia weight particle swarm optimization algorithm;The adaptive control parameter set is issued to the agricultural machinery actuator, the target travel speed, driving power and hydraulic system pressure are cooperatively regulated to optimize the mechanical operation state and form a closed-loop feedback control.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of agricultural machinery working condition prediction, in particular to an agricultural machinery working condition prediction and adaptive control method based on multi-sensor fusion. BACKGROUND

[0002] Currently, agricultural machinery is increasingly widely used in modern agricultural production, but its operating environment is complex and variable, such as terrain slope, soil hardness, environmental temperature and humidity, and other external factors have a significant impact on the operating state of the machinery. At the same time, the internal operating state of the agricultural machinery, such as travel speed, hydraulic system pressure, drive motor current, etc., will also fluctuate due to changes in the work load. However, the control method of traditional agricultural machinery is usually based on fixed parameter design or simple feedback control logic, lacking comprehensive perception and analysis ability for multiple complex factors, which is easy to cause the machinery to run in unsuitable working conditions, and thus cause low work efficiency, excessive energy consumption, etc. In addition, these control methods have slow response speed to the operating state of the machinery, and it is difficult to adjust the operating parameters of the machinery in time and accurately to adapt to the rapidly changing working environment. These deficiencies obviously cannot meet the requirements of modern agriculture for efficient and precise operation, and a more intelligent control system is needed to cope with complex working conditions.

[0003] At the same time, some improved agricultural machinery control methods currently exist, although they attempt to introduce sensor technology to collect operating data and environmental data to optimize the control strategy of the machinery, but mostly rely on a single sensor data source or simple data fusion, which is difficult to fully reflect the complexity of the operating state of the machinery and the working environment. In addition, these methods usually set operating parameters based on static models or empirical rules, lack dynamic prediction ability for working condition changes, and cannot predict future working conditions in advance, so it is difficult to correct the control strategy in time, resulting in unstable operating state of the machinery and large fluctuations in work efficiency. Therefore, efficient fusion and real-time processing of multi-source sensor data, combined with advanced prediction models and optimization algorithms to realize intelligent decision-making and dynamic control of agricultural machinery, is a key problem that needs to be broken through in the current technical field.

[0004] The above information disclosed in the background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide an agricultural machinery working condition prediction and adaptive control method based on multi-sensor fusion to solve the problems raised in the background.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0007] The application discloses a multi-sensor fusion-based working condition prediction and self-adaptive control method for agricultural machinery, and specifically comprises the following steps:

[0008] Step 1: constructing a working condition data stack for caching original data collected by multi-source sensors in time sequence, wherein the original data comprises mechanical operation data and surrounding environment data;

[0009] Step 2: pre-processing multi-source heterogeneous data in the working condition data stack, sequentially comprising timestamp alignment, spatial position mapping and standardization processing, and generating a standardized feature matrix for representing a comprehensive operation state of the machinery;

[0010] Step 3: inputting the standardized feature matrix into a working condition prediction model, wherein the model is constructed based on a long short-term memory network and is used for outputting a predicted working condition parameter set of a future control period, and the predicted working condition parameter set comprises a predicted driving motor current, a predicted hydraulic system pressure and a predicted travel resistance;

[0011] Step 4: determining a predicted working condition adaptation degree coefficient based on the predicted working condition parameter set, wherein the predicted working condition adaptation degree coefficient is used for representing a matching degree of the predicted working condition and an ideal working state of the agricultural machinery;

[0012] Step 5: inputting the predicted working condition parameter set and the predicted working condition adaptation degree coefficient into a multi-objective optimization controller, wherein the multi-objective optimization controller takes a preset mechanical operation safety requirement as a constraint, takes maximizing the predicted working condition adaptation degree coefficient as an optimization target, and iteratively solves a self-adaptive control parameter set through an optimization algorithm;

[0013] Step 6: issuing the self-adaptive control parameter set to underlying actuators, and adjusting the mechanical operation state through the actuators, wherein the self-adaptive control parameter set comprises a target travel speed, a target driving power and a target hydraulic system pressure.

[0014] Further, the mechanical operation data comprises a real-time travel speed of the machinery, a working current of a driving motor, an operation pressure of a hydraulic system and a load borne by a machine body, and are collected according to a speed sensor installed on a travel system of the machinery, a current sensor at an output end of the driving motor, a pressure sensor of a pipeline of the hydraulic system and a load sensor at a middle part of the machine body respectively;

[0015] The surrounding environment data comprises a terrain slope at a front end of the machinery, a soil surface hardness, an environment temperature and an environment humidity, and are collected according to a terrain slope sensor installed at the front end of the machinery, a soil surface hardness sensor and an environment temperature and humidity sensor at a top part of the machine body respectively;

[0016] A data buffer based on a ring queue structure is defined as a working condition data stack, and the raw data collected above is received in parallel; before each frame of data is stored in the working condition data stack, a high-precision timestamp and corresponding spatial coordinate information uniformly time-stamped by the system are added, so that the multi-source heterogeneous raw data is organized and buffered according to time sequence and spatial position; the raw data includes mechanical operation data and surrounding environment data.

[0017] Further, the raw data in the working condition data stack is preprocessed, first, the timestamp alignment is performed, taking the time-stamped timestamp uniformly time-stamped by the system as the reference, the time difference data generated by the sampling delay of different sensors is linearly interpolated and corrected, to ensure that a complete set of mechanical-environmental data corresponds to the same time node; then, the spatial position mapping is performed, combining the spatial coordinate information associated with each frame of data, the terrain slope, soil surface hardness, environmental temperature and humidity are associated and bound with the mechanical operation data under the corresponding coordinates, to form the corresponding relationship of spatial coordinate-mechanical data-environmental data; finally, the standardization processing is performed to generate a standardized feature matrix for representing the comprehensive running state of the machine;

[0018] Wherein, the standardization processing adopts the range standardization method, and the collected mechanical operation data and surrounding environment data are converted to interval:

[0019] ;

[0020] Wherein is the parameter original value, is the parameter standardized value, , The minimum and maximum values of the parameter in the working condition data stack are respectively.

