Electric tractor whole machine heat-energy collaborative self-adaptive control method

Abstract

The application discloses a kind of electric tractor whole machine heat-energy collaborative self-adaptive control method, it is related to electric agricultural equipment control technical field, and the sensing equipment carried by electric tractor collects farmland working condition data and whole machine operating state data, after processing, storage, wherein, farmland working condition data and whole machine operating state data include farmland terrain fluctuation, soil humidity, operation type and whole machine running posture, two-dimensional position data.The application is by laser radar, soil humidity sensor, operation mode identifier, inertial measurement unit and global positioning module and other multi-source sensing equipment, comprehensively collects farmland terrain fluctuation, soil humidity, operation type, whole machine running posture and two-dimensional position data, constructs multidimensional perception system, all data is stored in history database after dimensionless processing, provides comprehensive input support for heat load prediction and intelligent decision-making, so that system can perceive potential influence of external working condition change on power system heat load in advance.

Description

Technical Field

[0001] This invention relates to the field of electric agricultural equipment control technology, specifically to a thermal-energy coordinated adaptive control method for an electric tractor. Background Technology

[0002] The power system of an electric tractor typically consists of multiple high-power components, including a power battery, a travel motor, a power take-off (PTO) motor, a hydraulic motor, and an electronic control module. Its operating status directly determines the overall machine's work efficiency, range, and service life. However, the farmland working environment is highly complex and time-varying, involving not only changes in work type, fluctuations in soil resistivity, and changes in terrain undulations, but also placing higher demands on the machine's stability and the accuracy of its work path tracking. When an electric tractor performs plowing, rotary tilling, sowing, or harvesting operations in the field, changes in the tractor's pitch and roll are significant. The positional offset within the two-dimensional plane often leads to severe coupled fluctuations in the thermal-energy load of various power components. If the thermal management system lacks the ability to perceive and respond to the above-mentioned multi-dimensional dynamic information, it is very easy to cause problems such as component temperatures exceeding the safety threshold, a sharp increase in system energy consumption, and waste of waste heat resources. This seriously restricts the reliability and continuity of operation of the whole machine in complex farmland scenarios. Therefore, it is urgent to develop an intelligent thermal management technology that can integrate farmland conditions, whole machine status, attitude changes, and position information to achieve coordinated optimization of heat flow and energy flow, so as to meet the application needs of modern agriculture for electric tractors that are efficient, intelligent, and highly adaptable.

[0003] Current thermal management technologies for electric tractors generally suffer from significant functional limitations. Traditional methods often employ fixed threshold control or simple proportional adjustment, failing to fully integrate multi-dimensional operating condition information such as farmland topography, soil properties, and operation type. This results in low accuracy in heat load prediction and difficulty in timely response to thermal shocks caused by attitude changes or load fluctuations. At the decision-making level, existing technologies often focus solely on temperature stability, neglecting system energy consumption optimization and the cascade recovery and utilization of waste heat resources. They also lack coupling consideration of the machine's two-dimensional position tracking accuracy and attitude stability, causing a disconnect between thermal management strategies and overall machine motion control. This further exacerbates energy waste and the risk of system instability. Furthermore, traditional adjustment methods... The control method lacks a closed-loop feedback mechanism, and the control parameters cannot be dynamically corrected according to the actual operating status. Fault handling mostly adopts a uniform response mode, which cannot be graded according to the degree of temperature anomaly, nor can it effectively combine attitude anomaly or path deviation information for comprehensive judgment. This leads to insufficient response or excessive load restriction, affecting operating efficiency and driving safety. In terms of waste heat utilization, the existing technology mostly adopts direct emission treatment for waste heat, and has failed to establish a graded recovery and targeted distribution mechanism, which further reduces the overall energy utilization level of the machine and makes it difficult to meet the actual needs of electric tractors for efficient and reliable operation under multi-dimensional coupled working conditions. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention is implemented through the following technical solution: a method for coordinated adaptive control of thermal and energy systems in an electric tractor, the specific steps of which are as follows: The S100 collects farmland working condition data and overall machine operation status data through the sensing equipment mounted on the electric tractor. After processing, the data is stored. The farmland working condition data and overall machine operation status data include farmland topography undulation, soil moisture, operation type, overall machine operation posture, and two-dimensional position data, providing comprehensive and accurate multi-dimensional data support for heat load prediction and intelligent decision-making. S200, based on the collected data, matches the differentiated component power heat loss coefficients, combines the working condition coupled heat load prediction algorithm with the tractor attitude angle and position changes, calculates the future heat load of each component, compares the heat load deviation and corrects the working condition coupling coefficient and the electronic control heat loss correction coefficient, and outputs the prediction result, so that the heat load prediction can dynamically adapt to the working condition changes, improve the prediction accuracy and response speed. The S300 employs three-dimensional target deep reinforcement learning and inner-layer fuzzy neural network, integrating attitude stability and path tracking deviation, and combining a four-dimensional decision matrix to generate control schemes for the cooling system power, heat flow distribution ratio, and waste heat utilization path of each heat source component of the whole machine. This achieves on-demand allocation of cooling resources and coordinated optimization of heat flow path, thereby improving the energy efficiency and operational stability of the whole machine. The S400 collects and classifies waste heat through a heat exchanger, then delivers it in a directional manner via an electromagnetic reversing valve. After recording the data and calculating the actual waste heat utilization rate, the data is uploaded, realizing the graded recycling and precise utilization of waste heat resources and improving the overall energy utilization efficiency of the machine. The S500 outputs control commands according to the control scheme, calculates the heat flow distribution correction coefficient through the waste heat feedback heat flow correction algorithm and feeds it back, while monitoring the component temperature and adjusting the strategy in real time in combination with attitude and position deviations, so as to execute the graded fault emergency mechanism, forming a closed-loop adaptive control system, enhancing the system's reliability and fault response capability under complex working conditions.

[0005] Preferably, S100 specifically includes: The system acquires farmland topographic relief data via a lidar sensor, collects soil moisture content data via a soil moisture sensor, and obtains operation type identification information via an operation mode recognizer. The operation types include four modes: tillage, sowing, harvesting, and transportation, ensuring that the thermal management strategy can match the differentiated needs of different operation modes. By collecting real-time operating power data of each motor through torque and speed sensors deployed on the walking motor, power output shaft motor and hydraulic motor, obtaining battery state of charge and charging / discharging current data through the battery management system, and obtaining the operating power of the electronic control system through the electronic control module monitoring unit, real-time status monitoring is provided for battery thermal management and energy scheduling, supporting accurate assessment and dynamic control of the thermal load of the electronic control unit. The tractor's overall operating attitude angle data, including pitch and roll angles, are collected by an inertial measurement unit. Two-dimensional position coordinate data of the tractor are obtained by a global positioning module. All collected data are stored in a historical database after being processed to be dimensionless. This provides a dynamic compensation basis for the impact of thermal load correction on driving stability, enabling thermal management to be calibrated in real time in conjunction with path tracking deviation.

[0006] Preferably, S200 specifically includes: The onboard lidar acquires the corresponding component power heat loss coefficient based on the collected component type identification information. Among them, motor components are uniformly matched with 0.85, batteries with 0.05, and electronic control modules with 0.1. The working condition coupling coefficient is matched based on the operation type identification information, with 0.9 for tillage, 0.7 for sowing, 1.0 for harvesting, and 0.5 for transportation. This enables heat load calculation to be differentiated and accurately matched according to component characteristics and operation mode. The future heat load of the component is calculated using a condition-coupled heat load prediction algorithm. The algorithm formula is as follows: ,in, For the future heat load of the components, The component's power heat loss coefficient, The operating condition coupling coefficient is... This is the correction factor for heat loss in electronic control systems. The real-time operating power of the component is expressed in kilowatts (kW). The soil moisture content is after dimensionless treatment. This represents the dimensionless variation of the terrain undulations. Input voltage to the electronic control module, in volts (V). The operating current of the electronic control module is expressed in amperes (A), enabling quantitative prediction and dynamic assessment of heat load under multiple coupled operating conditions. The working condition coupling coefficient is dynamically corrected by introducing real-time attitude angle and two-dimensional position change data of the tractor. When the pitch angle or roll angle exceeds the preset threshold, positive compensation is applied to the working condition coupling coefficient based on the angle deviation. When the path tracking deviation exceeds the preset range, the heat load prediction result is calibrated in real time based on the position deviation to ensure that the heat load prediction can dynamically adapt to the additional effects brought about by changes in driving attitude and path.

[0007] Preferably, S200 further includes: The predicted heat load is compared with the actual measured heat load, the relative deviation percentage is calculated, and the operating condition coupling coefficient and the electronic control heat loss correction coefficient are dynamically corrected using piecewise adaptive logic. When the deviation is ≤ ±1%, the original coefficient remains unchanged to ensure the long-term stability and reliability of the heat load prediction. When the deviation is 1% < ±2%, the coefficient correction range is ±0.02, which makes the prediction result converge quickly to the actual working condition. When the deviation is 2% < ±3%, the coefficient correction range is ±0.05. The correction direction is determined according to the sign of the deviation, which makes the heat load prediction result gradually approach the true value and enhances the system's dynamic adaptability to complex working conditions. When the absolute value of the deviation exceeds ±3%, in addition to the execution coefficient correction, the sensor fault self-test process is triggered simultaneously. The sensor fault self-test process includes the detection of the power-on status of the temperature sensor and the power sensor and the detection of the continuity of signal output. The detection results are uploaded to the control system to achieve a synergistic improvement in prediction accuracy and system reliability.

