Energy management method for hybrid power self-propelled plant protection machine

Abstract

The invention discloses an energy management method for a hybrid power self-propelled plant protection machine, and the method comprises the following steps: S1, building a workshop longitudinal dynamic model, the hybrid power self-propelled plant protection machine energy management system topological structure model comprises three energy sources of a fuel cell, a lithium battery and a super capacitor; the physical model comprises the fuel cell, the lithium battery and the super capacitor; s2, designing a plant protection machine operation condition identification method based on a BP neural network, identifying the operation condition type of the plant protection machine, and calculating the total demand power under the corresponding condition; s3, constructing a hierarchical energy management framework based on a fuzzy filter and a deep reinforcement learning algorithm, and performing frequency division processing and multi-energy source power distribution on the total demand power, according to the hybrid power self-propelled plant protection machine energy management method, a more reasonable energy distribution method can be decided according to the power requirements of different working conditions and the characteristics of different energy sources, and then the overall efficiency and the endurance mileage of the plant protection machine during operation are improved.

Description

Technical Field

[0001] This invention relates to the field of hybrid self-propelled plant protection machinery technology, specifically to an energy management method for hybrid self-propelled plant protection machinery. Background Technology

[0002] With the rapid development of agricultural modernization and new energy technologies, self-propelled plant protection machines equipped with new energy hybrid power are gradually becoming a new direction for development in the agricultural plant protection field due to their advantages such as high efficiency, zero emissions, and reliable power. The effectiveness of the energy management strategy of new energy plant protection machines largely determines the overall performance and energy utilization efficiency of the agricultural machinery.

[0003] Currently, domestic research on energy management methods for new energy self-propelled plant protection drones mostly focuses on single energy sources, but these drones suffer from the fatal problem of short driving range. In contrast, hybrid self-propelled plant protection drones significantly extend their driving range through the coordinated operation of multiple power sources, but research on their energy management methods is relatively lacking. Therefore, how to provide a more efficient energy management method for hybrid self-propelled plant protection drones has become an urgent technical challenge to be solved in this field. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the existing defects and provide an energy management method for a hybrid self-propelled plant protection machine. This method can determine a more reasonable energy allocation method based on the power requirements of different working conditions and the characteristics of different energy sources, thereby improving the overall efficiency and range of the plant protection machine during operation and effectively solving the problems in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a hybrid self-propelled plant protection machine energy management method, comprising the following steps:

[0006] S1: Establish a longitudinal dynamic model of the workshop, including the topological structure model of the energy management system of the hybrid self-propelled plant protection machine with three energy sources: fuel cell, lithium battery and supercapacitor, as well as the physical models of fuel cell, lithium battery and supercapacitor.

[0007] S2: Design a method for identifying the operating conditions of plant protection machines based on BP neural networks, identify the operating condition types of plant protection machines and calculate the total power demand under the corresponding operating conditions;

[0008] S3: Construct a hierarchical energy management framework based on fuzzy filters and deep reinforcement learning algorithms to perform frequency division processing and multi-energy source power allocation on the total required power, thereby realizing real-time energy management of the plant protection machine.

[0009] As a preferred embodiment of the present invention, the longitudinal dynamic model of the workshop in step S1 is as follows:

[0010]

[0011] in, Indicates the first Each sampling time, It is system output. It is the identity matrix. , and The coefficient matrix in the discrete state-space equations of the system is shown below:

[0012]

[0013] in, It is the system's control cycle.

[0014] As a preferred embodiment of the present invention, the topology model of the energy management system for a hybrid self-propelled plant protection machine, which includes three energy sources—fuel cell, lithium battery, and supercapacitor—in step S1 is as follows:

[0015]

[0016] in, These are the total power demand of the agricultural drone, the power of the fuel cell, the power of the lithium battery, and the power of the supercapacitor; during charging, and All are negative; during discharge, and It is positive.