[0021] Further, the specific execution process of step 3 is as follows:

[0022] The standardized feature matrix is divided into an input sequence according to time sequence, the time window length of the initial input sequence is set to 50 sampling periods, the input sequence is input into a working condition prediction model based on a long short-term memory network, the input layer dimension of the working condition prediction model is consistent with the dimension of the standardized feature matrix, the hidden layer is set to 2 layers, each layer has 32 neurons, the output layer has 3 neurons, respectively corresponding to the predicted driving motor current, the predicted hydraulic system pressure and the predicted travel resistance, the model learns the time sequence dependence of each parameter in the historical standardized feature matrix, and predicts the working condition parameters of the agricultural machine in the future one control period;

[0023] After the prediction is completed, the rationality of the model output result is verified: if the predicted value exceeds the preset normal range, the time window length of the prediction model input sequence is returned to 120% of the original time window length; the prediction is re-predicted, and if the predicted value is within the normal range, a prediction working condition parameter set containing the predicted driving motor current, the predicted hydraulic system pressure and the predicted travel resistance is output.

[0024] Further, based on the prediction working condition parameter set, a prediction working condition adaptation degree coefficient is determined, and the formula is as follows:

[0025] ;

[0026] In the formula, is the prediction working condition adaptation degree coefficient, which is used to represent the matching degree of the prediction working condition and the ideal working state of the agricultural machine; is the predicted driving motor current, is the ideal value of the driving motor current; is the predicted hydraulic system pressure, is the rated working pressure of the hydraulic system; is the predicted travel resistance, is the reference travel resistance, which is a typical resistance value measured under standard operating conditions; is a natural constant; , and are preset weights, , and satisfy .

[0027] Further, the prediction working condition parameter set and the prediction working condition adaptation degree coefficient are spatio-temporally associated and bound, that is, a mapping relationship is established between the prediction working condition parameter set and the prediction working condition adaptation degree coefficient according to the time stamp and the spatial coordinate information; the preset mechanical operation safety requirements are synchronously called from the agricultural machine control system as the constraint boundary, and the maximum prediction working condition adaptation degree coefficient is taken as the optimization objective, the target travel speed, the target driving power and the target hydraulic system pressure are set as the to-be-optimized variables, and the particle swarm optimization algorithm with dynamic inertia weight is used for iterative solution. Specifically, 50 particles are initialized, each particle corresponds to a combination of to-be-optimized variables, the maximum number of iterations is set to 30, the prediction working condition adaptation degree coefficient corresponding to each particle is calculated and the individual optimal and group optimal are updated each time, and when the iteration reaches 15 times without updating the group optimal, the adaptive control parameter set is output; if not, the particle swarm size is adjusted to 60 for re-iteration until the required parameter set is obtained, that is, the adaptive control parameter set; the adaptive control parameter set includes the target travel speed, the target driving power and the target hydraulic system pressure; the dynamic inertia weight refers to that in the iteration process, the inertia weight linearly decreases from the initial value 0.9 to the final value 0.4 with the number of iterations.

[0028] Further, the specific implementation process of step 6 is as follows:

[0029] The adaptive control parameter set is disassembled according to the actuator type into corresponding control instructions, wherein the target travel speed instruction is issued to the travel controller of the agricultural machinery walking system, the actual travel speed of the machinery is converged to the target value by adjusting the walking motor speed; the target driving power instruction is issued to the driving motor controller, the motor output current threshold is calculated in combination with the rated voltage and power factor of the driving motor, and the power is accurately output through current closed-loop control; the target hydraulic system pressure instruction is issued to the hydraulic controller, the hydraulic system pressure is maintained stable by adjusting the hydraulic pump displacement and the overflow valve opening; and the adjusted actual running parameters are fed back to the working condition data stack through the sensor collection, forming a closed-loop control, so as to complete the working condition prediction and adaptive control of the agricultural machinery.

[0030] Compared with the prior art, the beneficial effects of the present application are:

[0031] The present application fuses the mechanical running data and surrounding environment data collected by multiple source sensors to construct a working condition prediction and adaptive control method, which can realize real-time perception of the comprehensive running state of the agricultural machinery, and accurately predict the future working condition parameters based on the long short-term memory network. Compared with the traditional mechanical control method, the present application can dynamically adjust the control parameters of the machinery to ensure its efficient operation in the complex working environment. Specifically, by calculating the predicted working condition adaptation coefficient, the particle swarm optimization algorithm with dynamic inertia weight is combined to quickly iteratively solve the adaptive control parameter set, realizing the precise regulation and control of the target travel speed, driving power and hydraulic system pressure. At the same time, the method constructs a closed-loop control mechanism, which feeds back the actual running data to the control system in real time, continuously optimizes the working state of the machinery, thereby effectively reducing the working energy consumption and improving the working efficiency and running reliability of the agricultural machinery. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 It is a schematic diagram of the overall method of the present application;

[0033] Figure 2 It is a parallel coordinate graph of the current deviation term, the pressure deviation term and the resistance deviation term.

[0034] Figure 3 It is a bar graph of the current deviation term, the pressure deviation term and the predicted working condition adaptation coefficient. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the present application is further described in detail below in combination with specific embodiments.