[0008] Preferably, S300 specifically includes: The system retrieves standard energy consumption, actual operating energy consumption, real-time temperature of each component, target temperature of each component, upper and lower limits of safe temperature of each component, and current waste heat utilization rate data as input parameters for the three-dimensional target deep reinforcement learning reward algorithm. Through multi-dimensional parameter fusion evaluation, it ensures that the reward value can fully reflect the comprehensive performance of the thermal management strategy. The reward value for the current state is calculated using a three-dimensional target deep reinforcement learning reward algorithm. The algorithm formula is as follows: ,in, As a reward value, To optimize the weighting coefficients for energy consumption, This is the temperature stability weighting coefficient. This is the weighting coefficient for waste heat utilization. The standard energy consumption of the system is expressed in joules (J). The actual energy consumption of the system is expressed in joules (J). The actual temperature of the component, in degrees Celsius. o C), The target temperature of the component, in degrees Celsius (°C). o C), The upper limit of the safe temperature for the component, in degrees Celsius (°C). o C), The lower limit of the safe temperature of the component, in degrees Celsius ( o C), To improve waste heat utilization, a reward-driven strategy is dynamically switched to achieve adaptive adjustment and optimization of thermal management objectives under different operating conditions. The real-time attitude angle deviation and path tracking deviation of the tractor are used as supplementary dimensions of the state space and input into the deep reinforcement learning module. When the attitude angle deviation or path tracking deviation exceeds the preset threshold, the reward algorithm introduces a penalty term to guide the policy network to output a coordinated action that takes into account both thermal management and driving stability. The introduction of attitude and path deviation penalty mechanism effectively improves the coordinated control capability of thermal management and driving stability under complex terrain.

[0009] Preferably, S300 further includes: The deep reinforcement learning module is constructed using the deep deterministic policy gradient algorithm, which includes two independent network structures: a policy network and a target network. The policy network is responsible for making action decisions that output energy consumption optimization coefficients, while the target network is responsible for calculating the target action value function. The dual network structure decouples decision-making and evaluation, significantly improving the stability of the learning process and the policy convergence efficiency. The experience replay pool of the deep reinforcement learning module is set to have a capacity of 10,000 records. The reward value calculated by the 3D target deep reinforcement learning reward algorithm, together with the corresponding state space data, action space data, next state data and termination flag, are stored in the replay pool as experience data. 64 experience data are randomly sampled each time training is conducted, and the policy network is trained using the gradient descent method. The training iteration step size is set to 0.001. The parameters of the policy network are synchronized every 200 training steps of the target network. Experience replay is used to break the correlation of data, ensure the independent distribution of training samples, and improve the model's generalization ability and learning efficiency. The strategy network dynamically adjusts the direction of the output action based on the reward value. When the reward value is higher than 0.5, it shifts towards reducing energy consumption, and when the reward value is lower than 0.3, it shifts towards ensuring stable temperature. The original decision value is normalized and converted into an energy consumption optimization coefficient in the range of 0 to 1. After being processed by a filter with a smoothing window of 5 decision cycles, it is output to the inner fuzzy neural network. The optimization target is adaptively switched based on the reward feedback to achieve a dynamic balance between energy consumption and temperature control. The filtering process ensures that the control command is output smoothly and avoids frequent actions of the actuator.

[0010] Preferably, in step S300, the process of generating control schemes for the cooling system power, heat flow distribution ratio, and waste heat utilization path of each heat source component of the whole machine by combining the four-dimensional decision matrix is ​​as follows: The input to the inner fuzzy neural network includes the energy consumption optimization coefficient output by the deep reinforcement learning module, the difference between the real-time temperature of each component and the target temperature, and the real-time attitude angle deviation and path tracking deviation of the tractor. The temperature difference is calculated according to the target range of 25~35℃ for the battery, 40~60℃ for the motor, and 30~50℃ for the electronic control, so that the thermal management decision can take into account both the temperature control requirements and the overall driving stability of the machine. The inner fuzzy neural network has 15 built-in fuzzy rules, which are divided into 5 temperature priority rules, 5 energy consumption priority rules and 5 waste heat utilization priority rules. After the input data is fuzzified, it is defuzzified by rule reasoning and the centroid method is used. The output is the 0~1 interval cooling priority coefficient of the walking motor, power output shaft motor and hydraulic motor, realizing intelligent allocation of cooling resources under multi-objective collaborative optimization. Based on the operation type, soil moisture content, and terrain undulation, the main blocks and sub-blocks of the four-dimensional decision matrix are matched. The nearest neighbor interpolation method is used to select the benchmark decision data. Combined with the cooling priority coefficient and energy consumption optimization coefficient, the control scheme is generated. The control scheme includes the power of each motor cooling water pump, the speed of the electric fan, the heat flow distribution ratio, and the specific utilization paths of high-temperature waste heat to battery preheating and medium- and low-temperature waste heat to the hydraulic system. This ensures that the control commands are accurately adapted to the real-time working conditions and achieves synergistic optimization of cooling and waste heat utilization.

[0011] Preferably, S400 specifically includes: Waste heat is collected by heat exchangers deployed at the outlets of various heat source components and classified according to the waste heat temperature level, which includes high temperature level ≥65℃, medium temperature level 45~55℃ and low temperature level 30~45℃, so as to realize the classification of waste heat resources by quality and lay the foundation for cascade utilization. The waste heat after classification is directed to the target utilization unit through electromagnetic reversing valve group. High-temperature waste heat is preferentially transported to the battery preheating circuit, medium-temperature waste heat is transported to the hydraulic oil tank insulation circuit, and low-temperature waste heat is transported to the cab heating system or battery insulation circuit to ensure that waste heat is distributed as needed and improve the comprehensive energy utilization efficiency. The system records the waste heat collection volume of each heat exchanger, the opening status and duration of each electromagnetic reversing valve, and the temperature change data of each target utilization unit in real time. Based on the total waste heat collected and the actual waste heat utilized, the system calculates the actual waste heat utilization rate and uploads the data to the intelligent decision-making module to provide data support for heat flow distribution optimization and form a closed-loop feedback.

[0012] Preferably, S500 specifically includes: Based on the control scheme generated by the intelligent decision-making module, control commands are output to the actuators of the cooling water pump, electric fan and electromagnetic reversing valve. At the same time, the actual temperature and waste heat utilization rate of each component are collected in real time to ensure that the actuators respond accurately to the control commands and provide real-time data support for subsequent feedback and correction. The waste heat feedback heat flow correction algorithm is used to calculate the correction coefficient for the next round of heat flow distribution. The algorithm formula is as follows: ,in, Correction factor for the next round of heat flow allocation. This is the current heat flow distribution coefficient. To provide feedback and correct the sensitivity coefficient, For actual waste heat utilization efficiency, To achieve the target waste heat utilization efficiency, dynamic matching between waste heat recovery effect and expected target is realized, and the adaptive capability of heat flow distribution strategy is improved. The calculated heat flow distribution correction coefficient is fed back to the heat flow distribution control module to dynamically adjust the cooling medium flow distribution ratio of each heat source component in the next decision cycle, so that the actual waste heat utilization rate gradually approaches the target waste heat utilization rate, forming a closed-loop adaptive control, continuously optimizing the waste heat recovery efficiency, and ensuring that the heat flow distribution is always in the optimal operating state.

[0013] Preferably, S500 further includes: The temperature of each component is monitored in real time and compared with the preset safe temperature threshold. At the same time, the tractor attitude angle deviation and path tracking deviation are monitored. The graded fault emergency mechanism is triggered according to the degree to which the temperature exceeds the safe threshold, thereby improving the overall machine's operational safety and fault response speed under complex working conditions. The graded fault emergency mechanism adopts a three-level response strategy. The first-level fault is when the temperature exceeds the safety threshold by 10%~20%, only the cooling priority of the corresponding component is increased. The second-level fault is when the temperature exceeds the safety threshold by 20%, the cooling priority is increased to the highest level and unnecessary waste heat transfer paths are cut off. The third-level fault is when the temperature exceeds the safety threshold by 30%, the load rate of the corresponding motor is simultaneously limited to ≤50%. This suppresses the development of overheating from the source and prevents the fault from further deteriorating and causing shutdown. When a fault is triggered, the system synchronously records the fault occurrence time, faulty component information, component temperature change curve in the 10 seconds before the fault, cooling system operating parameters, control command history, and attitude angle and position deviation data at the time of the fault. All fault data is stored in the database for subsequent analysis and system optimization, providing real data support for system optimization and continuously improving fault response capabilities.

[0014] This invention provides a method for coordinated adaptive control of thermal and energy performance of an electric tractor. It offers the following advantages: (I) The above-mentioned thermal-energy coordinated adaptive control method for electric tractors comprehensively collects farmland topography undulation, soil moisture, operation type, overall machine operating posture and two-dimensional position data through multi-source sensing devices such as lidar, soil moisture sensor, operation mode recognizer, inertial measurement unit and global positioning module. It constructs a multi-dimensional perception system covering environmental conditions, overall machine status and driving dynamics. All data are stored in historical database after dimensionless processing, providing accurate and comprehensive input support for heat load prediction and intelligent decision-making, enabling the system to perceive the potential impact of external operating condition changes on the heat load of the power system in advance.

[0015] (II) The whole-machine thermal-energy coordinated adaptive control method of the electric tractor is based on the acquisition data matching the differentiated component power heat loss coefficient, combined with the working condition coupled heat load prediction algorithm, to make a preliminary prediction of the future heat load of core components such as motor, battery, and electronic control. Furthermore, the real-time attitude angle and two-dimensional position change data of the tractor are introduced to dynamically compensate and calibrate the working condition coupling coefficient, so that the heat load prediction result can dynamically adapt to the changes in the actual driving trajectory and working area. Through the segmented adaptive deviation correction logic, the prediction accuracy is continuously optimized to ensure that the heat load prediction module has high accuracy and real-time response capability under long-term complex working conditions.