[0017] As a preferred embodiment of the present invention, the physical models of the fuel cell, lithium battery, and supercapacitor in step S1 include constructing voltage models of the fuel cell, lithium battery, and supercapacitor, wherein the voltage models are as follows:

[0018]

[0019]

[0020]

[0021]

[0022]

[0023]

[0024] in, It refers to the number of cells on the fuel cell stack. It is the electromotive force of the fuel cell. This is the loss during fuel cell activation. It is the ohmic loss of the fuel cell. It is the temperature of the catalyst layer. It is the temperature correction offset of the catalyst layer. It is the universal gas constant. It is Faraday's constant. It is the pressure at the interface of the anode catalyst layer. It is the pressure at the interface of the cathode catalyst layer. It is the open-circuit voltage of the lithium battery. It is the output current of the lithium battery. It is the internal resistance of the lithium battery. It is the SoC in the current state of the lithium battery. It is the initial SoC for lithium batteries. This refers to the nominal capacity of the lithium battery. It is the charge / discharge switch for lithium batteries (positive for charging, negative for discharging). yes The output current of the lithium battery at all times. It is the open-circuit voltage of the supercapacitor. It is the output current of the supercapacitor. It is the internal resistance of the supercapacitor. It is the initial SoC for supercapacitors. This is the maximum charge capacity of a supercapacitor. This is the initial charge of the supercapacitor. yes The instantaneous voltage of the supercapacitor at any given moment.

[0025] As a preferred embodiment of the present invention, the specific process of step S2 is as follows:

[0026] (1) Determine the characteristic parameters of the plant protection machine's operating conditions, including speed, acceleration, load, battery state (SOC), nozzle pressure, and steering angle. Collect the characteristic parameters in the actual operating environment through sensors as operating condition samples.

[0027] (2) Construct a BP neural network model to divide the plant protection machine operation conditions into uniform spraying operation conditions, field turning operation conditions and boundary turning operation conditions. The BP neural network model outputs the recognition results as follows: uniform spraying operation condition outputs 1, field turning operation condition outputs 2, and boundary turning operation condition outputs 3.

[0028] (3) Based on the identified operating conditions, calculate the total power demand under different operating conditions. The power required for the spraying system is:

[0029]

[0030] in, For the spray flow rate, The pressure difference of the nozzle;

[0031] The total power requirement of the plant protection machine is:

[0032]

[0033] in, For air resistance, For rolling resistance, It is the slope resistance. To increase resistance, This is for steering resistance.

[0034] As a preferred embodiment of the present invention, the specific process in step S3 is as follows:

[0035] (1) By using the integral method and the formula Calculate the weighted SoC of lithium battery and supercapacitor;

[0036] (2) Input the obtained power requirement of the plant protection machine and the weighted SoC into the fuzzy filter, and finally obtain the required adjustment frequency after fuzzy processing. ;

[0037] Input demand power The fuzzy sets represent the power ranges as {negative large region (NB), negative medium region (NM), negative small region (NS), zero region (ZE), positive small region (PS), positive medium region (PM), positive large region (PB)}, and the weighted SoC fuzzy set represents the ranges as {small region (S), relatively small region (RS), medium region (M), relatively large region (RB), large region (B)}; output adjustment frequency The fuzzy sets represent the ranges of {small region (S), relatively small region (RS), medium region (M), relatively large region (RB), and large region (B)} respectively.

[0038] (3) The obtained adjustment frequency and required power The input is fed into an adaptive low-pass filter, where the required power is divided by frequency and allocated appropriately.

