An electric drive control system for a loader

Abstract

The application provides an electric loader driving control system and relates to the technical field of engineering machinery, and the system comprises a working device hydraulic assembly, a walking assembly, a power assembly and a whole vehicle control unit; a control method obtains whole machine state information through the whole vehicle control unit, determines a current working condition through an operation state recognition model, combines a loader dynamics model to calculate the theoretical demand of the hydraulic assembly and the walking assembly, further takes the optimal system efficiency of the whole machine as a core target, takes the working efficiency and the controllability as hard constraints, constructs a multi-agent collaborative optimization model, solves a hydraulic assembly motor speed, a walking motor torque, a hydraulic valve and a hydraulic pump displacement control function, and realizes the collaborative control and energy recovery of each subsystem. The application can effectively improve the whole machine energy efficiency of the electric loader and realize the recycling of negative load energy.

Description

Technical Field

[0001] This invention relates to the field of engineering machinery technology, and more specifically to an electric loader drive control system. Background Technology

[0002] Currently, electrification of construction machinery has become an important trend in the industry, especially in typical operating equipment such as loaders, where electrification is considered an effective means to improve energy efficiency and reduce emissions. However, in existing technologies, the drive systems of electric loaders mostly use the hydraulic transmission architecture of traditional fuel-powered loaders, simply replacing the engine with an electric motor, without fundamentally optimizing the system structure and control strategy. This simple replacement of the engine with an electric motor, while achieving energy savings in the power components to some extent, still results in low overall drive system efficiency. Especially in terms of energy recovery under heavy load conditions (such as boom descent and braking deceleration), existing systems lack effective recovery mechanisms, leading to significant energy waste. Furthermore, existing systems have shortcomings in the coordinated control of hydraulic and travel components, making it difficult to optimize overall machine efficiency, thus causing problems such as high energy consumption, large battery capacity requirements, and poor range in electric loaders.

[0003] In view of the above, this application is hereby submitted. Summary of the Invention

[0004] The present invention provides an electric loader drive control system that can at least partially improve the above-mentioned problems.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] An electric loader drive control system includes: a working device hydraulic component, a traveling component, a power component, and a vehicle control unit, wherein the data terminal of the vehicle control unit is electrically connected to the data terminals of the working device hydraulic component, the traveling component, and the power component, and the power component is electrically connected to the working device hydraulic component and the traveling component. The vehicle control unit is configured to perform the following steps by executing a computer program stored internally thereon: Obtain the overall vehicle status information and control signals of the loader at time t, and determine the working status of the loader at time t based on the overall vehicle status information at time t; Based on the control signal and operating status at time t, calculate the theoretical required flow rate of the hydraulic components of the working device and the theoretical required torque of the motor of the traveling component at time t. Using a trained multi-agent collaborative optimization model, the theoretical demand flow and the theoretical demand torque of the motor at time t are preprocessed to obtain the vehicle state information and control signal at time t+1. The vehicle state information and control signals at time t+1 are saved to the training dataset. Based on the updated training dataset, the multi-agent cooperative optimization model is retrained to obtain a newly trained multi-agent cooperative optimization model, which is used for control preprocessing at the next time step.

[0007] In summary, this invention provides an electric loader drive control system to solve the technical problems of low drive system efficiency and difficulty in recovering load in existing technologies. The system integrates a hydraulic component for the working device, a traveling component, a power component, and a vehicle control unit. The vehicle control unit is interconnected with each component at the data and electrical levels, forming a closed-loop control architecture. The vehicle control unit stores a computer program. During execution, it first acquires the loader's current vehicle status information and control signals, accurately identifying the loader's real-time operating status. Based on this, it calculates the theoretical flow rate requirement of the hydraulic component for the working device and the theoretical torque requirement of the traveling component's motor, combining the control signals and the operating status. Subsequently, it uses a pre-trained multi-agent cooperative optimization model to perform control preprocessing on the above theoretical requirements, generating the vehicle status and control information for the next moment, and storing it in the training dataset. Finally, based on the updated training data, the multi-agent cooperative optimization model is continuously iteratively optimized to adapt to actual operating conditions for precise control in subsequent moments. This invention achieves adaptive optimization and collaborative control of the electric loader drive system by constructing a data-driven model training and real-time control closed loop, effectively improving the overall energy efficiency, response speed and energy recovery capability of the machine.

[0008] Compared to the open-loop, static control methods in existing technologies, this invention achieves the following beneficial effects by establishing a closed-loop control link of "data acquisition - state recognition - collaborative optimization - model iteration": Through multi-agent collaborative optimization, the energy efficiency of hydraulic components and traveling components is balanced in real time, significantly improving the overall energy efficiency; by utilizing model prediction and iterative updates, effective energy recovery is achieved under heavy load conditions (such as boom descent and braking deceleration); and the control system possesses adaptive capabilities, continuously optimizing the control strategy according to changes in operating conditions, solving the problems of rigid control and poor adaptability in existing systems. Attached Figure Description

[0009] Figure 1 This is a structural framework diagram of the electric loader drive control system provided in an embodiment of the present invention.

[0010] Figure 2 This is a flowchart illustrating the electric loader drive control system provided in an embodiment of the present invention. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0012] refer to Figure 1 As shown, the first embodiment of the present invention discloses an electric loader drive control system, which includes: a working device hydraulic component, a walking component, a power component, and a vehicle control unit 1, wherein the data terminal of the vehicle control unit 1 is electrically connected to the data terminals of the working device hydraulic component, the walking component, and the power component, and the power component is electrically connected to the working device hydraulic component and the walking component; The hydraulic components of the working device include: a first electric pump 2, a second electric pump 3, a first relief valve 4, a second relief valve 5, a flow priority valve 6, a distribution valve group 7, a large displacement steering gear 8, a steering cylinder 9, a bucket cylinder 10, an accumulator 11, and a boom cylinder 12. The outlet of the first electric pump 2 is connected to the inlet of the first relief valve 4, the outlet of the first relief valve 4 is connected to the inlet of the flow priority valve 6, the first outlet of the flow priority valve 6 is connected to the inlet of the large displacement steering gear 8, the outlet of the second electric pump 3 is connected to the inlet of the second relief valve 5, the outlet of the second relief valve 5 and the distribution valve group 7 are connected to the second outlet of the flow priority valve 6, the large displacement steering gear 8 is connected to the steering cylinder 9, and the distribution valve group 7 is connected to the bucket cylinder 10, the boom cylinder 12, and the accumulator 11. The data terminal of the vehicle control unit is electrically connected to the first electric pump 2, the second electric pump 3, the flow priority valve 6, and the distribution valve group 7.