[0016] (III) The above-mentioned thermal-energy coordinated adaptive control method for electric tractors adopts a three-dimensional target deep reinforcement learning reward algorithm, which uses energy consumption optimization, temperature stability and waste heat utilization rate as comprehensive evaluation indicators to construct a multi-objective coordinated optimization mechanism. The deep reinforcement learning module makes online decisions based on the deep deterministic policy gradient algorithm, dynamically outputs the energy consumption optimization coefficient, and adaptively switches the control target according to the reward value, taking into account both energy efficiency improvement and temperature control protection. At the same time, the attitude angle deviation and path tracking deviation are introduced into the state space to guide the policy network to output coordinated actions that take into account thermal management and driving stability, thereby improving the overall operating performance of the machine in complex terrain. Attached Figure Description

[0017] Figure 1 This is a flowchart of a thermal-energy coordinated adaptive control method for an electric tractor according to the present invention; Figure 2 This is a schematic diagram illustrating the data transmission between each step of the thermal-energy coordinated adaptive control method for an electric tractor according to the present invention. Figure 3 This is a schematic diagram of the structure of the four-dimensional decision matrix in the intelligent decision-making process of the whole-machine thermal-energy coordinated adaptive control method of an electric tractor according to the present invention; Figure 4 This is a flowchart of the core algorithm for predicting the heat load of an electric tractor's overall thermal-energy coordinated adaptive control method, as described in this invention. Figure 5 This is a flowchart of the intelligent decision-making module of the whole-machine thermal-energy coordinated adaptive control method for electric tractors according to the present invention; Figure 6 This is a flowchart of the entire process of waste heat distribution in the whole-machine thermal-energy coordinated adaptive control method of an electric tractor according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1, please refer to Figures 1 to 6 This invention provides a technical solution: a method for coordinated adaptive control of thermal and energy performance of an electric tractor, the specific steps of which are as follows: The S100 uses sensors mounted on the electric tractor to collect farmland condition data and overall machine operating status data. This data is processed and stored. The farmland condition data and overall machine operating status data include farmland topography undulation, soil moisture, operation type, overall machine operating posture, and two-dimensional position data. This provides comprehensive and accurate multi-dimensional data support for heat load prediction and intelligent decision-making. The S100 uses an onboard LiDAR to acquire farmland topography undulation data, a soil moisture sensor to collect soil moisture content data, and an operation mode recognizer to obtain operation type identification information. Operation types include four modes: tillage, sowing, harvesting, and transportation. This ensures that the thermal management strategy can match the differentiated needs of different operation modes. This is achieved through sensors deployed on the drive motor and power take-off shaft. Torque and speed sensors of the electric motor and hydraulic motor collect real-time operating power data of each motor. The battery management system obtains battery state of charge and charging / discharging current data. The electronic control module monitoring unit obtains the operating power of the electronic control system, providing real-time status monitoring for battery thermal management and energy scheduling. This supports accurate assessment and dynamic control of the thermal load of the electronic control unit. The inertial measurement unit collects the tractor's overall operating attitude angle data, including pitch and roll angles. The global positioning module obtains the tractor's two-dimensional position coordinate data. All collected data is stored in a historical database after dimensionless processing, providing a dynamic compensation basis for the impact of thermal load correction on driving stability, enabling thermal management to be calibrated in real time in conjunction with path tracking deviation. It should be noted that before and during farmland operations, the electric tractor continuously scans the terrain ahead using its onboard LiDAR to acquire real-time data on terrain undulations. Simultaneously, a soil moisture sensor collects soil moisture information. The operation mode recognizer determines the current operation type—tillage, sowing, harvesting, or transportation—based on the driver's selection or automatic recognition logic. Farmland operating data, after dimensionless processing, is stored in a historical database as environmental input parameters. This allows the system to anticipate the potential impact of external operating conditions on the overall power system's thermal load based on the soil and terrain characteristics of different plots. During operation, torque and speed sensors deployed on the travel motor, power take-off shaft motor, and hydraulic motor collect real-time power data from each motor, accurately reflecting the actual load status of each power unit. The battery management system continuously monitors the battery's state of charge and... The charging and discharging current and the electronic control module monitoring unit synchronously acquire the operating power of the electronic control system. The overall machine operating status data is collected and dimensionless processed, and then stored in the historical database for real-time assessment of the current heating degree of each heat source component. The inertial measurement unit collects the pitch and roll angle data of the tractor in real time, and the global positioning module synchronously acquires the two-dimensional position coordinate information of the tractor. Together, they constitute the attitude and position perception system of the whole machine. The attitude angle data is used to judge the driving stability of the tractor in complex terrain, and the position coordinate data is used to track path deviation and locate the work area. All attitude and position data are also dimensionless processed and stored in the historical database to provide dynamic correction basis for heat load prediction and introduce driving stability constraints to the intelligent decision module, so that the thermal management strategy can take into account the attitude changes and path tracking accuracy under complex terrain, and improve the adaptive control capability of the whole machine in the variable farmland environment. The S200, based on the collected data, matches differentiated component power heat loss coefficients. Combined with the working condition coupled heat load prediction algorithm and the tractor's attitude angle and position changes, it calculates the future heat load of each component. After comparing the heat load deviation and correcting the working condition coupling coefficient and the electronic control heat loss correction coefficient, it outputs the prediction result. This allows the heat load prediction to dynamically adapt to changes in working conditions, improving prediction accuracy and response speed. The onboard LiDAR acquires the component type identification information and matches the corresponding component power heat loss coefficient. For motor components, it uniformly matches 0.85; for batteries, 0.05; and for electronic control modules, 0.1. The working condition coupling coefficient is matched based on the operation type identification information: 0.9 for tillage, 0.7 for sowing, 1.0 for harvesting, and 0.5 for transportation. This allows for differentiated and accurate matching of heat load calculation based on component characteristics and operating modes. The working condition coupled heat load prediction algorithm is used to calculate the future heat load of components. The algorithm formula is: ,in, For the future heat load of the components, The component's power heat loss coefficient, The operating condition coupling coefficient is... This is the correction factor for heat loss in electronic control systems. The real-time operating power of the component is expressed in kilowatts (kW). The soil moisture content is after dimensionless treatment. This represents the dimensionless variation of the terrain undulations. Input voltage to the electronic control module, in volts (V). The operating current of the electronic control module is measured in amperes (A). This enables quantitative prediction and dynamic assessment of heat load under multiple operating conditions. Real-time attitude angle and two-dimensional position change data of the tractor are introduced to dynamically correct the operating condition coupling coefficient. When the pitch angle or roll angle exceeds the preset threshold, positive compensation is applied to the operating condition coupling coefficient based on the angle deviation. When the path tracking deviation exceeds the preset range, the heat load prediction result is calibrated in real time based on the position deviation to ensure that the heat load prediction can dynamically adapt to the additional effects brought about by changes in driving attitude and path. It should be noted that during actual operation, based on the collected component type identification information, the corresponding power heat loss coefficient is automatically matched. For motor components such as the travel motor, power take-off shaft motor, and hydraulic motor, a uniform matching coefficient of 0.85 is used; the battery system matching coefficient is 0.05; and the electronic control module matching coefficient is 0.1. Simultaneously, based on the operation type identifier output by the operation mode recognizer, the working condition coupling coefficient table is called for matching: tillage operation matching coefficient 0.9, sowing operation matching 0.7, harvesting operation matching 1.0, and transportation operation matching 0.5. After completing the coefficient matching, the real-time operating power of each component, the dimensionless soil moisture content, and the terrain undulation amplitude data are read, and combined with the input voltage and operating current of the electronic control module, they are substituted into the working condition coupling heat load prediction algorithm to preliminarily calculate the future heat load value of each component. Based on the preliminary heat load prediction, the real-time attitude angle and two-dimensional position change data of the tractor are introduced to dynamically correct the working condition coupling coefficient. The inertial measurement unit collects the pitch angle and roll angle of the tractor in real time. When any angle exceeds the preset threshold, the current operation type is adjusted according to the angle deviation. The coupling coefficient corresponding to the type of operation is positively compensated to reflect the additional impact of complex terrain on the thermal load of the power system. At the same time, the global positioning module tracks the two-dimensional position coordinates of the tractor in real time. When the path tracking deviation exceeds the preset range, the thermal load prediction result is calibrated in real time according to the position deviation, so as to ensure that the thermal load prediction value can dynamically adapt to the changes in the actual driving trajectory and the working area, and improve the accuracy and adaptability of the prediction result. After dynamic correction of attitude angle and position deviation, the final future thermal load prediction result of each component is output as the core input parameter of the intelligent decision module. The prediction result not only reflects the static working condition characteristics such as operation type, soil moisture, and terrain undulation, but also integrates the dynamic attitude changes and path tracking deviation during the tractor's driving process, making the thermal load prediction more real-time and accurate. Based on this prediction result, combined with the three-dimensional target deep reinforcement learning algorithm and fuzzy neural network, a coordinated control scheme that takes into account energy consumption optimization, temperature stability and waste heat utilization is generated to achieve precise control of cooling power, heat flow distribution ratio and waste heat utilization path, and ensure the efficient and stable operation of the whole machine in complex farmland environment. Furthermore, the S200 also includes: comparing the predicted heat load value with the actual measured heat load value, calculating the relative deviation percentage, and using segmented adaptive logic to dynamically correct the operating condition coupling coefficient and the electronic control heat loss correction coefficient. When the deviation is ≤ ±1%, the original coefficient remains unchanged to ensure the long-term stability and reliability of the heat load prediction. When 1% < deviation ≤ ±2%, the coefficient correction range is ±0.02, so that the prediction result quickly converges to the actual operating condition. When 2% < deviation ≤ ±3%, the coefficient correction range is ±0.05. The correction direction is determined according to the positive or negative sign of the deviation, so that the heat load prediction result gradually approaches the true value, enhancing the system's dynamic adaptability to complex operating conditions. When the absolute value of the deviation exceeds ±3%, in addition to performing coefficient correction, the sensor fault self-check process is triggered simultaneously. The sensor fault self-check process includes the power-on status detection and signal output continuity detection of the temperature sensor and power sensor. The detection results are uploaded to the control system to achieve a synergistic improvement in prediction accuracy and system reliability. It should be noted that during the actual execution of heat load prediction, the predicted heat load values ​​for each component calculated by the algorithm are compared item by item with the real-time measured heat load values ​​collected by the temperature and power sensors deployed at the corresponding locations. The relative percentage deviation between the two is automatically calculated. This deviation value serves as the core indicator for measuring the accuracy of the prediction and is input into the segmented adaptive correction logic module. Based on the different ranges of the deviation, the module dynamically decides whether to adjust the operating condition coupling coefficient and the electronic control heat loss correction coefficient: when the deviation is within ±1%, the prediction result is considered to be highly consistent with the actual situation, and the original coefficients are maintained without change; when the deviation is between 1% and ±2%, the correlation coefficients are corrected by ±0.02; when the deviation is between 2% and ±3%, the correction range is increased to ±0.05, and the correction direction is strictly determined according to the sign of the deviation, ensuring that the heat load prediction result gradually approaches the true value during continuous operation; when the absolute value of the deviation exceeds ±3%, it is determined that the current prediction result deviates significantly from the actual operating conditions. In addition to immediately executing the above coefficient correction procedure, the sensor fault self-check process is triggered simultaneously. The self-test process checks the power-on status of the temperature and power sensors involved in the heat load measurement to confirm whether the sensor power supply is normal. It then checks the continuity of the sensor signal output to determine if there are any signal interruptions, jumps, or abnormal fluctuations. All test results are uploaded to the vehicle control system in real time, and prompts are generated on the human-machine interface, providing operators or maintenance personnel with clear