[0036] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly. Example

[0037] Please see Figure 1 The present invention provides a technical solution:

[0038] The method for predicting and adaptively controlling the operating conditions of agricultural machinery based on multi-sensor fusion includes the following steps:

[0039] Step 1: Construct a working condition data stack to cache raw data collected in real time by multiple sources of sensors in a time series manner. The raw data includes mechanical operation data and surrounding environment data.

[0040] In this embodiment, the mechanical operation data includes the real-time mechanical speed, the operating current of the drive motor, the operating pressure of the hydraulic system, and the load on the machine body, which are collected by the speed sensor installed on the mechanical walking system, the current sensor at the output end of the drive motor, the pressure sensor in the hydraulic system pipeline, and the load sensor in the middle of the machine body, respectively.

[0041] The surrounding environmental data includes the terrain slope, soil surface hardness, ambient temperature and ambient humidity at the front end of the machine, which are collected by the terrain slope sensor, soil surface hardness sensor and ambient temperature and humidity sensor installed at the front end of the machine respectively.

[0042] The core reason for collecting the above mechanical operation data and surrounding environment data in this embodiment is that the working condition of agricultural machinery is significantly affected by the coupling of "mechanical state-external environment", and multi-dimensional parameter collection is needed to achieve comprehensive perception and accurate representation of the working condition: the real-time driving speed of the machine directly reflects the working efficiency and the load state of the walking system, and its fluctuation will lead to uneven working distance or overload of the driving system, which needs to be monitored in real time to avoid efficiency loss or mechanical loss; the driving motor working current is a direct reflection of the motor load, and an abnormal increase in current means that the motor may be overloaded due to increased soil resistance, component jamming, etc., and real-time collection can provide early warning of motor failure; the hydraulic system operating pressure is related to the stability of key actions such as header lifting and turning, and excessive pressure can easily lead to hydraulic pipeline leakage, while low pressure can cause slow action response, so pressure data is needed to ensure the safe operation of the hydraulic system; the machine body load is related to the overall stress balance of the machine, and a sudden increase in load can cause the machine body to tilt due to hard soil or crop congestion, affecting the working accuracy. In the surrounding environment data, the terrain slope directly affects the mechanical driving resistance and working posture, and excessive slope can easily cause the machine to slide down the slope or overload on one side; the hardness of the soil surface determines the traction required by the driving system and the depth of the working components into the soil, and differences in hardness can cause significant fluctuations in the mechanical load; environmental temperature and humidity affect sensor accuracy, motor cooling efficiency, and hydraulic oil viscosity, and high temperature can reduce motor insulation performance, while high humidity can easily cause sensor data drift - in summary, collecting the above parameters can comprehensively cover the key influencing factors of "mechanical operation-environmental interference", providing complete and reliable raw data support for subsequent data fusion, working condition prediction and adaptive control, and avoiding misjudgment of working conditions or control errors due to missing parameters.

[0043] A data buffer area based on a ring queue structure is defined as a working condition data stack, and the above collected raw data is received in parallel; before each frame of data is stored in the working condition data stack, a high-precision timestamp and corresponding spatial coordinate information are added, which are uniformly time-stamped by the system, so that multi-source heterogeneous raw data is organized and cached according to time sequence and spatial position; the raw data includes mechanical operation data and surrounding environment data.

[0044] Step 1 organizes and caches the mechanical operation data and surrounding environment data collected by multi-source sensors in real time in the form of a ring queue by building a working condition data stack. This data structure not only efficiently manages multi-source heterogeneous data, but also neatly records data streams in time sequence, ensuring continuity and integrity. At the same time, each frame of data is labeled with high-precision timestamps and spatial coordinate information uniformly time-stamped by the system, enabling multi-dimensional association of time and space for different sensor data. This real-time caching and organization greatly improves the fusion efficiency of multi-source data, laying a good foundation for subsequent data preprocessing and model prediction.

[0045] Compared with the management mode of multi-source data in the prior art, such as simple linear cache or single data collection process, the ring queue structure and time-space marking method of step 1 effectively solve the problems of data discontinuity, incompleteness and different sensor data cannot be accurately synchronized. When dealing with complex working conditions, the traditional method often leads to inaccurate expression of working condition characteristics due to insufficient data integration capability. However, through accurate time stamp alignment and space coordinate mapping, the present application can ensure one-to-one correspondence between mechanical operation data and surrounding environment data, significantly improving the accuracy and timeliness of multi-sensor information fusion.

[0046] The implementation of step 1 provides a reliable data foundation for the overall scheme of the present application. Through accurate time and space marking, not only the smooth execution of subsequent data preprocessing process is ensured, but also a uniform and standardized input data format is provided for the working condition prediction model. In addition, the efficient cache capability of the working condition data stack enables the real-time changes of mechanical operation state and environment to be quickly captured, providing support for the accuracy of timely response and adaptive control of the subsequent prediction model. This step is an important prerequisite for realizing the function closed loop of the overall scheme.

[0047] Step 2: Preprocessing of multi-source heterogeneous data in the working condition data stack, including time stamp alignment, space position mapping and standardization processing in sequence, to generate a standardized feature matrix for representing the comprehensive operation state of the machine;

[0048] In this embodiment, the original data in the working condition data stack is preprocessed. First, time stamp alignment is performed, taking the time stamp of the system unified time as the reference, linearly interpolating and correcting the time difference data generated by different sensors due to sampling delay, to ensure that a complete set of machine-environment data corresponds to the same time node. Then, space position mapping is performed, combining the space coordinate information associated with each frame of data, to associate and bind the terrain slope, soil surface hardness, environment temperature and environment humidity with the mechanical operation data under the corresponding coordinates, forming a corresponding relationship of space coordinate-mechanical data-environment data. Finally, standardization processing is performed to generate a standardized feature matrix for representing the comprehensive operation state of the machine;

[0049] Among them, the standardization processing adopts the range standardization method, and the collected mechanical operation data and surrounding environment data are converted to interval according to the following formula:

[0050] ;

[0051] Among them, is the parameter original value, is the parameter standardized value, , are the minimum value and the maximum value of the parameter in the working condition data stack, respectively.