[0039] (4) Define the state space for deep reinforcement learning: set the width to be... The time window allows the system to acquire a continuous data sequence containing various information such as acceleration, deceleration, and constant speed within each detection cycle by continuously collecting driving data within a specific time range. The sampling sequence at time t is shown below:

[0040]

[0041] in, yes The status of the selected agricultural machinery vehicle at any given time. yes The status sequence of the agricultural drone at any given time;

[0042] (5) Design the reward function for deep reinforcement learning:

[0043]

[0044]

[0045] in, It represents the total hydrogen consumption at time t during the operation of the agricultural machinery. The hydrogen consumption of the fuel cell during the operation of the agricultural drone at time t. and These represent the equivalent hydrogen consumed by the supercapacitor and the lithium battery at time t, respectively. The losses are due to the frequent power fluctuations in fuel cells and lithium batteries. It is a penalty factor for fuel cells. and These are the equivalent factors for supercapacitors and lithium batteries, respectively.

[0046] (6) The designed reward function is incorporated into the deep reinforcement learning algorithm. Then, the deep reinforcement learning algorithm is trained with real working condition data to obtain the optimal energy management strategy for the plant protection machine. Finally, the trained deep reinforcement learning algorithm is used to perform real-time energy management on the hybrid self-propelled plant protection machine.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows: This hybrid self-propelled plant protection machine energy management method addresses the differences in power demand of the plant protection machine under different operating conditions by using three energy sources to provide power output in a coordinated manner; according to the physical characteristics of each energy source, the power is split, with the supercapacitor responsible for high-frequency power, and the fuel cell and lithium battery responsible for medium and low-frequency power, and the maximum charging and discharging current of the lithium battery is constrained to maximize the service life of the energy sources, resulting in better overall performance and longer driving range for the hybrid plant protection machine during operation, and strong universality and real-time performance for different operating conditions. Attached Figure Description

[0048] Figure 1 This is a flowchart of the present invention;

[0049] Figure 2 This is a schematic diagram of the energy source power output distribution of the present invention. Detailed Implementation

[0050] 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.

[0051] Please see Figure 1-2 This invention provides a technical solution: an energy management method for a hybrid self-propelled plant protection machine, comprising the following steps:

[0052] S1: Establish a longitudinal dynamic model of the workshop, including the topological structure model of the energy management system of the hybrid self-propelled plant protection machine with three energy sources: fuel cell, lithium battery, and supercapacitor, as well as the physical models of fuel cell, lithium battery, and supercapacitor:

[0053] (1) Constructing a longitudinal dynamic model of the workshop:

[0054]

[0055] in, Indicates the first Each sampling time, It is system output. It is the identity matrix. , and The coefficient matrix in the discrete state-space equations of the system is shown below:

[0056]

[0057] in, It is the system's control cycle;

[0058] (2) The topology model of the energy management system of the hybrid self-propelled plant protection machine, which includes three energy sources: fuel cell, lithium battery and supercapacitor, is as follows:

[0059]

[0060] in, These are the total power demand of the agricultural drone, the power of the fuel cell, the power of the lithium battery, and the power of the supercapacitor; during charging, and All are negative; during discharge, and It is positive;

[0061] (3) Construct voltage models for fuel cells, lithium batteries, and supercapacitors. The voltage models are as follows:

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068] in, It refers to the number of cells on the fuel cell stack. It is the electromotive force of the fuel cell. This is the loss during fuel cell activation. It is the ohmic loss of the fuel cell. It is the temperature of the catalyst layer. It is the temperature correction offset of the catalyst layer. It is the universal gas constant. It is Faraday's constant. It is the pressure at the interface of the anode catalyst layer. It is the pressure at the interface of the cathode catalyst layer. It is the open-circuit voltage of the lithium battery. It is the output current of the lithium battery. It is the internal resistance of the lithium battery. It is the SoC in the current state of the lithium battery. It is the initial SoC for lithium batteries. This refers to the nominal capacity of the lithium battery. It is the charge / discharge switch for lithium batteries (positive for charging, negative for discharging). yes The output current of the lithium battery at all times. It is the open-circuit voltage of the supercapacitor. It is the output current of the supercapacitor. It is the internal resistance of the supercapacitor. It is the initial SoC for supercapacitors. This is the maximum charge capacity of a supercapacitor. This is the initial charge of the supercapacitor. yes The instantaneous voltage of the supercapacitor at any given moment;