fault location information. This mechanism ensures that when the heat load prediction deviation is large, potential problems at the sensor hardware level can be quickly identified, avoiding subsequent control decision errors due to abnormal data sources. Through a closed-loop mechanism combining segmented adaptive correction and fault self-testing, the accuracy of the heat load prediction model is continuously optimized during actual operation. On the one hand, a refined coefficient correction strategy based on the deviation range allows the operating condition coupling coefficient and the electronic control heat loss correction coefficient to be dynamically adjusted according to changes in the operating environment, gradually eliminating prediction errors. On the other hand, the timely intervention of the sensor fault self-testing process effectively prevents the risk of data distortion caused by hardware failure, significantly improving the operational reliability and adaptability of the heat load prediction module under long-term complex operating conditions. The S300 employs 3D target deep reinforcement learning and an inner-layer fuzzy neural network, fusing attitude stability and path tracking deviation. Combined with a four-dimensional decision matrix, it generates control schemes for the cooling system power, heat flow distribution ratio, and waste heat utilization paths of each heat source component. This achieves on-demand allocation of cooling resources and coordinated optimization of heat flow paths, improving overall energy efficiency and operational stability. It retrieves system standard energy consumption, actual system operating energy consumption, real-time temperature of each component, target temperature of each component, upper and lower limits of safe temperature for each component, and current waste heat utilization rate data as input parameters for the 3D target deep reinforcement learning reward algorithm. Through multi-dimensional parameter fusion evaluation, it ensures that the reward value comprehensively reflects the overall performance of the thermal management strategy. The 3D target deep reinforcement learning reward algorithm calculates the reward value in the current state. The algorithm formula is as follows: ,in, As a reward value, To optimize the weighting coefficients for energy consumption, This is the temperature stability weighting coefficient. This is the weighting coefficient for waste heat utilization. The standard energy consumption of the system is expressed in joules (J). The actual energy consumption of the system is expressed in joules (J). The actual temperature of the component, in degrees Celsius. o C), The target temperature of the component, in degrees Celsius (°C). o C), The upper limit of the safe temperature for the component, in degrees Celsius (°C). o C), The lower limit of the safe temperature of the component, in degrees Celsius ( o C), To improve waste heat utilization, a reward-driven strategy is dynamically switched to achieve adaptive adjustment and optimization of thermal management objectives under different operating conditions. The real-time attitude angle deviation and path tracking deviation of the tractor are used as supplementary dimensions of the state space and input into the deep reinforcement learning module. When the attitude angle deviation or path tracking deviation exceeds the preset threshold, the reward algorithm introduces a penalty term to guide the strategy network to output a coordinated action that takes into account both thermal management and driving stability. The introduction of attitude and path deviation penalty mechanism effectively improves the coordinated control capability of thermal management and driving stability under complex terrain. It should be noted that during the actual operation of the intelligent decision-making module, the system retrieves pre-calibrated standard energy consumption, real-time collected actual operating energy consumption, current temperature of each component, preset target temperature and safe temperature limits, and key parameters of waste heat utilization rate calculated in the current cycle from the historical database. This data is then uniformly input into the calculation unit of the three-dimensional target deep reinforcement learning reward algorithm. Based on the real-time status of each parameter, the algorithm comprehensively evaluates the overall performance of the current thermal management strategy in three dimensions: energy consumption optimization, temperature stability, and waste heat recovery, generating a reward value to guide the training of the deep reinforcement learning module. The deep reinforcement learning module is constructed using a deep deterministic policy gradient algorithm. The policy network dynamically outputs energy consumption optimization coefficients to guide the overall system thermal management based on the current state space input and reward value feedback. When the reward value is higher than a preset threshold, the policy network automatically shifts its output action towards reducing system energy consumption, prioritizing energy efficiency improvements through methods such as reducing cooling power. When the reward value is lower than a preset threshold, the algorithm shifts its output action towards reducing system energy consumption. When the threshold is reached, the strategy network adjusts its actions to shift towards ensuring stable component temperature, prioritizing enhanced cooling capacity to prevent component overheating. The original decision values ​​output by the strategy network are normalized and smoothed to generate stable energy consumption optimization coefficients, which are then passed to the inner fuzzy neural network for further decision fusion. To balance the overall machine's driving stability and thermal management performance in complex farmland environments, the deep reinforcement learning module uses the tractor's real-time attitude angle deviation and path tracking deviation as supplementary dimensions of the state space as input. When the inertial measurement unit detects that the pitch angle or roll angle exceeds the preset stability threshold, or the global positioning module detects that the path tracking deviation exceeds the allowable range, the reward algorithm automatically introduces a corresponding penalty term to reduce the reward value in the current state. This guides the strategy network to focus on thermal management objectives in subsequent decisions, while also considering driving stability constraints. The output coordinates thermal management requirements with overall machine attitude control and path tracking accuracy, ensuring optimal overall performance of the machine during complex terrain operations. Furthermore, the S300 also includes: a deep reinforcement learning module constructed using a deep deterministic policy gradient algorithm, comprising two independent network structures: a policy network and a target network. The policy network is responsible for outputting action decisions with energy consumption optimization coefficients, while the target network is responsible for calculating the target action value function. This dual-network structure decouples decision-making and evaluation, significantly improving the stability of the learning process and the policy convergence efficiency. The deep reinforcement learning module's experience replay pool capacity is set to 10,000 records. The reward value calculated by the 3D target deep reinforcement learning reward algorithm, along with the corresponding state space data, action space data, next state data, and termination flag, are combined to form experience data stored in the replay pool. During each training iteration, 64 experience data records are randomly sampled, and the policy network is trained using gradient descent. The step size is set to 0.001. The target network synchronizes the parameters of the policy network every 200 training steps. Experience replay is used to break the correlation of data, ensure the independent distribution of training samples, and improve the generalization ability and learning efficiency of the model. The policy network dynamically adjusts the output action direction according to the reward value. When the reward value is higher than 0.5, it shifts towards reducing energy consumption. When the reward value is lower than 0.3, it shifts towards ensuring temperature stability. The original decision value output is normalized and converted into an energy consumption optimization coefficient in the range of 0 to 1. After being processed by a filter with a smoothing window of 5 decision cycles, it is output to the inner fuzzy neural network. The optimization target is adaptively switched according to the reward feedback to achieve a dynamic balance between energy consumption and temperature control. The filtering process ensures that the control command is output smoothly and avoids frequent action of the actuator. It should be noted that during the actual operation of the intelligent decision-making module, the deep reinforcement learning module continuously performs online decision-making and optimization based on a dual-network architecture constructed using a deep deterministic policy gradient algorithm. The policy network, as the action output unit, generates the original decision value of the energy consumption optimization coefficient in real time based on the current state space input. The target network independently calculates the target action value function, providing a stable benchmark for the policy network's updates. An experience replay pool with a capacity of 10,000 experience data points is set up within the module. The state, action, reward, next state, and termination flag generated in each decision cycle are combined to form an experience tuple and stored in the replay pool. Once the replay pool has accumulated sufficient samples, 64 experience data points are randomly sampled and used to train the policy network using gradient descent. The iteration step size is set to 0.001 to ensure smooth update of network parameters. Simultaneously, the target network synchronizes the policy network parameters every 200 training steps to maintain the stability and convergence of the learning process. The deep reinforcement learning module dynamically adjusts the output direction of the policy network based on the reward value calculated by the three-dimensional target reward algorithm, achieving adaptive switching of the thermal management target. When the reward value is higher than 0.5, it indicates that the current thermal management... The strategy demonstrates excellent performance across three dimensions: energy consumption optimization, temperature stability, and waste heat utilization. The strategy network automatically shifts its output actions towards reducing system energy consumption, prioritizing improvements in overall energy efficiency through methods such as reducing cooling power. When the reward value falls below 0.3, indicating a risk of temperature runaway or a decline in overall performance, the strategy network immediately adjusts its actions to ensure component temperature stability, prioritizing enhanced cooling capacity to prevent component overheating. The original decision values ​​output by the strategy network must undergo normalization and smoothing filtering before being passed to the next-level decision module. Normalization maps the original output values ​​to the 0-1 range, forming standardized energy consumption optimization coefficients that are compatible with the input interface of subsequent fuzzy neural networks. The smoothing filter employs a sliding window mechanism with a window width of 5 decision cycles, performing mean filtering on the continuously output energy consumption optimization coefficients. This effectively suppresses drastic coefficient jumps caused by sensor noise or network output fluctuations. The filtered energy consumption optimization coefficients exhibit good temporal stability, smoothly guiding the inner fuzzy neural network to make cooling priority decisions and avoiding ineffective actions or oscillations in the field actuators caused by frequent fluctuations in control parameters. Furthermore, in the S300, the process of generating control schemes for the cooling system power, heat flow distribution ratio, and waste heat utilization path of each heat source component in the whole machine by combining a four-dimensional decision matrix is ​​as follows: The input of the inner fuzzy neural network includes the energy consumption optimization coefficient output by the deep reinforcement learning module, the difference between the real-time temperature and the target temperature of each component, and the real-time attitude angle deviation and path tracking deviation of the tractor. The temperature difference is calculated according to the target range of 25~35℃ for the battery, 40~60℃ for the motor, and 30~50℃ for the electronic control system, so that the thermal management decision can simultaneously take into account the temperature control requirements and the overall driving stability of the machine. The inner fuzzy neural network has 15 built-in fuzzy rules, divided into 5 temperature priority rules, 5 energy consumption priority rules, and 5 waste heat utilization priority rules. The input data is processed... After fuzzification, defuzzification is completed through rule-based reasoning and the centroid method. The cooling priority coefficients of the walking motor, power output shaft motor, and hydraulic motor in the 0~1 interval are output to realize intelligent allocation of cooling resources under multi-objective collaborative optimization. The main blocks and sub-blocks of the four-dimensional decision matrix are matched according to the operation type, soil moisture content and terrain undulation. The nearest neighbor interpolation method is used to select the benchmark decision data. The control scheme is generated by combining the cooling priority coefficient and energy consumption optimization coefficient. The control scheme includes the power of each motor cooling water pump, the speed of the electric fan, the heat flow distribution ratio and the specific utilization path of high temperature waste heat to battery preheating and medium and low temperature waste heat to the hydraulic system. This ensures that the control command is accurately adapted to the real-time working conditions and realizes the collaborative optimization of cooling and waste heat utilization. It should be noted that the inner fuzzy neural network receives the energy consumption optimization coefficients output by the deep reinforcement learning module, and simultaneously imports the differences between the real-time temperature and the target temperature of each component, as well as the real-time attitude angle deviation and path tracking deviation of the tractor. Based on the target temperature ranges of 25~35℃ for the battery, 40~60℃ for the motor, and 30~50℃ for the electronic control system, the temperature differences are calculated to ensure that the temperature control requirements of each heat source component are accurately quantified. The attitude angle deviation and path tracking deviation are input into the network as driving stability constraints, enabling the decision-making process to take into account the overall dynamic characteristics of the machine under complex terrain. After fuzzification processing, the input data... Mapped to the corresponding membership function; the network has 15 built-in fuzzy rules, divided into three categories of 5 rules each: temperature priority, energy consumption priority, and waste heat utilization priority, covering the main control objectives of the overall machine's thermal management. In the rule inference stage, the corresponding rules are activated according to the current input state, and the cooling requirements of each motor under the current operating conditions are comprehensively evaluated. The centroid method is used to complete the defuzzification process, and the inference results are converted into precise numerical outputs, generating cooling priority coefficients in the 0~1 interval for the travel motor, power output shaft motor, and hydraulic motor. These coefficients directly reflect the relative priority of each heat source component in obtaining cooling resources in the current cycle. The higher the value, the