[0052] Step 2 addresses the issues of different sampling frequencies, sampling delays, and various data formats by preprocessing the multi-source heterogeneous data in the working condition data stack, including timestamp alignment, spatial position mapping, and standardization. Timestamp alignment ensures the synchronization of data collected by different sensors, spatial position mapping accurately associates mechanical operation data with surrounding environmental data based on space, and standardization eliminates the interference of different data dimensions on model input, making the input data more comparable and consistent. This clear and logical preprocessing process not only improves data quality, but also ensures that the training and reasoning of the subsequent working condition prediction model can be based on high-quality, standardized data input, achieving stability and accuracy of prediction.

[0053] Compared with the simple data preprocessing method in the prior art, the multi-step preprocessing scheme of Step 2 significantly improves the integrity and consistency of data. Traditional methods are often inefficient in dealing with sensor data delays or noise problems, and easily miss the spatio-temporal association between multi-source data, resulting in insufficient accuracy of model input data. The system unified time and spatial coordinate association technology adopted by the invention not only ensures the synchronization of data in time and space dimensions, but also realizes the efficient combination of mechanical operation data and environmental data, so that the model can more comprehensively perceive the comprehensive running state of agricultural machinery, thereby improving the overall scheme's ability to handle complex working conditions.

[0054] The implementation of Step 2 provides a high-quality input data basis for the overall scheme, which is an important link to improve the accuracy of the working condition prediction model. Through timestamp alignment and spatial position mapping, mechanical operation data and environmental data form a complete spatio-temporal association feature, providing key support for the working condition prediction model to identify complex time series relationships. At the same time, standardization effectively improves the training efficiency and prediction stability of the model, avoiding model bias caused by data dimension differences. These preprocessing steps provide standardized and unified high-quality input data for the construction and optimization of the subsequent long short-term memory network model, ensuring the reliability and efficiency of the entire working condition prediction and adaptive control method.

[0055] Step 3: input the standardized feature matrix into the working condition prediction model, which is based on a long short-term memory network and is used to output a predicted working condition parameter set for the next control period, including predicted drive motor current, predicted hydraulic system pressure, and predicted travel resistance.

[0056] In this embodiment, the specific execution process of Step 3 is as follows:

[0057] The standardized feature matrix is divided into an input sequence according to a time sequence, the length of a time window of an initial input sequence is set to 50 sampling periods, the input sequence is input into a working condition prediction model constructed based on a long short-term memory network, the input layer dimension of the working condition prediction model is consistent with the dimension of the standardized feature matrix, 2 layers of hidden layers are set, each layer has 32 neurons, 3 neurons are set in an output layer, and the 3 neurons correspond to a predicted driving motor current, a predicted hydraulic system pressure and a predicted travel resistance respectively, the model learns the time sequence dependency of each parameter in the historical standardized feature matrix, and working condition parameters of the agricultural machine in a future control period are predicted;

[0058] After the prediction is completed, the output result of the model is verified for rationality: if the predicted value exceeds a preset normal range, the length of the time window of the input sequence of the working condition prediction model is returned to 120% of the original time window length; the prediction is re-performed, and if the predicted value is within the normal range, a predicted working condition parameter set including the predicted driving motor current, the predicted hydraulic system pressure and the predicted travel resistance is output.

[0059] Step 3 uses the working condition prediction model based on the long short-term memory network to mine the time sequence dependency of the mechanical running state and the environmental working condition data by inputting the time sequence of the standardized feature matrix, so that the working condition parameters of the machine in a future control period can be predicted with high precision. The multi-layer hidden layer structure of the LSTM model and the time window division strategy enable it to capture complex historical data patterns and working condition change trends, and it is especially suitable for dynamic agricultural production environments. This prediction capability provides a reliable basis for intelligent decision-making and pre-adjustment of agricultural machines.

[0060] Compared with the simple linear regression or prediction method based on static rules in the prior art, the LSTM model has stronger nonlinear expression ability and time sequence modeling ability. The traditional method usually depends on fixed experience parameters or static mathematical models, and it is difficult to cope with the nonlinear changes and dynamic load characteristics of the working conditions of the agricultural machine, while the present application learns the time sequence dependency of the multi-source data through the LSTM, can accurately capture the law of the change of the working conditions of the machine with time, and significantly improves the accuracy and robustness of the prediction. In addition, through the rationality verification, the length of the time window is adjusted when the predicted value exceeds the normal range, which can further improve the stability and adaptability of the model prediction.

[0061] Step 3 is the core of the overall scheme, through high-precision working condition prediction, the future running state can be perceived in advance, and key parameter support is provided for adaptive control. Predicting working condition parameters such as motor current, hydraulic system pressure and travel resistance can directly provide accurate input basis for the subsequent steps, so that the system can adjust the operating parameters faster and more accurately to meet the requirements of complex working environment. In addition, this step also significantly improves the response speed of agricultural machinery to dynamic working conditions, providing a guarantee for efficient, energy-saving and stable mechanical operation.

[0062] Step 4: Based on the predicted working condition parameter set, determine the predicted working condition adaptation coefficient, which is used to represent the matching degree of the predicted working condition and the ideal working state of the agricultural machinery.