[0013] The walking assembly includes a walking motor controller 13, a walking motor 14, a multi-speed gearbox 15, a drive shaft 16, and a drive axle (front axle 17 + rear axle 18). The power assembly includes a power battery pack 19, a charging socket 20, a power battery energy management module 21, a power battery thermal management module 22, and an all-in-one controller 23. The data terminal of the vehicle control unit is electrically connected to the data terminals of the power battery thermal management module 22, the power battery energy management module 21, the control terminal of the all-in-one controller 23, the control terminal of the walking motor controller 13, and the control terminal of the multi-speed gearbox 15.

[0014] Specifically, in this embodiment, the system includes a hydraulic component for the working device, a traveling component, a power component, and a vehicle control unit. The vehicle control unit, as the system's control core, is electrically connected to the data terminals of the hydraulic component, the traveling component, and the power component, thereby enabling real-time monitoring of the status of each subsystem and the transmission of control commands. Simultaneously, the power component is electrically connected to the hydraulic component and the traveling component, providing the electrical energy required for their operation. Through this architecture, the vehicle control unit can acquire the overall machine's operating status in real time and coordinate the scheduling of each subsystem based on a preset control strategy. Compared to the open-loop control method in existing technologies that simply replace the power source, the closed-loop electrical connection architecture of this invention lays the hardware foundation for achieving overall machine energy efficiency optimization and energy recovery.

[0015] Specifically, the hydraulic components of the working device include a first electric pump 2, a second electric pump 3, a first relief valve 4, a second relief valve 5, a flow priority valve 6, a distribution valve group 7, a large-displacement steering gear 8, a steering cylinder 9, a bucket cylinder 10, an accumulator 11, and a boom cylinder 12. In terms of hydraulic pipeline connections, the outlet of the first electric pump 2 is connected to the inlet of the first relief valve 4, the outlet of the first relief valve 4 is connected to the inlet of the flow priority valve 6, and the first outlet of the flow priority valve 6 is connected to the inlet of the large-displacement steering gear 8. The outlet of the second electric pump 3 is connected to the inlet of the second relief valve 5, and the outlet of the second relief valve 5 and the distribution valve group 7 are connected to the second outlet of the flow priority valve 6. The large-displacement steering gear 8 is connected to the steering cylinder 9 and is used to convert hydraulic energy into mechanical energy for the steering cylinder 9, thereby enabling the loader's steering operation. The distribution valve assembly 7 is connected to the bucket cylinder 10, boom cylinder 12, and accumulator 11, and is used to distribute hydraulic oil to the bucket cylinder 10, boom cylinder 12, or accumulator 11 according to operational needs. In this hydraulic architecture, the first motor pump 2 and the second motor pump 3 adopt a dual-pump oil supply method, and steering priority control is achieved through the flow priority valve 6, that is, the flow demand of the steering cylinder 9 is prioritized, and the excess flow is then distributed to the distribution valve assembly 7 for the operation of the bucket and boom, thereby ensuring the steering safety of the loader under any working condition. At the same time, the introduction of the accumulator 11 allows the hydraulic potential energy under heavy load conditions such as boom descent to be recovered and stored, and released for auxiliary oil supply when the system needs it, effectively improving the energy utilization efficiency of the hydraulic components. The data terminal of the vehicle control unit is electrically connected to the first motor pump 2, the second motor pump 3, the flow priority valve 6, and the distribution valve group 7, thereby enabling real-time adjustment of the motor pump speed, the flow priority valve 6 diversion ratio, and the valve core opening of the distribution valve group 7, achieving precise control of the flow and pressure of the hydraulic components, and further improving the action response speed and stability of the working device.

[0016] The travel assembly includes a travel motor controller 13, a travel motor 14, a multi-speed gearbox 15, a drive shaft 16, and a drive axle. The travel motor controller 13 receives commands from the vehicle control unit and controls the output torque and speed of the travel motor 14. The power from the travel motor 14 is transmitted to the drive axle via the multi-speed gearbox 15 and drive shaft 16, ultimately driving the loader. The multi-speed gearbox 15 allows the travel motor 14 to operate in its efficient speed range according to different working conditions, thereby reducing energy consumption. The power assembly includes a power battery pack 19, a charging socket 20, a power battery energy management module 21, a power battery thermal management module 22, and a multi-function controller 23. The power battery pack 19 serves as the vehicle's energy source, supplying power to the first motor pump 2, the second motor pump 3, and the travel motor 14 via the multi-function controller 23. The power battery energy management module 21 monitors the battery pack's state of charge, voltage, current, and other parameters, and feeds this information back to the vehicle control unit so that the vehicle control unit can rationally allocate power output based on the battery status. The power battery thermal management module 22 is used to maintain the battery pack within a suitable temperature range, preventing battery performance degradation or lifespan reduction due to excessively high or low temperatures. The data terminals of the vehicle control unit are electrically connected to the data terminals of the power battery thermal management module 22, the power battery energy management module 21, the control terminals of the multi-function controller 23, the drive motor controller 13, and the multi-speed transmission 15. Through these electrical connections, the vehicle control unit can obtain real-time power battery status information and coordinate the control of the multi-function controller 23, the drive motor controller 13, and the multi-speed transmission 15 to achieve drive and braking energy recovery of the drive motor 14. When the loader brakes or decelerates, the drive motor 14 enters a power generation state, converting mechanical energy into electrical energy, which is then fed back to the power battery pack 19 via the multi-function controller 23, thereby recovering driving and braking energy and further reducing overall power consumption.

[0017] Please see Figure 2 The vehicle control unit is configured to perform the following steps by executing a computer program stored internally thereon: S1, obtain the overall vehicle status information and control signals of the loader at time t, and determine the working status of the loader at time t based on the overall vehicle status information at time t; Specifically, step S1 further includes: vehicle status information at time t. The formula is: , The characteristics of the hydraulic components of the working device at time t. The characteristics of the walking component at time t, The characteristics of the power component at time t, The characteristics of the vehicle control unit at time t are as follows: For transpose; Control signal at time t The formula is: , For brake pedal opening, This refers to the accelerator pedal opening. For boom handle signal, For the bucket handle signal, For steering wheel signals; in, , The outlet pressure of the first motor pump 2, The outlet pressure of the second motor pump 3. The first outlet pressure of the flow priority valve 6, The second outlet pressure of the flow priority valve 6, The pressure at the return port. The pressure of the left steering cylinder. This refers to the pressure of the right steering cylinder. This refers to the pressure in the large chamber of the bucket cylinder 10. The pressure in the small chamber of the bucket cylinder 10, For the pressure of accumulator 11, For the large chamber pressure of boom cylinder 12, The pressure in the small chamber of boom cylinder 12; , This is the torque of the walking motor 14. v is the rotational speed of the travel motor 14, v is the vehicle speed, and a is the vehicle acceleration. This is the 15th gear of a multi-gear transmission; , For battery power, Battery voltage, Battery current, For battery cell temperature, For the power of the all-in-one controller 23, For the voltage of the all-in-one controller 23, For the current of the all-in-one controller 23, Temperature for the all-in-one controller 23; , For steering angle, For the steering angular velocity, The speed of bucket cylinder 10, The speed of boom cylinder 12.