earlier the corresponding component receives cooling power. Based on the current operation type, soil moisture content, and terrain undulation, the system automatically matches the corresponding main blocks and sub-blocks in the four-dimensional decision matrix. Nearest neighbor interpolation is used to select suitable baseline decision data from the sub-blocks. The four-dimensional decision matrix is ​​divided into four main blocks according to operation type (cultivation, sowing, harvesting, transportation). Each main block is further divided into nine sub-blocks based on combinations of soil moisture content (10%~20%, 20%~30%, 30%~40%) and terrain undulation (0°~3°, 3°~6°, 6°~10°). Each sub-block stores 10 sets of optimal baseline decision data. Combining the cooling priority coefficient output by the inner fuzzy neural network and the energy consumption optimization coefficient output by the deep reinforcement learning module, the baseline decision data is dynamically adjusted to generate the final overall machine control scheme. This scheme specifically includes the power of each motor cooling water pump, the speed of the electric fan, the heat flow distribution ratio, and the directional utilization path of high-temperature waste heat to the battery preheating circuit and medium- and low-temperature waste heat to the hydraulic system. Through multi-level fusion and dynamic matching, it ensures that the control commands can accurately adapt to real-time operating condition changes and realize the on-demand allocation of cooling resources and efficient recovery of waste heat. The S400 system collects and grades waste heat through heat exchangers, then directs it via electromagnetic reversing valves. Data is recorded and the actual waste heat utilization rate is calculated and uploaded, enabling graded recovery and precise utilization of waste heat resources, thus improving the overall energy efficiency of the unit. Waste heat is collected by heat exchangers deployed at the outlets of each heat source component, and graded according to temperature levels: high temperature (≥65℃), medium temperature (45~55℃), and low temperature (30~45℃). This classification of waste heat resources by quality lays the foundation for tiered utilization. The graded waste heat is then directed to [the appropriate location / location] via electromagnetic reversing valves. In the target utilization unit, high-temperature waste heat is preferentially transported to the battery preheating circuit, medium-temperature waste heat is transported to the hydraulic oil tank insulation circuit, and low-temperature waste heat is transported to the cab heating system or battery insulation circuit to ensure that waste heat is distributed as needed, improve the overall energy utilization efficiency, and record the waste heat collection volume of each heat exchanger, the opening status and opening duration of each electromagnetic reversing valve, and the temperature change data of each target utilization unit in real time. Based on the total amount of waste heat collected and the actual waste heat utilized, the actual waste heat utilization rate is calculated and uploaded to the intelligent decision-making module to provide data support for heat flow distribution optimization and form a closed-loop feedback. It should be noted that during the actual operation of the electric tractor, the waste heat classification and targeted utilization are automatically executed through heat exchangers deployed at the outlets of each heat source component. The heat exchangers monitor the temperature of the cooling medium flowing through in real time and classify the waste heat online according to preset temperature thresholds. High-temperature waste heat (≥65℃), medium-temperature waste heat (45~55℃), and low-temperature waste heat (30~45℃) are marked and temporarily stored in corresponding buffer pipelines. The classification process is synchronized with the overall machine operation and does not rely on manual intervention, ensuring that waste heat resources can be identified and classified according to quality in a timely manner. At the same time, the temperature sensor built into the heat exchanger continuously uploads the classification data to the control system to support the dynamic optimization of the waste heat distribution strategy. The electromagnetic reversing valve group automatically switches the conduction state according to the control instructions issued by the intelligent decision module, and directionally delivers the classified waste heat to the corresponding target utilization unit. High-temperature waste heat is preferentially conducted to the battery preheating circuit to maintain the battery pack in low-temperature environments. The operating temperature is adjusted to improve discharge efficiency and service life; medium-temperature waste heat is introduced into the hydraulic oil tank insulation circuit to reduce hydraulic oil viscosity and reduce hydraulic system operating power consumption; low-temperature waste heat is selected to be delivered to the cab heating system or battery insulation circuit according to the cab temperature requirement or battery insulation requirement. The opening status and opening duration of the valve group are adjusted in real time by the control system to ensure that waste heat is distributed as needed and to avoid energy waste or heat flow conflict; during the waste heat distribution process, the control system simultaneously records the waste heat collection amount of each heat exchanger, the action parameters of each electromagnetic reversing valve, and the temperature change data of each target utilization unit. Based on the ratio of the total waste heat collected to the actual waste heat utilized, the actual waste heat utilization rate of the current cycle is calculated online and uploaded to the intelligent decision module as the input parameter for the next round of heat flow distribution correction. Through continuous data recording and feedback, the waste heat distribution module can adaptively adjust the delivery strategy so that the waste heat utilization rate gradually approaches the target value, realizing efficient recovery and precise utilization of waste heat resources; The S500 outputs control commands based on the control scheme, calculates the heat flow distribution correction coefficient using a waste heat feedback heat flow correction algorithm, and feeds it back. Simultaneously, it monitors component temperatures and adjusts strategies in real time based on attitude and position deviations to implement a graded fault emergency mechanism, forming a closed-loop adaptive control system. This enhances the system's reliability and fault response capabilities under complex operating conditions. Based on the control scheme generated by the intelligent decision module, it outputs control commands to the cooling water pump, electric fan, and electromagnetic reversing valve actuators. At the same time, it collects real-time data on the actual temperature and waste heat utilization rate of each component to ensure accurate response of the actuators to the control commands, providing real-time data support for subsequent feedback corrections. The waste heat feedback heat flow correction algorithm is used to calculate the heat flow distribution correction coefficient for the next round. The algorithm formula is as follows: ,in, Correction factor for the next round of heat flow allocation. This is the current heat flow distribution coefficient. To provide feedback and correct the sensitivity coefficient, For actual waste heat utilization efficiency, To achieve the target waste heat utilization efficiency, the system dynamically matches the waste heat recovery effect with the expected target, enhances the adaptive capability of the heat flow distribution strategy, feeds back the calculated heat flow distribution correction coefficient to the heat flow distribution control module, dynamically adjusts the cooling medium flow distribution ratio of each heat source component in the next decision cycle, so that the actual waste heat utilization rate gradually approaches the target waste heat utilization rate, forms a closed-loop adaptive control, continuously optimizes the waste heat recovery efficiency, and ensures that the heat flow distribution is always in the optimal operating state. It should be noted that during the closed-loop control phase of the entire machine, the control system generates specific execution commands based on the control scheme issued by the intelligent decision-making module and transmits them to the terminal actuators such as the cooling water pump, electric fan, and solenoid directional valve. The cooling water pump adjusts the medium flow rate of each motor's cooling circuit according to the commands, the electric fan adjusts its speed to enhance or weaken its heat dissipation capacity as needed, and the solenoid directional valve group switches its conduction state according to the heat flow distribution ratio, directionally delivering the cooling medium to the target heat source component. Simultaneously, temperature sensors and heat exchangers deployed at the outlets of each component continuously collect real-time temperature data and waste heat utilization data, and... The collected data is uploaded to the control system as input parameters for the next round of regulation. This execution and data acquisition process is repeated cyclically within each decision cycle to ensure that regulation commands can be applied to the physical system in a timely manner. It also provides real-time data support for subsequent feedback corrections, forming the basic execution layer for closed-loop control. After obtaining the actual waste heat utilization rate for the current cycle, the control system initiates a waste heat feedback heat flow correction algorithm to dynamically optimize the heat flow allocation strategy. The algorithm substitutes the current heat flow allocation coefficient, the actual waste heat utilization rate, and the target waste heat utilization rate into a preset correction formula to calculate the heat flow allocation correction coefficient for the next round. The calculation process of this correction coefficient fully considers the deviation between the actual waste heat recovery effect and the expected target. The deviation is reasonably scaled using a feedback correction sensitivity coefficient to ensure that the correction amplitude can quickly respond to changes in deviation while avoiding system oscillations caused by over-adjustment. The introduction of the correction coefficient gives the heat flow allocation strategy adaptive capability, enabling it to adjust the cooling medium flow distribution ratio of each heat source component in real time according to changes in the waste heat recovery effect, gradually narrowing the gap between the actual waste heat utilization rate and the target value. After the correction coefficient is calculated, the control system sends the new heat flow allocation coefficient to the heat flow allocation control module, which then re-plans the next... Within a decision cycle, the conduction sequence and conduction duration of each electromagnetic reversing valve are dynamically adjusted to optimize the flow distribution ratio of the cooling medium among each heat source component. As the operating conditions continue to change and the waste heat recovery effect is fed back in real time, this correction process is repeated in each decision cycle, so that the actual waste heat utilization rate always dynamically approaches the target value. Through the above closed-loop adaptive control mechanism, the whole machine thermal management system can continuously optimize the waste heat recovery efficiency and the rationality of heat flow distribution without relying on manual intervention, ensuring that waste heat resources are fully utilized in actual operation, while ensuring that the cooling needs of each heat source component are met. In addition, the S500 also includes: real-time monitoring of the temperature of each component and comparison with preset safe temperature thresholds, while monitoring the tractor's attitude angle deviation and path tracking deviation. Based on the degree to which the temperature exceeds the safe threshold, a graded fault emergency mechanism is triggered to improve the overall machine's operational safety and fault response speed under complex working conditions. The graded fault emergency mechanism adopts a three-level response strategy. Level 1 faults are when the temperature exceeds the safe threshold by 10%~20%, only the cooling priority of the corresponding component is increased. Level 2 faults are when the temperature exceeds the safe threshold by 20%, the cooling priority is increased to the highest level and unnecessary waste heat transport paths are cut off. Level 3 faults are when the temperature exceeds the safe threshold by 30%, the load rate of the corresponding motor is simultaneously limited to ≤50%. This suppresses overheating from the source and prevents the fault from further deteriorating and causing shutdown. When a fault is triggered, the fault occurrence time, fault component information, component temperature change curve in the 10 seconds before the fault, cooling system operating parameters, control command history, and attitude angle and position deviation data at the time of the fault are recorded simultaneously. All fault data are stored in the database for subsequent analysis and system optimization, providing real data support for system optimization and continuously improving fault response capabilities. It should be noted that during the actual operation of the electric tractor, the control system monitors the temperature data of each power component in real time and compares it item by item with the preset safe temperature threshold. Simultaneously, it collects pitch and roll angle data output by the inertial measurement unit and path tracking deviation information fed back by the global positioning module. Based on the different degrees to which the temperature exceeds the safe threshold, a graded fault emergency mechanism is automatically triggered to ensure rapid response and targeted measures in case of abnormal temperature, preventing further escalation of the fault. This monitoring process runs continuously in each control cycle, providing real-time data support for early fault identification and graded handling, ensuring the operational safety of the entire machine in complex farmland environments. The graded fault emergency mechanism adopts a three-level response strategy, implementing differentiated control measures according to the degree of temperature exceedance. When the component temperature exceeds the safe threshold by 10% to 20%, it is judged as a level one fault. The system only increases the cooling priority of the corresponding component, enhancing heat dissipation capacity by increasing the cooling medium flow rate or fan speed. When the temperature exceeds... When the temperature drops by 20%, it is classified as a Level 2 fault. The system prioritizes the cooling of the corresponding component to the highest level and simultaneously cuts off unnecessary waste heat transport paths, concentrating cooling resources to protect high-risk components. When the temperature exceeds 30%, it is classified as a Level 3 fault. In addition to implementing all the measures for a Level 2 fault, the system further limits the load rate of the corresponding motor to no more than 50%, reducing heat generation at the source and preventing shutdown or damage caused by continuous temperature rise. At the same time as the fault is triggered, the control system automatically starts the fault data recording process, completely saving the fault occurrence time, fault component information, component temperature change curves in the 10 seconds before the fault, current cooling system operating parameters, historical control command records, and attitude angle and position deviation data at the time of the fault. All fault data is uniformly stored in the database for fault cause tracing and system optimization analysis. Through the mechanism of combining graded response and data recording, the whole machine thermal management system can quickly intervene and accumulate operating experience when the temperature is abnormal, continuously improving fault response capabilities and long-term operational reliability.