[0063] In this embodiment, based on the predicted working condition parameter set, the predicted working condition adaptation coefficient is determined according to the following formula:

[0064]

[0065] In the formula, is the predicted working condition adaptation coefficient, used to represent the matching degree of the predicted working condition and the ideal working state of the agricultural machinery. is the predicted driving motor current, is the ideal value of the driving motor current; is the predicted hydraulic system pressure, is the rated working pressure of the hydraulic system; is the predicted travel resistance, is the reference travel resistance, which is the typical resistance value measured under standard working conditions; is a natural constant; , and is the preset weight, , and satisfies .

[0066] The weight setting of is a realistic consideration based on the actual influence of each parameter on the working condition adaptation and the safety and efficiency of the mechanical operation during the agricultural machinery operation: the driving motor current is the core indicator of the matching between the mechanical power output and the load, and its deviation from the ideal value will directly lead to motor overload and burning, energy consumption surge or insufficient power, which seriously affects the safety and efficiency of the mechanical operation, and is the primary factor determining the working condition adaptation, so it is given the highest weight; the hydraulic system pressure is related to the stability of key execution actions such as header lifting and turning, and the pressure deviation from the rated value will lead to action delay or pipeline leakage, but compared with motor current anomaly, its influence on the core power system of the machinery and the overall working efficiency is secondary, so the weight is lower than ; the travel resistance will affect the mechanical travel speed and energy consumption, but its deviation from the baseline value can be compensated by adjusting the travel speed, driving power and other parameters, which has the minimum threat to the safety of mechanical core components and work efficiency, so it is given the lowest weight. This weight distribution not only conforms to the functional priority of each system of agricultural machinery, but also matches the degree of harm and controllability of parameter abnormalities in actual operation, ensuring that the predicted working condition adaptation coefficient can accurately reflect the dominant influence of key parameters on working condition adaptability.

[0067] In the above formula, the dependent variable is the predicted working condition adaptation coefficient, which quantitatively represents the matching degree of the predicted working condition of the agricultural machinery in the future control period and the ideal working state. The higher the value, the closer the predicted working condition to the ideal state of high efficiency, low consumption and safe operation of the machinery. The lower the value, the more significant the deviation of the working condition from the ideal state. From the technical effect, It provides a clear quantitative basis for the optimization target of the subsequent multi-objective optimization controller, avoiding the ambiguity of "working condition adaptability" relying on qualitative judgment, so that the controller can maximize the precision of the adaptive control parameter set, and thus ensure that the machinery always adjusts the operating state in the direction of "power matching, stable system pressure and controllable travel resistance" in complex working environments, improving the scientificity and operability of the control strategy.

[0068] The independent variables in the formula are the current deviation term, the pressure deviation term and the resistance deviation term. The correlation between the three and the dependent variable is derived from the core influencing mechanism of the working condition adaptability of agricultural machinery: the current deviation term directly reflects the deviation of the driving motor load from the ideal load, the pressure deviation term reflects the difference between the hydraulic system operating state and the rated state, and the resistance deviation term represents the deviation of the external resistance from the baseline resistance during the mechanical travel process. As key dimensions affecting the working condition adaptability of the machinery, the three jointly determine the value of the dependent variable, which is the core input variable for quantitative calculation of the working condition matching degree.

[0069] The dependent variable is negatively correlated with the three independent variables, i.e., as the current deviation term, pressure deviation or resistance deviation term increases, the dependent variable will decrease accordingly; on the contrary, when any independent variable decreases, the dependent variable will increase accordingly. This negative correlation is further reflected in the different function forms in the formula: the current deviation term has an impact through the exponential function: when the deviation increases, the exponential term rapidly decays, weakening , the pressure deviation term has an impact through the logarithmic function: when the deviation increases, the logarithmic term slowly rises, and the weakening effect on is relatively flat, and the resistance deviation term has an impact through the square function: when the deviation increases, the square term grows quadratically, and the weakening effect on The weakening effect lies between the two, ultimately forming a negative correlation change pattern that matches the influence weight of each parameter.

[0070] Table 1: Statistics of Predicted Working Condition Adaptability Coefficient

[0071]

[0072] It should be noted that in the above text and Table 1, the current deviation term, pressure deviation term, and resistance deviation term correspond to the following formulas: , , item.

[0073] The overall trend of the 15 sets of data shows a significant correlation between the predicted operating condition fit coefficient and the current deviation, pressure deviation, and resistance deviation terms: when the three deviation terms increase simultaneously, the fit coefficient generally decreases, indicating that as the deviation between the predicted operating condition and the ideal working state of agricultural machinery (ideal value of drive motor current, rated pressure of hydraulic system, and reference travel resistance) increases, the fit of the operating condition decreases accordingly. This trend is consistent with the calculation logic of the fit coefficient—the three deviation terms directly affect the coefficient result through exponential, logarithmic, and square terms; the larger the deviation, the more obvious the weakening effect on the coefficient, reflecting that when the mechanical operating parameters deviate from the ideal state, its comprehensive performance of "high efficiency and low consumption" will gradually decline.

[0074] Further observation of some data reveals that in scenarios where deviations change asynchronously, the change in the fit coefficient better reflects the differences in the influence weights of different deviations: when the current deviation is at a low level but the pressure or resistance deviation increases, the decrease in the fit coefficient is relatively gradual; however, when the current deviation increases significantly, even if the increases in the other two deviations are small, the fit coefficient will still show a significant decrease. This aligns with the formula... The weight settings are consistent, indicating that the deviation of the drive motor current from the ideal value has a more critical impact on the adaptability of the working conditions, followed by the hydraulic system pressure deviation, while the impact of the travel resistance deviation is relatively weak. This also provides data support for the subsequent multi-objective optimization controller to prioritize the adjustment of current-related control parameters, ensuring that the optimization strategy can specifically improve the deviation items that have the greatest impact on adaptability.