[0018] The formula for determining the work status is: , Let t be the job status at time t. For job status identification mapping function, For trainable parameters, For normalized exponential functions, It is a multilayer perceptron. The encoder layer consists of a multi-head self-attention network and a feedforward network. The formula for the multi-head self-attention mechanism is as follows: , , This is a single-head self-attention output, where Q is the query matrix, K is the key matrix, and V is the value matrix. Scaling factor For multi-head self-attention output, For splicing operations, The first self-focused head, Let h be the h-th self-attention head, where h is the number of self-attention heads. Let be the transformation matrix.

[0019] In this embodiment, the vehicle state information at time t is constructed as a state vector, which consists of four sub-features: features of the hydraulic components of the working device, features of the traveling component, features of the power component, and features of the vehicle control unit. The features of the hydraulic components of the working device are further expanded into a vector containing twelve pressure parameters, specifically including the outlet pressure of the first motor pump 2, the outlet pressure of the second motor pump 3, the first outlet pressure of the flow priority valve 6, the second outlet pressure of the flow priority valve 6, the return port pressure, the left steering cylinder pressure, the right steering cylinder pressure, the large chamber pressure of the bucket cylinder 10, the small chamber pressure of the bucket cylinder 10, the accumulator 11 pressure, the large chamber pressure of the boom cylinder 12, and the small chamber pressure of the boom cylinder 12. The features of the traveling component include the torque of the traveling motor 14, the speed of the traveling motor 14, the vehicle speed, the vehicle acceleration, and the gear position of the multi-gear transmission 15. The characteristics of the power components include battery power, battery voltage, battery current, battery cell temperature, power of the multi-function controller 23, voltage of the multi-function controller 23, current of the multi-function controller 23, and temperature of the multi-function controller 23. The characteristics of the vehicle control unit include steering angle, steering angular velocity, speed of bucket cylinder 10, and speed of boom cylinder 12. Through the collection of the above multi-dimensional status information, the vehicle control unit can comprehensively perceive the real-time operating status of the loader in terms of hydraulic components, travel components, power components, and overall machine attitude, providing a rich data foundation for accurate judgment of subsequent operating status.

[0020] Simultaneously, the control signals at time t are constructed into a control signal vector, specifically including brake pedal opening, accelerator pedal opening, boom handle signal, bucket handle signal, and steering wheel signal. These control signals directly reflect the driver's operating intentions and are crucial for judging the loader's operating status. Combining the vehicle status information with the control signals allows the system to comprehensively understand the current operating condition from both the "actual vehicle response" and "driver's expectations," thereby improving the accuracy and robustness of operating status identification.

[0021] After acquiring the state information and control signals, the vehicle control unit uses a pre-defined operation state recognition model to determine the operation state at time t. The encoder layer of this model consists of a multi-head self-attention network and a feedforward network. In the encoder layer, the multi-head self-attention mechanism is used to extract the correlation features between different state signals. Through this mechanism, the model can calculate attention weights from multiple different representation subspaces, allowing different attention heads to focus on different types of signal correlation patterns. For example, one head might focus on the correlation between steering operation and steering cylinder 9 pressure, another on the correlation between bucket operation and bucket cylinder 10 speed, and yet another on the temporal state change patterns. Compared to traditional single-feature extraction methods, the multi-head self-attention mechanism significantly enhances the representation ability of complex operating conditions, thereby improving the accuracy of operation state recognition. Based on this, the normalized exponential function and multilayer perceptron are combined to complete the final classification output, obtaining the current operation state of the loader, such as forward digging, digging and lifting, fully loaded reverse, forward unloading, unloading, or empty return, etc. Accurate identification of the operating status is a prerequisite for achieving refined control. Only by clearly knowing which operating stage the loader is currently in can we formulate targeted coordinated control strategies for the hydraulic and travel components, thereby avoiding the problem of poor energy efficiency under different operating conditions caused by the use of uniform control parameters in existing technologies.

[0022] S2, based on the control signal and working status at time t, calculate the theoretical required flow rate of the hydraulic components of the working device and the theoretical required torque of the motor of the traveling component at time t respectively; Specifically, step S2 further includes: calculating the theoretical flow rate of the steering cylinder 9 at time t based on the control signal and operating status at time t, combined with the flow diversion characteristics of the flow priority valve 6. Theoretical flow rate of the working cylinder Its formula is: , , , , ; in, For steering gear displacement, For steering speed, This is the hydraulic leakage compensation coefficient. The job status at time t The corresponding flow coefficient of the steering cylinder 9 that the entire vehicle can pass through. The effective working area of ​​the bucket cylinder 10 is... Let t be the theoretical speed of the bucket cylinder 10 at time t. The effective working area of ​​boom cylinder 12 is... Let t be the theoretical speed of boom cylinder 12 at time t. The job status at time t The corresponding hydraulic cylinder flow coefficient that the entire vehicle can pass through. The steering speed constant is The speed constant of the bucket cylinder 10 is... The speed constant of boom cylinder 12; Calculate the theoretical motor drive torque at time t based on the control signal and operating status at time t. and theoretical motor braking torque Its formula is: , ,in, Main reduction ratio, Let be the real-time gear ratio of the transmission at time t. The total transmission efficiency of the drive axle. This is the theoretical motor drive torque constant. This is the theoretical braking torque constant of the motor. The job status at time t The corresponding torque coefficient of the walking motor that the whole vehicle can provide is 14.

[0023] In this embodiment, the vehicle control unit calculates the theoretical flow rate requirement of the hydraulic components of the working device and the theoretical torque requirement of the motor of the traveling component at time t, based on the control signal and operating status at time t. The core of this step lies in combining the driver's control intention with the current operating stage of the loader, transforming it into specific execution parameters for the hydraulic and traveling components, thereby providing clear demand targets for subsequent multi-agent collaborative optimization.