[0020] Example 2, as Figures 1 to 6 As shown, based on Embodiment 1, the present invention provides a technical solution: In a plowing operation scenario (high load condition), the lidar detects that the terrain undulation is 8°, the soil moisture sensor obtains a soil moisture content of 35%, and the operation mode recognizer identifies the operation type as plowing; at the same time, it captures that the walking motor and hydraulic motor have high speed and large torque, the battery discharge current increases significantly, and the real-time temperature of each component rises rapidly, specifically the walking motor 75°C, the hydraulic motor 70°C, and the battery 45°C. All collected data are processed and stored. Based on the collected component type information, the corresponding component power heat loss coefficients are matched, with the motor at 0.85, the battery at 0.05, and the electronic control at 0.1. For the high-load operation type of plowing, the system calls the working condition coupling coefficient table to match and obtain a working condition coupling coefficient of 0.9. The prediction result shows that the heat load will continue to remain at the edge of the safety threshold for the next 5 minutes. After the segmented adaptive logic detection of heat load prediction deviation correction, the prediction deviation is 1.2%, which is in the range of 1% < deviation ≤ ±2%. Therefore, the correlation coefficient is corrected by +0.02 to compensate for the additional heat generation trend brought about by high resistance operation, so that the heat load prediction result is more in line with the actual heat generation situation in operation. Due to the extremely high heat load of each component in the overall power system, after retrieving the actual energy consumption and actual temperature data of the components, the reward value calculated by the three-dimensional target DRL reward algorithm is less than 0.3. Based on the current operation type, soil moisture content and terrain undulation data, the system matches the benchmark decision data of the main plowing block and the corresponding high slope and high moisture content sub-block in the four-dimensional decision matrix, adjusts the cooling power and fan speed in combination with the cooling priority coefficient, corrects the heat flow distribution ratio in combination with the energy consumption optimization coefficient, and determines the waste heat utilization path based on the waste heat classification prediction results. Finally, a strong cooling control scheme is generated: the power of the walking motor cooling water pump is adjusted to 90%, the cooling flow rate of the hydraulic motor is 18 L / min, and the speed of the electric fan is 2800 r / min, to suppress the temperature rise of the components and avoid the interruption of operation due to overheating, thus ensuring the continuity of plowing operation. The heat exchanger continuously collects high-temperature waste heat generated during operation. If the waste heat temperature is detected to be higher than 65°C, the system classifies it and directs most of the high-temperature waste heat to the radiator for dissipation through the electromagnetic reversing valve. Only a small amount of high-temperature flow path is retained to maintain the battery at the optimal discharge temperature. If the battery temperature is detected to be within the appropriate range, the flow path is immediately cut off. The system monitors the temperature of each component in real time. If the motor temperature is found to be trending towards exceeding the safety threshold, the first-level fault emergency mechanism is immediately triggered to further limit the auxiliary load and prioritize the heat dissipation needs of the walking power system. When a fault occurs, the system automatically records the fault occurrence time, fault component information, component temperature change curve in the 10 seconds before the fault, cooling system operating parameters, and control command history, providing a basis for subsequent fault investigation and system optimization, and ensuring that the system can still operate stably under extreme conditions.