[0069] S2: Design a method for identifying the operating conditions of agricultural machinery based on a BP neural network, which identifies the operating condition type of the agricultural machinery and calculates the total power demand under the corresponding condition. The process is as follows:

[0070] (1) Obtaining working conditions

[0071] Determine the characteristic parameters of the plant protection machine's operating conditions, including speed, acceleration, load (such as spray volume), battery state of charge (SOC), nozzle pressure, and steering angle. Then, in the actual operating environment, use sensors to collect field working data of the plant protection machine containing these characteristic parameters as the operating conditions.

[0072] (2) Operating condition identification based on BP neural network

[0073] The operating conditions of plant protection machines are classified as follows: 1. Uniform speed spraying operation, 2. Field turning operation, 3. Boundary turning operation.

[0074] Construct a BP neural network model for working condition identification, classify the working conditions of the plant protection machine, output 1 if the identification result is uniform spraying, output 2 if the identification result is field turning, and output 3 if the identification result is boundary turning.

[0075] (3) Calculation of demand power prediction based on different operating conditions

[0076] Based on the identified operating condition types, calculate the total power demand under different operating conditions. In different operating conditions, the vehicle driving state of the plant protection machine is different, so the changes in its various resistance values ​​are also different.

[0077] Specifically, the power requirement for the spraying system is:

[0078]

[0079] in, For the spray flow rate, The pressure difference of the nozzle;

[0080] The total power requirement of the plant protection machine is:

[0081]

[0082] in, For air resistance, For rolling resistance, It is the slope resistance. To increase resistance, For steering resistance;

[0083] S3: Construct a hierarchical energy management framework based on fuzzy filters and deep reinforcement learning algorithms, perform frequency division processing on the total power demand and allocate power from multiple energy sources, and realize real-time energy management of the plant protection machine. The specific process is as follows:

[0084] (1) By using the integral method and the formula Calculate the weighted SoC of lithium battery and supercapacitor;

[0085] (2) Input the obtained power requirement of the plant protection machine and the weighted SoC into the fuzzy filter, and finally obtain the required adjustment frequency after fuzzy processing. ;

[0086] Input demand power The fuzzy sets represent the power ranges as {negative large region (NB), negative medium region (NM), negative small region (NS), zero region (ZE), positive small region (PS), positive medium region (PM), positive large region (PB)}, and the weighted SoC fuzzy set represents the ranges as {small region (S), relatively small region (RS), medium region (M), relatively large region (RB), large region (B)}; output adjustment frequency The fuzzy sets represent the ranges of {small region (S), relatively small region (RS), medium region (M), relatively large region (RB), and large region (B)} respectively.

[0087] (3) The obtained adjustment frequency and required power The input is fed into an adaptive low-pass filter, where the required power is divided by frequency and allocated appropriately.

[0088] Among them, processes (1) to (3) are the first distinction of the required power, and the high-frequency power in the required positive power is provided by the supercapacitor. This not only makes effective and reasonable use of the physical characteristics of the supercapacitor, but also greatly reduces the life loss of fuel cells and lithium batteries.

[0089] The allocation of low- and medium-frequency power needs to be optimized by using the SoC of the energy source storage system and the overall hydrogen consumption of the vehicle as the reward function of the deep reinforcement learning algorithm.

[0090] (4) Define the state space of deep reinforcement learning

[0091] Set the width to The time window allows the system to acquire a continuous data sequence containing various information such as acceleration, deceleration, and constant speed within each detection cycle by continuously collecting driving data within a specific time range. The sampling sequence at time t is shown below:

[0092]

[0093] in, yes The status of the selected agricultural machinery vehicle at any given time. yes The status sequence of the agricultural drone at any given time;

[0094] (5) Incorporate multi-objective optimization problems into the reward function of deep reinforcement learning algorithms:

[0095]

[0096]

[0097] in, It represents the total hydrogen consumption at time t during the operation of the agricultural machinery. The hydrogen consumption of the fuel cell during the operation of the agricultural drone at time t. and These represent the equivalent hydrogen consumed by the supercapacitor and the lithium battery at time t, respectively. The losses are due to the frequent power fluctuations in fuel cells and lithium batteries. It is a penalty factor for fuel cells; and These are the equivalent factors for supercapacitors and lithium batteries, respectively.