[0024] The theoretical flow rate requirement for the hydraulic components of the working device is determined by the operational needs of the three actuators: steering, bucket, and boom. Specifically, the vehicle control unit calculates the theoretical flow rate of the steering cylinder 9 and the working cylinder based on the control signal and operating status at time t, combined with the flow-diverting characteristics of the flow priority valve 6. The expression for the theoretical flow rate of the steering cylinder 9 is: the theoretical flow rate of the steering cylinder 9 equals the steering gear displacement multiplied by the steering speed, then multiplied by the hydraulic leakage compensation coefficient, and finally multiplied by the flow rate coefficient of the steering cylinder 9 that the entire vehicle can pass under the operating status at time t. Here, the steering gear displacement is determined by the structural parameters of the large-displacement steering gear 8 itself; the steering speed reflects the speed at which the driver turns the steering wheel; the hydraulic leakage compensation coefficient is used to compensate for hydraulic oil leakage losses at pipelines and joints; and the flow rate coefficient of the steering cylinder 9 is dynamically determined according to the current operating status. The theoretical flow rate of the working cylinder is calculated as follows: The theoretical flow rate of the working cylinder equals the effective working area of ​​the bucket cylinder 10 multiplied by the theoretical speed of the bucket cylinder 10, plus the effective working area of ​​the boom cylinder 12 multiplied by the theoretical speed of the boom cylinder 12, and then multiplied by the working cylinder flow rate coefficient that the entire vehicle can pass through at time t corresponding to the working state. The theoretical speed of the bucket cylinder 10 is further expressed as the product of the speed constant of the bucket cylinder 10 and the bucket handle signal; the theoretical speed of the boom cylinder 12 is further expressed as the product of the speed constant of the boom cylinder 12 and the boom handle signal; and the steering speed is expressed as the product of the steering speed constant and the steering wheel signal. Through this calculation method, the vehicle control unit can reasonably determine the theoretical required flow rate of the hydraulic components based on the driver's operation range on the handle and steering wheel, combined with the current working stage of the loader (e.g., a larger thrust but slower speed is required in the digging stage, while a faster action response is required in the unloading stage). It should be noted that the introduction of the flow coefficients of the steering cylinder 9 and the working cylinder reflects the flow-dividing characteristics of the flow priority valve 6. That is, the system prioritizes flow distribution to the steering and working devices differently under different operating conditions. When the loader is traveling in a straight line or making slight turns, more flow can be allocated to the working device to improve operating efficiency. When the loader is making large-angle turns or operating in complex terrain, the system prioritizes the flow supply to the steering cylinder 9 to ensure operational safety. This dynamic flow distribution strategy based on operating conditions significantly improves the energy efficiency and operational safety of hydraulic components compared to the fixed-ratio flow distribution method in existing technologies.

[0025] The torque requirement of the travel motor 14 is determined by the driver's accelerator and brake pedal inputs, combined with the speed ratio of the multi-speed transmission 15 and the motor's output capacity under actual operating conditions. Specifically, for the calculation of the theoretical torque requirement of the travel component motor, the vehicle control unit calculates the theoretical motor drive torque and theoretical motor braking torque based on the control signal and operating state at time t. The expression for calculating the theoretical motor drive torque is: theoretical motor drive torque equals the theoretical motor drive torque constant multiplied by the accelerator pedal opening, then multiplied by the final drive ratio, the real-time gear ratio of the transmission, the total transmission efficiency of the drive axle, and the torque coefficient of the travel motor 14 that the vehicle can provide corresponding to the operating state at time t. The expression for calculating the theoretical motor braking torque is: theoretical motor braking torque equals the theoretical motor braking torque constant multiplied by the brake pedal opening, then multiplied by the final drive ratio, the real-time gear ratio of the transmission, the total transmission efficiency of the drive axle, and the torque coefficient of the travel motor 14 that the vehicle can provide corresponding to the operating state at time t. The main reduction ratio and the overall transmission efficiency of the drive axle are determined by the mechanical structure of the traveling assembly. The real-time gear ratio of the gearbox reflects the current gear position of the multi-gear gearbox 15, while the torque coefficient of the traveling motor 14 is dynamically determined based on the current operating state. Through the above calculation method, the vehicle control unit can reasonably determine the theoretical torque requirement of the traveling motor 14 based on the driver's operation of the accelerator and brake pedals, combined with the current operating stage of the loader (e.g., a larger driving torque is needed to overcome insertion resistance during the digging stage, while speed is more important than torque during the unloaded return stage). In particular, the calculation of the theoretical motor braking torque provides a quantitative basis for the recovery of traveling braking energy—when the driver depresses the brake pedal, the vehicle control unit controls the traveling motor 14 to enter the generator state based on the calculated theoretical motor braking torque, converting the loader's kinetic energy into electrical energy to be fed back to the power battery pack 19, thereby realizing the recovery and utilization of braking energy. Compared to existing technologies that rely solely on mechanical braking to dissipate energy, this invention recovers braking energy to the greatest extent possible while ensuring braking safety through precise calculation and dynamic adjustment of the theoretical motor braking torque, effectively reducing overall power consumption and extending runtime.

[0026] S3. Using the trained multi-agent collaborative optimization model, the theoretical demand flow and the theoretical demand torque of the motor at time t are preprocessed to obtain the vehicle state information and control signal at time t+1. S4. Save the vehicle state information and control signals at time t+1 to the training dataset. Based on the updated training dataset, retrain the multi-agent cooperative optimization model to obtain a newly trained multi-agent cooperative optimization model, which is used for control preprocessing in the next time step.

[0027] Preferably, in this embodiment, when At this time, the multi-agent cooperative optimization model is pre-trained using a training dataset, and the data in the training dataset only includes pre-shipment training data; when At that time, the data in the training dataset includes pre-factory training data as well as all vehicle status information and control signals before time t.

[0028] In this embodiment, the multi-agent cooperative optimization model treats the hydraulic component, walking component, and power component as agents with their own objectives and constraints. Through cooperative interaction and game theory among the agents, the optimal control variables are obtained while simultaneously satisfying hard constraints such as work efficiency and maneuverability, under the core objective of optimizing the overall system efficiency. Specifically, the theoretical required flow rate and the theoretical required torque of the motor are used as inputs. Combined with the vehicle state information at time t (including various pressure parameters of the hydraulic component of the working device, torque, speed and vehicle speed parameters of the walking component, battery and multi-in-one controller 23 parameters of the power component, and steering and cylinder speed parameters of the whole machine), the model performs forward calculation through a pre-built neural network structure, and outputs the vehicle state information and control signal at time t+1. It should be noted that the vehicle state information output by the model at time t+1 does not directly replace the actual sensor data, but rather reflects the expected state of the system under the current control strategy. This information is used to compare with subsequent actual data to evaluate the control effect. The control signals output at time t+1 include specific execution commands such as the target speeds of the first motor pump 2 and the second motor pump 3, the target opening degrees of each solenoid valve in the distribution valve group 7, the target flow priority valve 6, and the target torque of the travel motor 14. These commands are sent to the various execution components via the data terminal of the vehicle control unit to achieve coordinated control of the hydraulic and travel components. By using a multi-agent collaborative optimization model to preprocess the theoretical requirements, rather than simply issuing the theoretical requirements directly as control commands, the system can comprehensively consider the coupling relationships and energy flow paths between the various subsystems of the machine, and reasonably modify and redistribute the requirements while meeting the driver's operating intentions. For example, when the theoretical flow demand is high but the current state of charge of the power battery is low, the model can appropriately reduce the flow output of the hydraulic components and reserve more electrical energy for the walking components, or adjust the flow priority valve 6 to prioritize steering needs and appropriately reduce the operating speed of the working device, thereby achieving optimal overall energy efficiency while ensuring basic operational capabilities. This preprocessing mechanism based on multi-agent collaboration significantly improves the overall performance of the machine under complex working conditions compared to the independent control and uncoordinated operation of each subsystem in existing technologies.