[0075] Step 4 quantifies the deviation degree of the actual working condition of the agricultural machine from the ideal working state by calculating the predicted working condition adaptation coefficient based on the predicted working condition parameter set. The calculation formula of the adaptation coefficient comprehensively considers the deviations of the three key parameters: drive motor current, hydraulic system pressure, and travel resistance. Through weighted evaluation, it can reflect the comprehensive performance of the machine running state in multiple dimensions. The introduction of exponential, logarithmic, and quadratic non-linear operators in the formula makes the sensitivity processing of different parameter deviations more flexible, allowing for prioritizing the optimization of more important working condition parameters and providing a more scientific and reasonable adaptation evaluation method for precise guidance of subsequent optimization control.

[0076] Compared with the evaluation method in the prior art that simply optimizes a certain working condition parameter as a single target, the adaptation coefficient calculation formula proposed by the present application has the core of multi-target index comprehensive evaluation and has significant advantages. The traditional method often ignores the synergistic effect between multiple working condition parameters, which may lead to deviations in the optimization results and cannot fully reflect the machine running state. The present application, by reasonably allocating weight parameters, enables the adaptation coefficient to consider the safety, energy efficiency, and stability of the machine running, ensuring that the optimization results match the actual application requirements. This multi-target comprehensive evaluation method significantly improves the adaptability and control effect of the agricultural machine in complex working conditions.

[0077] Step 4 is the key link to realize precise optimization control, and the predicted working condition adaptation coefficient provides a clear optimization target and scientific evaluation basis for the multi-objective optimization controller. Through quantitative evaluation of the deviation between the predicted working condition and the ideal state, the adaptation coefficient not only guides the optimization algorithm to dynamically adjust the optimization strategy, but also evaluates the pros and cons of the output results, providing a guarantee for the adjustment of actual control parameters. In the overall scheme, the accurate and efficient execution of Step 4 directly affects the precise adjustment ability of the optimization controller to the machine running state, thereby ensuring that the agricultural machine can run efficiently, energy-efficiently, and stably in complex working scenarios.

[0078] Step 5: input the predicted working condition parameter set and the predicted working condition adaptation coefficient into the multi-objective optimization controller, which takes the preset machine running safety requirement as a constraint and maximizes the predicted working condition adaptation coefficient as an optimization target, and iteratively solves the adaptive control parameter set through an optimization algorithm.

[0079] In this embodiment, the predicted working condition parameter set and the predicted working condition fitness coefficient are spatio-temporally associated and bound, that is, a mapping relationship is established between the predicted working condition parameter set and the predicted working condition fitness coefficient according to the time stamp and the spatial coordinate information; the preset mechanical operation safety requirement is synchronously called from the agricultural machinery control system as a constraint boundary, and then the target travel speed, the target driving power and the target hydraulic system pressure are set as to-be-optimized variables, and a dynamic inertia weight particle swarm optimization algorithm is used for iterative solving, specifically: 50 particles are initialized, each particle corresponds to a combination of to-be-optimized variables, the maximum number of iterations is set to 30, the predicted working condition fitness coefficient corresponding to each particle is calculated and the individual optimal and group optimal are updated each time, and when the iteration reaches 15 times without updating the group optimal, the adaptive control parameter set is output; if not, the particle swarm size is adjusted to 60 for re-iteration until the required parameter set is obtained, that is, the adaptive control parameter set; the adaptive control parameter set includes the target travel speed, the target driving power and the target hydraulic system pressure; the dynamic inertia weight refers to that in the iteration process, the inertia weight linearly decreases from the initial value 0.9 to the final value 0.4 with the number of iterations.

[0080] Step 5: The adaptive control parameter set is obtained by inputting the predicted working condition parameter set and the predicted working condition fitness coefficient into a multi-objective optimization controller, taking the agricultural machinery operation safety requirement as a constraint condition, taking the maximum working condition fitness coefficient as an optimization target, and using a dynamic inertia weight particle swarm optimization algorithm. This step uses the global search ability of the particle swarm algorithm and the local search enhancement characteristics of the dynamic inertia weight, which can efficiently converge in a short time and obtain the optimal control parameter combination that meets the complex working condition. At the same time, this optimization process can take into account the synergistic effect of multiple working condition parameters, achieving multi-dimensional and adaptive optimization of the agricultural machinery operation state.

[0081] Compared with the single-objective optimization or static rule optimization method in the prior art, the multi-objective optimization control strategy is adopted in step 5 of the present application, so that the optimization result can comprehensively consider the safety, stability and energy efficiency of the agricultural machinery working condition, and is not limited to the optimization of a single parameter. By introducing the dynamic inertia weight particle swarm optimization algorithm, the global search efficiency can be improved while the local search precision is enhanced, and the problem of falling into local optimum is avoided. In addition, the strategy of dynamically adjusting the particle swarm size further enhances the robustness of the algorithm, effectively solving the limitation that the traditional optimization algorithm is difficult to adapt to complex and dynamic working conditions.

[0082] Step 5 is the core step of realizing the adaptive control of the working condition of the agricultural machinery. Through the iterative solution of the multi-objective optimization controller, the adaptive control parameter set that can meet the operation demand is accurately output. The optimized control parameter set not only directly guides the adjustment of the running state of the agricultural machinery, but also provides a scientific basis for the subsequent control of the actuator, thereby improving the intelligent level and execution efficiency of the whole scheme. At the same time, through the optimization of the working condition adaptation degree coefficient, this step can continuously improve the matching degree of the mechanical running state and the ideal working condition, ensuring that the whole scheme realizes efficient, energy-saving and stable agricultural machinery operation effect, and significantly improves the adaptability and reliability of the system in the actual working environment.