[0021] In summary, plowing is a high-load operation with significant terrain undulations, high soil moisture content, and extremely strong operational resistance. The entire machine's power system experiences peak thermal load. This invention comprehensively captures the operating conditions and system status through multi-dimensional data acquisition, laying a solid data foundation for subsequent processes. It accurately predicts thermal load trends using a condition-coupled thermal load prediction algorithm, with deviation correction ensuring accuracy. The intelligent decision-making process prioritizes temperature to generate a strong cooling control scheme, effectively suppressing temperature rise. Waste heat distribution focuses on the rational dissipation and limited recovery of high-temperature waste heat to avoid heat accumulation. Closed-loop control, combined with a fault emergency mechanism and correction algorithm, continuously optimizes the operating strategy. The entire process works synergistically to ensure stable system operation under high loads and guarantee operational continuity.

[0022] Example 3, as Figures 1 to 6 As shown, based on Embodiments 1 and 2, the present invention provides a technical solution: In a rotary tillage operation scenario (medium load condition), the lidar detects a terrain undulation of 4°, the soil moisture sensor measures a soil moisture content of 25%, and the operation mode recognizer identifies the operation type as rotary tillage; in the overall machine operation status data, the PTO motor load is stable, the walking motor load is at a medium level, and the temperature of each component is in the middle of the temperature rise, with the PTO motor at 55°C and the walking motor at 50°C. All collected data are processed and stored to comprehensively capture the operating condition characteristics and system operation status. Based on the matching of the corresponding component power heat loss coefficient for each component type, and combined with the medium load characteristics of rotary tillage, the system calls the working condition coupling coefficient table to match and obtain a working condition coupling coefficient of 0.75. The prediction result shows that the heat load is at a medium level and the temperature rise rate is slow. After the segmented adaptive logic detection of the heat load prediction deviation correction, the prediction deviation is within the allowable range and no additional correction coefficient is required, thus accurately grasping the trend of heat load change. The system retrieves actual energy consumption and component temperature data. The reward value calculated by the three-dimensional target DRL reward algorithm is between 0.3 and 0.5. Based on the current operation type, soil moisture content, and terrain undulation data, the system matches the baseline decision data of the corresponding main block and sub-block in the four-dimensional decision matrix. The nearest neighbor interpolation method is used to select the appropriate baseline decision data. The cooling power and fan speed are adjusted in combination with the cooling priority coefficient. The heat flow distribution ratio is corrected in combination with the energy consumption optimization coefficient. At the same time, the waste heat utilization path is determined based on the waste heat classification prediction results. Finally, a neutral control scheme is generated: PTO motor cooling fan speed 1800 r / min, walking motor water pump power 50%, hydraulic motor flow rate 10 L / min. This avoids slow overheating of components due to insufficient cooling and prevents energy waste due to excessive cooling, thus achieving a dynamic balance between energy consumption and temperature. The heat exchanger collects medium-temperature waste heat generated during operation. The waste heat temperature is detected to be between 45℃ and 55℃. After classification, the system uses an electromagnetic reversing valve to directionally transport the waste heat to the hydraulic oil tank for insulation. During this process, the system records data such as the amount of waste heat collected, the transport path, and the temperature change of the hydraulic oil tank. It also calculates the actual waste heat utilization rate and uploads the data. This effectively recovers medium-temperature waste heat, reduces hydraulic oil viscosity, reduces hydraulic pump power consumption, and improves overall energy utilization efficiency. The system monitors the temperature of each component in real time to ensure that the component temperature is always maintained within a safe range, which not only ensures the stable operation of the system, but also avoids unnecessary energy consumption, so that the power system of the whole machine is always in the optimal operating state of energy consumption and temperature balance during rotary tillage. In summary, rotary tillage is a medium-load operation with mild conditions and small fluctuations in heat load. The core objective is to balance energy consumption and temperature. Multi-dimensional data acquisition accurately captures the characteristics and system status under medium load conditions. Heat load prediction clearly indicates a moderate heat load and a gradual temperature rise, providing a basis for balanced regulation. Intelligent decision-making, through a three-dimensional target DRL reward algorithm and balancing rules, generates a solution that balances energy consumption and temperature control, avoiding over-cooling or under-cooling. Waste heat distribution recovers medium-temperature waste heat for hydraulic tank insulation, improving energy utilization. Closed-loop control continuously fine-tunes and optimizes, ensuring that the entire machine's power system maintains efficient and balanced operation throughout the entire operation, achieving a win-win situation for energy consumption and reliability.

[0023] Example 4, as Figures 1 to 6 As shown, based on Embodiments 1, 2, and 3, the present invention provides a technical solution: In a transportation operation scenario (low-load condition), the lidar detects a terrain undulation of 1°, the operation mode recognizer identifies the operation type as transportation, and the soil moisture sensor collects soil moisture content data; in the three overall machine operation status data, the walking motor is in a low-torque, high-speed cruising state, the PTO motor and hydraulic motor are basically in standby or low-load state, and the temperature of each component is low, with the battery at 28°C and the walking motor at 40°C. All collected data are processed and stored to accurately capture the environmental characteristics and system operation status under low-load conditions; Based on the matching of the corresponding component power heat loss coefficient for each component type, and combined with the low load characteristics of transportation operations, the system calls the operating condition coupling coefficient table to match and obtain an operating condition coupling coefficient of 0.5. After the segmented adaptive logic detection of heat load prediction deviation correction, the prediction deviation is within a reasonable range, and no correction coefficient is required, thus clarifying the state of heat load. The system retrieves actual energy consumption and component temperature data. The reward value calculated by the three-dimensional target DRL reward algorithm is higher than 0.5. Based on the current operation type, soil moisture content and terrain undulation data, the system matches the baseline decision data of the main transportation block and the corresponding smooth road condition sub-block in the four-dimensional decision matrix. It adjusts the cooling power and fan speed in combination with the cooling priority coefficient, corrects the heat flow distribution ratio in combination with the energy consumption optimization coefficient, and determines the waste heat utilization path based on the waste heat classification prediction results. Finally, a low power consumption control scheme is generated: the power of the walking motor water pump is reduced to 30% to maintain the minimum circulation, the speed of the electric fan is reduced to 800 r / min or intermittently stopped, and the hydraulic cooling circuit is maintained at only 5 L / min basic flow rate to minimize the power consumption of the cooling system. The heat exchanger continuously collects low-temperature waste heat generated during operation. The waste heat temperature is detected to be between 30℃ and 45℃. After classification, the system implements a strategy to maximize recovery. All waste heat is directed to the battery pack through an electromagnetic reversing valve to maintain the battery pack in the optimal operating temperature range and prevent the battery from becoming too cold due to the air cooling effect. If a cab heating system is equipped, some waste heat can also be transferred to this system. During this process, the system records data such as waste heat collection volume, transport path, and battery temperature changes in real time, and calculates and uploads the actual waste heat utilization rate to fully explore the utilization value of low-grade waste heat and improve energy recovery efficiency. The system focuses on monitoring whether there is local heat accumulation in each component. If no abnormality is detected, it will continue to maintain a low power consumption operation mode. Under the premise of ensuring stable system operation, it will minimize energy consumption, extend the driving range of electric tractors, and give full play to the energy optimization advantages under low load conditions. In summary, transportation operations are characterized by low load conditions, smooth road surfaces, and low heat load. The key focus is on energy consumption optimization and waste heat recovery. Multi-dimensional data acquisition accurately captures system and operating condition information under low load conditions; heat load prediction clearly identifies extremely low and stable heat load states, providing support for low-power consumption control; intelligent decision-making prioritizes energy consumption, significantly reducing cooling system power consumption; waste heat distribution maximizes the recovery of low-temperature waste heat to maintain the battery's optimal operating temperature, fully exploiting energy value; closed-loop control focuses on monitoring heat accumulation and optimizing heat flow distribution, minimizing energy consumption and extending range while ensuring system stability, fully leveraging the energy optimization advantages under low load conditions.

[0024] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for coordinated adaptive control of thermal and energy performance of an electric tractor, characterized in that, The specific steps of this method are as follows: S100 collects farmland working condition data and overall machine operation status data through the sensing equipment mounted on the electric tractor, and stores the data after processing. The farmland working condition data and overall machine operation status data include farmland topography undulation, soil moisture, operation type, overall machine operation posture, and two-dimensional position data. S200, based on the collected data, matches the differentiated component power heat loss coefficients, combines the working condition coupling heat load prediction algorithm with the tractor attitude angle and position changes, calculates the future heat load of each component, compares the heat load deviation, and corrects the working condition coupling coefficient and the electronic control heat loss correction coefficient before outputting the prediction result; The S300 employs three-dimensional target deep reinforcement learning and an inner-layer fuzzy neural network, integrating attitude stability and path tracking deviation. It combines a four-dimensional decision matrix to generate control schemes for the power of the cooling system of each heat source component, the heat flow distribution ratio, and the waste heat utilization path. The input of the inner-layer fuzzy neural network includes the energy consumption optimization coefficient output by the deep reinforcement learning module, the difference between the real-time temperature of each component and the target temperature, and the real-time attitude angle deviation and path tracking deviation of the tractor. The temperature difference is calculated based on the target range of 25~35℃ for the battery, 40~60℃ for the motor, and 30~50℃ for the electronic control system. The inner fuzzy neural network has 15 built-in fuzzy rules, which are divided into 5 temperature priority rules, 5 energy consumption priority rules and 5 waste heat utilization priority rules. After the input data is fuzzified, it is defuzzified by rule reasoning and the centroid method is used to complete the defuzzification. The output is the 0~1 interval cooling priority coefficient of the walking motor, power output shaft motor and hydraulic motor. Based on the operation type, soil moisture content, and terrain undulation amplitude, the main blocks and sub-blocks of the four-dimensional decision matrix are matched. The nearest neighbor interpolation method is used to select the benchmark decision data. The control scheme is generated by combining the cooling priority coefficient and the energy consumption optimization coefficient. The control scheme includes the power of each motor cooling water pump, the speed of the electric fan, the heat flow distribution ratio, and the specific utilization paths of high temperature waste heat to battery preheating and medium and low temperature waste heat to the hydraulic system. The four-dimensional decision matrix is ​​divided into 4 main blocks according to the operation type of tillage, sowing, harvesting, and transportation. Each main block is divided into 9 sub-blocks according to the combination of soil moisture content of 10%~20%, 20%~30%, 30%~40% and terrain undulation amplitude of 0°~3°, 3°~6°, 6°~10°. Each sub-block stores 10 sets of optimal benchmark decision data. S400 collects and classifies waste heat through a heat exchanger, then delivers it in a directional manner through an electromagnetic reversing valve, records the data, calculates the actual waste heat utilization rate, and uploads it. The S500 outputs control commands according to the control scheme, calculates the heat flow distribution correction coefficient through the waste heat feedback heat flow correction algorithm and feeds it back, while monitoring the component temperature and adjusting the strategy in real time in combination with attitude and position deviations to execute the graded fault emergency mechanism.