[0029] Next, the vehicle state information and control signals at time t+1 are saved to the training dataset. Based on the updated training dataset, the multi-agent cooperative optimization model is retrained to obtain a newly trained multi-agent cooperative optimization model, which is used for control preprocessing in the next time step. Specifically, the training dataset is a historical data accumulation library stored in the internal or external memory of the vehicle control unit. Each data record contains the vehicle state information and corresponding control signals at a certain time. After the vehicle control unit executes step S3 and obtains the vehicle state information and control signals at time t+1, it appends this set of data to the end of the training dataset, forming an expanded training dataset. Subsequently, the vehicle control unit uses the updated training dataset to retrain the multi-agent cooperative optimization model. The training process employs reinforcement learning algorithms (such as proximal policy optimization algorithms), continuously adjusting the network parameters within the model through iterative interaction between the model and the training data, enabling the model to output better control preprocessing results in the next control cycle. The new model obtained after training is used for control preprocessing in the next time step (between time step t+1 and time step t+2), thus forming a complete closed loop of "acquisition - computation - preprocessing - storage - training - update".

[0030] It is important to note that the sources of the model's training data are differentiated. Specifically, when t equals 0, representing the initial state before the loader is put into actual operation, the multi-agent collaborative optimization model is pre-trained using a training dataset. At this point, the training dataset only contains pre-shipment training data (e.g., 1000 training iterations). This pre-shipment training data is collected in a laboratory environment or standard test site before the loader leaves the factory, under preset standard operating conditions. Its purpose is to ensure the model has basic control capabilities before delivery to the user, guaranteeing the loader's normal operation upon first use. When t is greater than 0, meaning the loader has been put into actual operation and has begun accumulating on-site operational data, the training dataset includes both pre-shipment training data and all vehicle status information and control signals up to time t. This means that as the loader's actual operating time increases, the training dataset continuously expands, the amount of training data upon which the model is based increases continuously, and all new data originates from the loader's actual operating conditions. The benefits of this design are multifaceted: First, the model can continuously learn and adapt to the unique working conditions and operating habits of loaders in actual operation, such as the digging resistance characteristics under different material densities, the driving and steering characteristics under different site conditions, and the differences in operating styles of different drivers, thereby continuously optimizing the control strategy from "general" to "personalized". Second, with the accumulation of training data, the model's prediction accuracy and control effect gradually improve, and the overall energy efficiency and energy recovery efficiency show a continuous improvement trend in actual use, rather than the fixed control parameters in existing technologies that prevent performance improvement. Third, because the training data retains both pre-delivery training data and all historical field data, the model will not forget the basic control capabilities at the time of delivery and the effective strategies learned in the early stages while adapting to new working conditions, avoiding the problem of "catastrophic forgetting". Finally, this online continuous learning mechanism enables different loaders of the same model to develop differentiated optimal control strategies according to their respective actual usage conditions, achieving refined control of "one machine, one strategy", further improving the adaptability and economy of the whole machine in different application scenarios.

[0031] Preferably, in this embodiment, the specific steps of training the multi-agent collaborative optimization model are as follows: the control process of the loader is modeled as an MDP process, and combined with embodied perception characteristics, an embodied closed loop of perception-decision-execution-feedback is realized, wherein the state-space function at time t is... The formula is: The formula for the action space function A is: , The rotational speed of the first motor pump 2, This refers to the rotational speed of the second motor pump 3. The displacement of the hydraulic pump motor in the first electric pump 2 is... The displacement of the hydraulic pump motor in the second motor pump 3; reward function The formula is: Where s represents the state of the control system, and a represents the action of the control system. Let be the transmission efficiency of the control system at time t. Let be the actual recovered energy at time t. Theoretically maximum recoverable energy. To constrain violations and penalties, , , All are weighting coefficients; Based on the reward function at time t The optimal strategy for the MDP process is solved using a near-end policy optimization algorithm. Its formula is: , For the strategy maximization operator, For strategy The expected operator below, Let be the state of the control system at time t. Let t be the action of the system at time t. As a discount factor, For strategy The state distribution under; The policy network and value network are iteratively updated to ultimately output the optimal control variable, as shown in the formula: , Let be the optimal speed of the first motor pump at time t. Let be the optimal speed of the second motor pump at time t. Let be the optimal torque of the walking motor at time t. Let be the optimal displacement of the hydraulic pump motor in the first motor pump at time t. Let t be the optimal displacement of the hydraulic pump motor in the second electric motor pump. Output the optimal strategy given the state and vehicle information; (The recoverable energy of the whole machine is divided into two parts: hydraulic load energy and walking braking energy. The accumulator 11 recovers hydraulic energy, and the power battery recovers walking energy through braking feedback.) Actual recovered energy at time t The formula is: , , ,in, The recoverable energy of the hydraulic negative load at time t comes from the gravity of the boom / bucket cylinder, and the hydraulic oil flows into the accumulator 11 through the distribution valve group 7. The energy is converted from the potential energy of the cylinder. The energy recoverable during driving and braking at time t (from the deceleration / braking of the whole machine, the driving motor 14 enters the power generation state, converting kinetic energy into electrical energy and feeding it back to the power battery). The losses during the energy recovery process, For the time of the downward movement, Let g be the mass of the lowered portion of the bucket, and g be the acceleration due to gravity. The 10 pistons of the bucket cylinder are adjusted to control the travel braking and recover energy to lower the bucket to the specified height. The mass of the lowered boom section. The lowering height of the boom cylinder 12 piston for regenerative braking is [not specified]. To improve the energy recovery efficiency of accumulator 11, Braking time or deceleration time The walking motor 14 generates torque to recover energy for walking braking. The power generation capacity of the walking motor 14 is for regenerative braking energy. Improve battery charging efficiency.

[0032] With the core objective of optimizing the overall system efficiency and with work efficiency and operability as hard constraints, the optimal set of parameters is selected to obtain a well-trained multi-agent collaborative optimization model. The objective function for optimizing the overall system efficiency The formula is: , , Let be the efficiency of the hydraulic components of the working device at time t. Let be the efficiency of the walking component at time t. Let be the efficiency of the power components at time t. Let be the pressure of the i-th hydraulic actuator at time t. Let be the flow rate of the i-th hydraulic actuator at time t. Let be the power output from the battery to the hydraulic components of the working device at time t. Let be the torque of the walking motor at time t. Let be the rotational speed of the walking motor at time t. Let t be the power output from the battery to the walking component at time t; Work efficiency constraint function The formula is: , This represents the theoretical minimum work efficiency threshold. The formula for the controllability constraint function is: , , , For maximum action response delay, To the maximum allowable response latency, For the torque ripple rate of the walking motor 14, For the maximum permissible torque ripple rate, For hydraulic flow rate fluctuation, Maximum allowable flow fluctuation rate; Recyclable energy maximization function The formula is: .