[0083] Step 6: The adaptive control parameter set is sent to the bottom layer actuator, and each actuator adjusts the running state of the machinery cooperatively; the adaptive control parameter set includes the target travel speed, the target driving power and the target hydraulic system pressure.

[0084] In this embodiment, the adaptive control parameter set is disassembled into corresponding control instructions according to the type of the actuator. The target travel speed instruction is sent to the travel controller of the agricultural machinery walking system, the actual travel speed of the machinery is converged to the target value by adjusting the walking motor speed; the target driving power instruction is sent to the driving motor controller, the motor output current threshold is calculated by combining the rated voltage and power factor of the driving motor, and the power is accurately output through current closed-loop control; the target hydraulic system pressure instruction is sent to the hydraulic controller, the hydraulic system pressure is maintained stable by adjusting the hydraulic pump displacement and the overflow valve opening; and the adjusted actual running parameters are fed back to the working condition data stack through the sensor to form a closed-loop control, so as to complete the working condition prediction and adaptive control of the agricultural machinery.

[0085] Step 6 disassembles the adaptive control parameter set into specific control instructions according to the type of the actuator, and sends it to the bottom layer actuator of the agricultural machinery, which can accurately adjust the travel speed, driving power and hydraulic system pressure of the machinery, so as to realize the adaptive adjustment of the agricultural machinery to the complex working environment. In addition, this step adopts a closed-loop control mechanism, collects the adjusted actual running parameters in real time and feeds them back to the working condition data stack, dynamically corrects the control strategy, and ensures that the running state of the machinery always tends to the optimization target. This closed-loop control structure of decomposition-execution-feedback significantly improves the adaptability and response ability of the machinery to the dynamically changing working condition.

[0086] Unlike traditional agricultural machinery control techniques that use fixed control parameters or simple feedback control, the present application significantly enhances the intelligent level and precision of mechanical operation through dynamic adjustment of adaptive control parameter sets and closed-loop control. Traditional techniques often struggle to handle complex and variable working environments, and control strategies are prone to lag or inaccuracy, resulting in low mechanical efficiency or excessive energy consumption. The present application dynamically adjusts the travel speed, driving power and hydraulic system pressure, combined with multi-objective optimization results, to maximize work efficiency and reduce energy consumption while meeting work safety requirements. In addition, the closed-loop feedback mechanism ensures the real-time and reliability of the control strategy, solving the problems of feedback lag and lack of dynamic optimization in traditional techniques.

[0087] Step 6 is the key link in the overall scheme to convert the prediction and optimization results into actual mechanical operation adjustments. By issuing adaptive control parameter sets to the underlying execution mechanism and combining closed-loop feedback, seamless connection from working condition prediction to real-time control is achieved. This step ensures that agricultural machinery can operate according to the optimized target parameters, effectively dealing with complex and variable working environments, achieving efficient, energy-saving and stable operation. At the same time, the closed-loop control mechanism provides real-time and dynamic operation data for the working condition data stack, further improving the accuracy of data prediction and control, ultimately forming an efficient closed-loop adaptive control system, and promoting the leap of agricultural machinery from traditional passive control to intelligent and adaptive control.

[0088] The above formulas are dimensionless numerical calculations, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the most recent real situation. The preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0089] The above embodiments can be realized wholly or partially by software, hardware, firmware or any combination thereof. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software methods depends on the specific application and design constraints of the technical solutions.

[0090] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, which can be located in one place or distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the present embodiment.

[0091] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application.

Claims

1. A method for working condition prediction and adaptive control of agricultural machinery based on multi-sensor fusion, characterized in that, The specific steps include: Step 1: Constructing a working condition data stack for caching raw data collected by multi-source sensors in real time in time series, the raw data including mechanical operation data and surrounding environment data; Step 2: Preprocessing the multi-source heterogeneous data in the working condition data stack, including time stamp alignment, spatial position mapping and standardization processing in sequence, to generate a standardized feature matrix for representing the comprehensive operation state of the machine; Step 3: Inputting the standardized feature matrix into a working condition prediction model, which is constructed based on a long short-term memory network, for outputting a predicted working condition parameter set for a future control period, the predicted working condition parameter set including a predicted driving motor current, a predicted hydraulic system pressure and a predicted travel resistance; Step 4: Determining a predicted working condition adaptation degree coefficient based on the predicted working condition parameter set, the predicted working condition adaptation degree coefficient being used to represent the matching degree between the predicted working condition and the ideal working state of the agricultural machine; Step 5: Inputting the predicted working condition parameter set and the predicted working condition adaptation degree coefficient into a multi-objective optimization controller, the multi-objective optimization controller taking the preset mechanical operation safety requirement as a constraint and maximizing the predicted working condition adaptation degree coefficient as an optimization objective, and iteratively solving a set of adaptive control parameters by an optimization algorithm; Step 6: Distributing the set of adaptive control parameters to the underlying actuators, and adjusting the mechanical operation state by the actuators in coordination, the set of adaptive control parameters including a target travel speed, a target driving power and a target hydraulic system pressure; Based on the predicted working condition parameter set, a predicted working condition adaptation degree coefficient is determined according to the following formula: ； In the formula, is a predicted working condition adaptation coefficient, used to represent the matching degree between the predicted working condition and the ideal working state of the agricultural machine; is a predicted driving motor current, is an ideal value of the driving motor current; is a predicted hydraulic system pressure, is a rated working pressure of the hydraulic system; is a predicted travel resistance, is a reference travel resistance, which is a typical resistance value measured under standard operating conditions; is a natural constant; , and are preset weights, , and satisfy .