2. The method for coordinated adaptive control of thermal energy in an electric tractor according to claim 1, characterized in that: Specifically, S100 includes: The system acquires farmland topographic relief data via a lidar sensor, collects soil moisture content data via a soil moisture sensor, and obtains operation type identification information via an operation mode recognizer. The operation types include four modes: tillage, sowing, harvesting, and transportation. The torque and speed sensors deployed on the walking motor, power output shaft motor and hydraulic motor collect real-time operating power data of each motor, obtain battery state of charge and charging / discharging current data through the battery management system, and obtain the operating power of the electronic control system through the electronic control module monitoring unit. The tractor's overall operating attitude angle data, including pitch and roll angles, are collected by an inertial measurement unit. The tractor's two-dimensional position coordinate data are obtained by a global positioning module. All collected data are stored in a historical database after being processed to be dimensionless.

3. The method for coordinated adaptive control of thermal and energy performance of an electric tractor according to claim 1, characterized in that: Specifically, S200 includes: The onboard lidar acquires the corresponding component power heat loss coefficient based on the collected component type identification information. Among them, motor components are uniformly matched with 0.85, batteries with 0.05, and electronic control modules with 0.

1. The working condition coupling coefficient is matched based on the operation type identification information, with 0.9 for tillage, 0.7 for sowing, 1.0 for harvesting, and 0.5 for transportation. The future heat load of the component is calculated using a condition-coupled heat load prediction algorithm. The algorithm formula is as follows: ,in, For the future heat load of the components, The component's power heat loss coefficient. The coupling coefficient is the operating condition coefficient. This is the correction factor for heat loss in electronic control systems. For the real-time operating power of the components, The soil moisture content is after dimensionless treatment. This represents the dimensionless variation of the terrain undulations. Input voltage to the electronic control module. This refers to the operating current of the electronic control module; The real-time attitude angle and two-dimensional position change data of the tractor are introduced to dynamically correct the working condition coupling coefficient. When the pitch angle or roll angle exceeds the preset threshold, positive compensation is applied to the working condition coupling coefficient based on the angle deviation. When the path tracking deviation exceeds the preset range, the heat load prediction result is calibrated in real time based on the position deviation.

4. The method for coordinated adaptive control of thermal energy in an electric tractor according to claim 3, characterized in that: The S200 further includes: The predicted heat load is compared with the actual measured heat load, the relative deviation percentage is calculated, and the operating condition coupling coefficient and the electronic control heat loss correction coefficient are dynamically corrected using piecewise adaptive logic. When the deviation is ≤ ±1%, the original coefficient remains unchanged; when 1% < deviation ≤ ±2%, the coefficient correction range is ±0.02; when 2% < deviation ≤ ±3%, the coefficient correction range is ±0.

05. The correction direction is determined according to the sign of the deviation, so that the heat load prediction result gradually approaches the true value. When the absolute value of the deviation exceeds ±3%, in addition to performing coefficient correction, the sensor fault self-test process is triggered simultaneously. The sensor fault self-test process includes power-on status detection and signal output continuity detection of temperature sensor and power sensor, and the detection results are uploaded to the control system.

5. The method for coordinated adaptive control of thermal energy in an electric tractor according to claim 1, characterized in that: Specifically, S300 includes: The system standard energy consumption, actual operating energy consumption, real-time temperature of each component, target temperature of each component, upper and lower limits of safe temperature of each component, and current waste heat utilization rate data are retrieved as input parameters for the three-dimensional target deep reinforcement learning reward algorithm. The reward value for the current state is calculated using a three-dimensional target deep reinforcement learning reward algorithm. The algorithm formula is as follows: ,in, As a reward value, To optimize the weighting coefficients for energy consumption, This is the temperature stability weighting coefficient. This is the weighting coefficient for waste heat utilization. This is the standard energy consumption of the system. This represents the actual energy consumption of the system during operation. This refers to the actual temperature of the component. The target temperature for the component. The upper limit of the safe temperature for the component. The lower limit of the safe temperature of the component, Waste heat utilization rate; The real-time attitude angle deviation and path tracking deviation of the tractor are used as supplementary dimensions of the state space and input into the deep reinforcement learning module. When the attitude angle deviation or path tracking deviation exceeds the preset threshold, the reward algorithm introduces a penalty term to guide the policy network to output a coordinated action that takes into account both thermal management and driving stability.

6. The method for coordinated adaptive control of thermal and energy performance of an electric tractor according to claim 5, characterized in that: The S300 also includes: The deep reinforcement learning module is constructed using the deep deterministic policy gradient algorithm and includes two independent network structures: a policy network and a target network. The policy network is responsible for making action decisions that output energy consumption optimization coefficients, while the target network is responsible for calculating the target action value function. The capacity of the experience replay pool for the deep reinforcement learning module is set to 10,000 records. The reward value calculated by the 3D target deep reinforcement learning reward algorithm, together with the corresponding state space data, action space data, next state data and termination flag, are stored in the replay pool as experience data. 64 experience data are randomly sampled each time training is conducted, and the policy network is trained using the gradient descent method. The training iteration step size is set to 0.001, and the target network synchronizes the parameters of the policy network once every 200 training steps. The policy network dynamically adjusts the direction of the output action based on the reward value. When the reward value is higher than 0.5, it shifts towards reducing energy consumption, and when the reward value is lower than 0.3, it shifts towards ensuring stable temperature. The original decision value output is normalized and converted into an energy consumption optimization coefficient in the range of 0 to 1. Then, it is processed by a filter with a smoothing window of 5 decision cycles before being output to the inner fuzzy neural network.

7. The method for coordinated adaptive control of thermal energy in an electric tractor according to claim 1, characterized in that: Specifically, S400 includes: Waste heat is collected by heat exchangers deployed at the outlets of each heat source component and classified according to the waste heat temperature level, which includes high temperature level ≥65℃, medium temperature level 45~55℃ and low temperature level 30~45℃. The waste heat after classification is directed to the target utilization unit through the electromagnetic reversing valve group. High-temperature waste heat is preferentially transported to the battery preheating circuit, medium-temperature waste heat is transported to the hydraulic oil tank insulation circuit, and low-temperature waste heat is transported to the cab heating system or battery insulation circuit. The system records the waste heat collection volume of each heat exchanger, the opening status and duration of each electromagnetic reversing valve, and the temperature change data of each target utilization unit in real time. Based on the total waste heat collected and the actual waste heat utilized, the system calculates the actual waste heat utilization rate and uploads the data to the intelligent decision-making module.

8. The method for coordinated adaptive control of thermal energy in an electric tractor according to claim 1, characterized in that: The S500 specifically includes: Based on the control scheme generated by the intelligent decision-making module, control commands are output to the cooling water pump, electric fan and electromagnetic reversing valve actuator, while real-time data on the actual temperature and waste heat utilization rate of each component are collected. The waste heat feedback heat flow correction algorithm is used to calculate the correction coefficient for the next round of heat flow distribution. The algorithm formula is as follows: ,in, Correction factor for the next round of heat flow allocation. This is the current heat flux distribution coefficient. To provide feedback and correct the sensitivity coefficient, For actual waste heat utilization efficiency, Target waste heat utilization efficiency; The calculated heat flow distribution correction coefficient is fed back to the heat flow distribution control module to dynamically adjust the cooling medium flow distribution ratio of each heat source component in the next decision cycle, so that the actual waste heat utilization rate gradually approaches the target waste heat utilization rate, forming a closed-loop adaptive control.

9. The method for coordinated adaptive control of thermal and energy performance of an electric tractor according to claim 8, characterized in that: The S500 also includes: The temperature of each component is monitored in real time and compared with the preset safe temperature threshold. At the same time, the tractor attitude angle deviation and path tracking deviation are monitored, and the graded fault emergency mechanism is triggered according to the degree to which the temperature exceeds the safe threshold. The graded fault emergency mechanism adopts a three-level response strategy. The first-level fault is when the temperature exceeds the safety threshold by 10%~20%, only the cooling priority of the corresponding component is increased. The second-level fault is when the temperature exceeds the safety threshold by 20%, the cooling priority is increased to the highest level and unnecessary waste heat transfer paths are cut off. The third-level fault is when the temperature exceeds the safety threshold by 30%, the load rate of the corresponding motor is simultaneously limited to ≤50%. When a fault is triggered, the fault occurrence time, fault component information, component temperature change curve 10 seconds before the fault, cooling system operating parameters, control command history, and attitude angle and position deviation data at the time of the fault are recorded synchronously. All fault data are stored in the database.

Images