[0033] Specifically, in this embodiment, the control process of the loader is first modeled as a Markov Decision Process (MDP), and combined with embodied perception characteristics to achieve an embodied closed loop of perception-decision-execution-feedback. Embodied perception characteristics refer to the fact that the agent's state perception directly originates from the physical feedback signals of the sensors and actuators actually installed on the loader, rather than relying on theoretical models or simulation data, thus making the model's decision-making process closely coupled with the loader's actual physical state. In this MDP framework, the state-space function at time t includes the pressure parameters of the hydraulic components of the working device, the torque, speed, and vehicle speed parameters of the traveling components, the battery and multi-function controller 23 parameters of the power components, and the overall machine's steering angle, steering angular velocity, bucket cylinder 10 speed, and boom cylinder 12 speed. The motion space vector includes the rotational speed of the first motor pump 2, the rotational speed of the second motor pump 3, the displacement of the hydraulic pump motor in the first motor pump 2, the displacement of the hydraulic pump motor in the second motor pump 3, and the torque of the traveling motor 14. In this system, the rotational speeds of the first electric pump 2 and the second electric pump 3 determine the oil supply capacity of the hydraulic components; the displacement of the hydraulic pump motor determines the proportional relationship between the pump's output flow rate and rotational speed; and the torque of the travel motor 14 directly determines the loader's driving force and braking capability. By explicitly defining the state space and action space mathematically, the MDP process provides a complete description of the decision-making environment for the multi-agent cooperative optimization model.

[0034] At each time step of the MDP process, the model receives corresponding reward feedback based on the current state and the selected action. The design of this reward function reflects the core optimization objectives of this invention: the first term guides the model to maximize the overall system efficiency, the second term guides the model to maximize the actual recovered energy, and the third term penalizes behaviors that violate efficiency or controllability constraints. By adjusting the relative magnitudes of the three weight coefficients, the priority between efficiency, recovery capability, and constraint compliance can be flexibly balanced according to the needs of different application scenarios. It should be noted that the theoretical maximum recoverable energy is a normalized baseline value under the upper limit of the system design, used to scale the actual recovered energy to an order of magnitude close to the overall system efficiency, facilitating gradient propagation during model training.

[0035] Based on the reward function at time t, the model employs the Proximal Policy Optimization (PPO) algorithm to solve for the optimal policy in the MDP process. The PPO algorithm avoids the training instability problem caused by excessively large policy update steps in traditional reinforcement learning algorithms by limiting the policy change magnitude at each policy update. It is particularly suitable for engineering applications requiring high reliability and stability, such as loader control. By iteratively updating the policy network and value network, the model continuously optimizes its decision-making ability, ultimately outputting the optimal control variables. This set of optimal control variables includes the optimal first motor pump speed, the optimal second motor pump speed, the optimal travel motor torque, the optimal hydraulic pump motor displacement in the first motor pump, and the optimal hydraulic pump motor displacement in the second motor pump. These optimal control variables are the specific command values ​​ultimately issued by the vehicle control unit to each executing component.

[0036] In calculating the actual recovered energy, the actual recovered energy at time t consists of two parts: the recoverable energy from the hydraulic load and the recoverable energy from the travel braking, minus the losses during the energy recovery process. The formula for calculating the recoverable energy from the hydraulic load is: the mass of the bucket lowering portion multiplied by the gravitational acceleration and then multiplied by the lowering height of the bucket cylinder 10 piston, plus the mass of the boom lowering portion multiplied by the gravitational acceleration and then multiplied by the lowering height of the boom cylinder 12 piston, multiplied by the recovery efficiency of the accumulator 11, and then divided by the lowering action time. The physical meaning of this formula is that when the boom or bucket descends under gravity, the potential energy that would otherwise be dissipated as heat is captured and stored by the accumulator 11 through the hydraulic components. The recovery efficiency of the accumulator 11 reflects the degree of loss in this energy conversion process. The formula for calculating the recoverable energy from the travel braking is: integrating the power generated by the travel motor 14 during the braking or deceleration time and then multiplying by the battery charging efficiency, where the power generated by the travel motor 14 is equal to the torque generated by the travel motor 14 multiplied by the speed of the travel motor 14. When the driver depresses the brake pedal or releases the accelerator pedal to slow the loader, the travel motor 14 switches from motor mode to generator mode, converting the loader's kinetic energy into electrical energy, which is then fed back to the power battery pack 19 via the multi-function controller 23. The battery charging efficiency reflects the degree of energy loss during the energy feedback process. Through the combined application of the two energy recovery formulas mentioned above, this system can simultaneously recover the load energy of the hydraulic components and the braking energy of the travel components, achieving energy recovery coverage under all working conditions. Compared with existing technologies that can only recover one type of energy or cannot recover any at all, this significantly improves the overall energy utilization efficiency of the machine.

[0037] During the training of the multi-agent collaborative optimization model, the model takes the optimization of the overall system efficiency as its core objective, with work efficiency and operability as hard constraints, and selects the optimal set of parameters to obtain the trained multi-agent collaborative optimization model. The product of the efficiency of the hydraulic components of the working device, the efficiency of the traveling components, and the efficiency of the power components is used as the overall evaluation index of the overall system efficiency, reflecting the technical approach of this invention to optimize at the system level rather than the component level. The operability constraint function includes three sub-constraints: the actual action response delay does not exceed the maximum permissible action response delay, the torque fluctuation rate of the traveling motor 14 does not exceed the maximum permissible torque fluctuation rate, and the hydraulic flow fluctuation rate does not exceed the maximum permissible flow fluctuation rate. These three sub-constraints correspond to the three operability dimensions of operational timeliness, driving smoothness, and working device stability, respectively, and together constitute a comprehensive guarantee for the dynamic response quality of the system. The recoverable energy maximization function guides the model to maximize the recoverable energy potential under the current working conditions while satisfying the work efficiency and operability constraints. Through the above-mentioned multi-objective and multi-constraint optimization framework, the multi-agent cooperative optimization model trained by this invention can achieve the optimal balance between efficiency, recovery capability, lower limit of working efficiency and upper limit of operability, realizing high-performance cooperative control of electric loader drive system under complex working conditions.