2. The multi-sensor fusion based working condition prediction and adaptive control method for agricultural machinery according to claim 1, characterized in that: The mechanical operation data includes real-time travel speed of the machine, working current of the driving motor, operating pressure of the hydraulic system and load carried by the machine body, which are collected by a speed sensor installed on the machine travel system, a current sensor at the output end of the driving motor, a pressure sensor on the hydraulic system pipeline and a load sensor in the middle of the machine body, respectively; The surrounding environment data includes terrain slope in front of the machine, soil surface hardness, ambient temperature and ambient humidity, which are collected by a terrain slope sensor installed in front of the machine, a soil surface hardness sensor and an ambient temperature and humidity sensor on the top of the machine body, respectively; A data buffer area based on a ring queue structure is defined as the working condition data stack, and the above collected raw data is received in parallel; Before each frame of data is stored in the working condition data stack, a high-precision time stamp uniformly granted by the system and corresponding spatial coordinate information are added, so that the multi-source heterogeneous raw data is organized and cached in time series and spatial position; the raw data includes mechanical operation data and surrounding environment data.

3. The multi-sensor fusion based working condition prediction and adaptive control method for agricultural machinery according to claim 2, characterized in that: The original data in the working condition data stack is preprocessed, first, time stamp alignment is performed, taking the time stamp of the system unified time service as the reference, the time difference data generated by different sensors due to sampling delay is corrected by linear interpolation, and it is ensured that a complete set of mechanical-environmental data corresponds to a time node; then, spatial position mapping is performed, combining the spatial coordinate information associated with each frame of data, the terrain slope, soil surface hardness, environmental temperature and environmental humidity are associated and bound with the mechanical operation data under the corresponding coordinates to form a corresponding relationship between spatial coordinates, mechanical data and environmental data; finally, standardization processing is performed to generate a standardized feature matrix for representing the comprehensive operation state of the machine; Among them, the standardization processing adopts the range standardization method, and the collected mechanical operation data and surrounding environment data are converted to Interval: ； wherein is the parameter raw value, is the parameter normalized value, , are the minimum and maximum values of the parameter in the operating data stack, respectively.

4. The multi-sensor fusion based working condition prediction and adaptive control method for agricultural machinery according to claim 1, characterized in that: The specific execution process of step 3 is as follows: The standardized feature matrix is divided into an input sequence according to the time sequence, the time window length of the initial input sequence is set to 50 sampling periods, the input sequence is input into the working condition prediction model based on the long short-term memory network, the input layer dimension of the working condition prediction model is consistent with the dimension of the standardized feature matrix, the hidden layer is set to 2 layers, each layer has 32 neurons, the output layer has 3 neurons, respectively corresponding to the predicted driving motor current, the predicted hydraulic system pressure and the predicted travel resistance, the model learns the time sequence dependence of each parameter in the historical standardized feature matrix to predict the working condition parameters of the agricultural machine in the future control period; After the prediction is completed, the rationality of the model output result is verified: if the predicted value exceeds the preset normal range, return the time window length of the working condition prediction model input sequence to 120% of the original time window length; re-predict, if the predicted value is within the normal range, output the predicted working condition parameter set including the predicted driving motor current, the predicted hydraulic system pressure and the predicted travel resistance.

5. The multi-sensor fusion based working condition prediction and adaptive control method for agricultural machinery according to claim 1, characterized in that: The predicted working condition parameter set and the predicted working condition adaptation coefficient are spatio-temporally associated and bound, that is, the mapping relationship between the predicted working condition parameter set and the predicted working condition adaptation coefficient is established according to the time stamp and the spatial coordinate information; Synchronously, the preset mechanical operation safety requirements are called from the agricultural machine control system as constraint boundaries, the maximum predicted working condition adaptation coefficient is taken as the optimization objective, the target travel speed, the target driving power and the target hydraulic system pressure are set as the to-be-optimized variables, and the particle swarm optimization algorithm with dynamic inertia weight is used for iterative solution, specifically: 50 particles are initialized, each particle corresponds to a combination of to-be-optimized variables, the maximum number of iterations is set to 30, the predicted working condition adaptation coefficient corresponding to each particle is calculated and the individual optimal and group optimal are updated each time, and when the iteration reaches 15 times without updating the group optimal, the adaptive control parameter set is output; if not, adjust the particle swarm size to 60 to re-iterate until the required parameter set is obtained, that is, the adaptive control parameter set; the adaptive control parameter set includes the target travel speed, the target driving power and the target hydraulic system pressure; the dynamic inertia weight refers to that in the iteration process, the inertia weight decreases linearly from the initial value 0.9 to the final value 0.4 with the number of iterations.

6. The multi-sensor fusion based working condition prediction and adaptive control method for agricultural machinery according to claim 1, characterized in that: The specific execution process of step 6 is as follows: The adaptive control parameter set is disassembled by actuator type into corresponding control instructions, wherein the target travel speed instruction is issued to the travel controller of the agricultural machinery walking system, the actual travel speed of the machinery is converged to the target value by adjusting the walking motor speed; The target driving power instruction is issued to the driving motor controller, the motor output current threshold is calculated in combination with the driving motor rated voltage and power factor, and the power is accurately output through current closed-loop control; the target hydraulic system pressure instruction is issued to the hydraulic controller, the hydraulic system pressure is maintained stable by adjusting the hydraulic pump displacement and the overflow valve opening; and the adjusted actual operating parameters are fed back to the working condition data stack through the sensor collection, forming a closed-loop control, so as to complete the working condition prediction and adaptive control of the agricultural machinery.

Images