[0038] In summary, this invention addresses the technical problems of existing electric loaders, such as simply replacing the engine with an electric motor, low drive system efficiency, inability to recover energy under negative loads, and static control strategies that cannot be adaptively optimized. It constructs a closed-loop control architecture comprising a hydraulic component for the working device, a traveling component, a power component, and a vehicle control unit. The hydraulic component for the working device uses a dual-motor pump for oil supply, which is delivered to the large-displacement steering gear 8 and the distribution valve group 7 via a flow priority valve 6. The distribution valve group 7 connects to the bucket cylinder 10, the boom cylinder 12, and the accumulator 11, achieving steering priority control and hydraulic potential energy recovery. The traveling component achieves electric drive through a traveling motor controller 13, a traveling motor 14, a multi-speed gearbox 15, and a drive axle. The power component provides power management for the entire vehicle through a power battery pack 19, a power battery energy management module 21, a power battery thermal management module 22, and a multi-function controller 23. The vehicle control unit achieves data and electrical interconnection with various components, and executes the following steps by running an internally stored computer program: acquiring the vehicle state information and control signals at time t; determining the current operating state based on an operating state recognition model that includes a multi-head self-attention mechanism; calculating the theoretical required flow rate of the hydraulic components based on the control signals and operating state, combined with the flow diversion characteristics of the flow priority valve 6, and calculating the theoretical required torque of the motor based on the speed ratio characteristics of the multi-gear transmission 15; using a trained multi-agent cooperative optimization model to perform control preprocessing on the theoretical requirements, obtaining the vehicle state information and control signals at time t+1; saving the data at time t+1 to the training dataset, and retraining the model based on the updated training dataset for control preprocessing at the next time step. The training process of the multi-agent cooperative optimization model models the loader control process as a Markov decision process, employing a proximal policy optimization algorithm to solve for the optimal strategy, with optimal overall system efficiency as the core objective, and work efficiency and maneuverability as hard constraints. A reward function guides the model to simultaneously optimize overall system efficiency, actual energy recovery, and penalize behaviors that violate constraints.

[0039] Compared with existing technologies, this application has the following advantages: By using dual-motor pump oil supply and flow priority valve 6 for flow diversion control, combined with the introduction of accumulator 11, it achieves steering priority protection and effective recovery of hydraulic negative load energy, solving the problem that existing systems cannot recover potential energy under working conditions such as boom descent; By theoretically calculating and dynamically controlling the braking torque of the travel motor 14, it achieves feedback recovery of travel braking energy, complementing hydraulic energy recovery and covering the energy recovery needs of the loader under all working conditions; By using an operation state recognition model that includes a multi-head self-attention mechanism, it can automatically extract the correlation features between different signals from multi-source sensor signals, achieving accurate recognition of the loader's digging, lifting, unloading, and returning operation states, providing a reliable basis for refined control; By using a multi-agent collaborative optimization model to preprocess the theoretical requirements, it treats the hydraulic components, travel components, and power components as collaborative intelligent systems. The system solves for the optimal control variables under the objective of optimal system efficiency, achieving collaborative optimization among multiple subsystems and avoiding energy efficiency losses caused by independent control of each subsystem in existing technologies. By saving the vehicle state information and control signals at each moment to the training dataset and continuously retraining the model based on the updated dataset, the model can continuously evolve with the accumulation of actual operation data, realizing the transformation of the control strategy from "factory preset" to "continuous adaptive optimization". The training data retains both pre-factory data and all historical field data, effectively avoiding catastrophic forgetting problems. Through the multi-objective design of the reward function, which includes the overall system efficiency, actual recovered energy, and constraint violation penalty terms, combined with the lower limit constraint of working efficiency and the upper limit constraint of maneuverability (action response delay, torque fluctuation rate, flow fluctuation rate), the system improves energy efficiency and recovery capabilities while ensuring that the basic operating capabilities and driving comfort of the loader are not affected.

[0040] In summary, this invention effectively improves the overall energy efficiency, energy recovery capability, and working condition adaptability of electric loaders, demonstrating significant technological advancement and engineering application value.

[0041] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A drive control system for an electric loader, characterized in that, include: The system includes a hydraulic component for the working device, a traveling component, a power component, and a vehicle control unit. The data terminal of the vehicle control unit is electrically connected to the data terminals of the hydraulic component for the working device, the traveling component, and the power component. The power component is also electrically connected to the hydraulic component for the working device and the traveling component. The vehicle control unit is configured to perform the following steps by executing a computer program stored internally thereon: Obtain the overall vehicle status information and control signals of the loader at time t, and determine the working status of the loader at time t based on the overall vehicle status information at time t; Based on the control signal and operating status at time t, calculate the theoretical required flow rate of the hydraulic components of the working device and the theoretical required torque of the motor of the traveling component at time t. Using a trained multi-agent collaborative optimization model, the theoretical demand flow and the theoretical demand torque of the motor at time t are preprocessed to obtain the vehicle state information and control signal at time t+1. The vehicle state information and control signals at time t+1 are saved to the training dataset. Based on the updated training dataset, the multi-agent cooperative optimization model is retrained to obtain a newly trained multi-agent cooperative optimization model, which is used for control preprocessing at the next time step. Vehicle status information at time t The formula is: , The characteristics of the hydraulic components of the working device at time t. The characteristics of the walking component at time t, The characteristics of the power component at time t, The characteristics of the vehicle control unit at time t are as follows: For transpose; Control signal at time t The formula is: , For brake pedal opening, This refers to the accelerator pedal opening. For boom handle signal, For the bucket handle signal, For steering wheel signals; in, , The outlet pressure of the first motor pump. The outlet pressure of the second motor pump. The first outlet pressure of the flow priority valve. The second outlet pressure of the flow priority valve. The pressure at the return port. The pressure of the left steering cylinder. This refers to the pressure of the right steering cylinder. This refers to the pressure in the large chamber of the bucket cylinder. The pressure in the small chamber of the bucket cylinder. For accumulator pressure, This refers to the pressure in the large chamber of the boom cylinder. This refers to the pressure in the small chamber of the boom cylinder; , This refers to the torque of the walking motor. v is the rotational speed of the drive motor, v is the vehicle speed, and a is the vehicle acceleration. For multi-gear transmissions; , For battery power, Battery voltage, Battery current, For battery cell temperature, For the power of the all-in-one controller, For the voltage of the all-in-one controller, For the current of the all-in-one controller, Temperature for the all-in-one controller; , For steering angle, For the steering angular velocity, For the speed of the bucket cylinder, The speed of the boom cylinder.

2. The electric loader drive control system according to claim 1, characterized in that, The hydraulic components of the working device include: a first electric pump, a second electric pump, a first relief valve, a second relief valve, a flow priority valve, a distribution valve group, a large-displacement steering gear, a steering cylinder, a bucket cylinder, an accumulator, and a boom cylinder. The outlet of the first electric pump is connected to the inlet of the first relief valve, the outlet of the first relief valve is connected to the inlet of the flow priority valve, the first outlet of the flow priority valve is connected to the inlet of the large-displacement steering gear, the outlet of the second electric pump is connected to the inlet of the second relief valve, the outlet of the second relief valve and the distribution valve group are connected to the second outlet of the flow priority valve, the large-displacement steering gear is connected to the steering cylinder, and the distribution valve group is connected to the bucket cylinder, boom cylinder, and accumulator. The data terminal of the vehicle control unit is electrically connected to the first electric pump, the second electric pump, the flow priority valve, and the distribution valve group.

3. The electric loader drive control system according to claim 2, characterized in that, The walking assembly includes a walking motor controller, a walking motor, a multi-speed gearbox, a drive shaft, and a drive axle. The power assembly includes a power battery pack, a charging socket, a power battery energy management module, a power battery thermal management module, and an all-in-one controller. The data terminal of the vehicle control unit is electrically connected to the data terminals of the power battery thermal management module, the power battery energy management module, the all-in-one controller, the walking motor controller, and the multi-speed gearbox.

4. The electric loader drive control system according to claim 3, characterized in that, The formula for determining the work status is: , Let t be the job status at time t. For job status identification mapping function, For trainable parameters, For normalized exponential functions, It is a multilayer perceptron. The encoder layer consists of a multi-head self-attention network and a feedforward network. The formula for the multi-head self-attention mechanism is as follows: , , This is a single-head self-attention output, where Q is the query matrix, K is the key matrix, and V is the value matrix. Scaling factor For multi-head self-attention output, For splicing operations, The first self-focused head, Let h be the h-th self-attention head, where h is the number of self-attention heads. Let be the transformation matrix.

5. The electric loader drive control system according to claim 4, characterized in that, Based on the control signal and operating status at time t, the theoretical required flow rate of the hydraulic components of the working device and the theoretical required torque of the motor of the traveling component at time t are calculated respectively, as follows: Based on the control signal and operating status at time t, and combined with the flow priority valve's diversion characteristics, the theoretical flow rate of the steering cylinder at time t is calculated. Theoretical flow rate of the working cylinder Its formula is: , , , , ; in, For steering gear displacement, For steering speed, This is the hydraulic leakage compensation coefficient. The job status at time t The corresponding flow coefficient of the steering cylinder that the entire vehicle can pass through. The effective working area of ​​the bucket cylinder. Let t be the theoretical speed of the bucket cylinder. The effective working area of ​​the boom cylinder. Let t be the theoretical speed of the boom cylinder. The job status at time t The corresponding hydraulic cylinder flow coefficient that the entire vehicle can pass through. The steering speed constant is Let be the speed constant of the bucket cylinder. The speed constant of the boom cylinder; Calculate the theoretical motor drive torque at time t based on the control signal and operating status at time t. and theoretical motor braking torque Its formula is: , ,in, Main reduction ratio, Let be the real-time gear ratio of the transmission at time t. The total transmission efficiency of the drive axle. This is the theoretical motor drive torque constant. This is the theoretical braking torque constant of the motor. The job status at time t The corresponding torque coefficient of the walking motor that the whole vehicle can provide.

6. The electric loader drive control system according to claim 1, characterized in that, when At this time, the multi-agent cooperative optimization model is pre-trained using a training dataset, and the data in the training dataset only includes pre-shipment training data; when At that time, the data in the training dataset includes pre-factory training data as well as all vehicle status information and control signals before time t.

7. The electric loader drive control system according to claim 5, characterized in that, The specific steps for training the multi-agent collaborative optimization model are as follows: The control process of the loader is modeled as an MDP process, and combined with embodied perception characteristics, an embodied closed loop of perception-decision-execution-feedback is achieved, where the state-space function at time t is... The formula is: The formula for the action space function A is: , The rotational speed of the first motor pump. The rotational speed of the second motor pump. This refers to the displacement of the hydraulic pump motor in the first electric pump. This refers to the displacement of the hydraulic pump motor in the second electric pump. reward function The formula is: Where s represents the state of the control system, and a represents the action of the control system. Let be the transmission efficiency of the control system at time t. Let be the actual recovered energy at time t. Theoretically maximum recoverable energy. To constrain violations and penalties, , , All are weighting coefficients; Based on the reward function at time t The optimal strategy for the MDP process is solved using a near-end policy optimization algorithm. Its formula is: , For the strategy maximization operator, For strategy The expected operator below, Let be the state of the control system at time t. Let t be the action of the system at time t. As a discount factor, For strategy The state distribution under; The policy network and value network are iteratively updated to ultimately output the optimal control variable, as shown in the formula: , Let be the optimal speed of the first motor pump at time t. Let be the optimal speed of the second motor pump at time t. Let be the optimal torque of the walking motor at time t. Let be the optimal displacement of the hydraulic pump motor in the first motor pump at time t. Let t be the optimal displacement of the hydraulic pump motor in the second electric motor pump. Output the optimal strategy given the state and vehicle information; Actual recovered energy at time t The formula is: , , ,in, The recoverable energy under hydraulic negative load at time t. The energy that can be recovered during driving braking at time t. The losses during the energy recovery process, For the time of the downward movement, Let g be the mass of the lowered portion of the bucket, and g be the acceleration due to gravity. Let t be the height at which the bucket cylinder piston descends. The mass of the lowered boom section. Let t be the lowering height of the boom cylinder piston. To improve the energy recovery efficiency of the accumulator, Braking time or deceleration time Let be the torque generated by the walking motor at time t. Let be the power output of the walking motor at time t. Improve battery charging efficiency.

8. The electric loader drive control system according to claim 7, characterized in that, Also includes: With the core objective of optimizing the overall system efficiency and with work efficiency and operability as hard constraints, the optimal set of parameters is selected to obtain a well-trained multi-agent collaborative optimization model. The objective function for optimizing the overall system efficiency The formula is: , , Let be the efficiency of the hydraulic components of the working device at time t. Let be the efficiency of the walking component at time t. Let be the efficiency of the power components at time t. Let be the pressure of the i-th hydraulic actuator at time t. Let be the flow rate of the i-th hydraulic actuator at time t. Let be the power output from the battery to the hydraulic components of the working device at time t. Let be the torque of the walking motor at time t. Let be the rotational speed of the walking motor at time t. Let t be the power output from the battery to the walking component at time t; Work efficiency constraint function The formula is: , This represents the theoretical minimum work efficiency threshold. The formula for the controllability constraint function is: , , , For maximum action response delay, To the maximum allowable response latency, For the torque ripple rate of the walking motor, For the maximum permissible torque ripple rate, For hydraulic flow rate fluctuation, Maximum allowable flow fluctuation rate; Recyclable energy maximization function The formula is: .

Images