[0098] (6) The designed reward function is incorporated into the deep reinforcement learning algorithm. Then, the deep reinforcement learning algorithm is trained with real working condition data to obtain the optimal energy management strategy for the plant protection machine. Finally, the trained deep reinforcement learning algorithm is used to perform real-time energy management on the hybrid self-propelled plant protection machine.

[0099] The above method addresses the varying power demands of agricultural machinery under different operating conditions by employing three energy sources—fuel cells, lithium batteries, and supercapacitors—to provide power output in synergy. Based on the physical characteristics of each energy source, power is distributed: the supercapacitor handles high-frequency power, while the fuel cell and lithium battery handle mid-to-low-frequency power. Furthermore, the maximum charge / discharge current of the lithium battery is constrained to maximize the lifespan of the energy sources, thereby extending the agricultural machinery's operating range and significantly improving overall operational efficiency.

[0100] The parts of the invention not described in detail are prior art. 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 variations 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 energy management of a hybrid self-propelled plant protection machine, characterized in that: Includes the following steps: S1: Establish a longitudinal dynamic model of the workshop, including the topological structure model of the energy management system of the hybrid self-propelled plant protection machine with three energy sources: fuel cell, lithium battery and supercapacitor, as well as the physical models of fuel cell, lithium battery and supercapacitor. S2: Design a method for identifying the operating conditions of plant protection machines based on BP neural networks, identify the operating condition types of plant protection machines and calculate the total power demand under the corresponding operating conditions; S3: Construct a hierarchical energy management framework based on fuzzy filters and deep reinforcement learning algorithms to perform frequency division processing and multi-energy source power allocation on the total required power, thereby realizing real-time energy management of the plant protection machine.

2. The energy management method for a hybrid self-propelled plant protection machine according to claim 1, characterized in that: The longitudinal dynamic model of the workshop in step S1 is as follows: in, Indicates the first Each sampling time, It is system output. It is the identity matrix. , and The coefficient matrix in the discrete state-space equations of the system is shown below: in, It is the system's control cycle.

3. The energy management method for a hybrid self-propelled plant protection machine according to claim 1, characterized in that: The topology model of the energy management system for the hybrid self-propelled agricultural drone, which includes three energy sources—fuel cells, lithium batteries, and supercapacitors—in step S1 is as follows: in, These are the total power demand of the agricultural drone, the power of the fuel cell, the power of the lithium battery, and the power of the supercapacitor; during charging, and All are negative; during discharge, and It is positive.

4. The energy management method for a hybrid self-propelled plant protection machine according to claim 1, characterized in that: The physical models of the fuel cell, lithium battery, and supercapacitor in step S1 include constructing voltage models for the fuel cell, lithium battery, and supercapacitor. The voltage models are as follows: in, It refers to the number of cells on the fuel cell stack. It is the electromotive force of the fuel cell. This is the loss during fuel cell activation. It is the ohmic loss of the fuel cell. It is the temperature of the catalyst layer. It is the temperature correction offset of the catalyst layer. It is the universal gas constant. It is Faraday's constant. It is the pressure at the interface of the anode catalyst layer. It is the pressure at the interface of the cathode catalyst layer. It is the open-circuit voltage of the lithium battery. It is the output current of the lithium battery. It is the internal resistance of the lithium battery. It is the SoC in the current state of the lithium battery. It is the initial SoC for lithium batteries. This refers to the nominal capacity of the lithium battery. It is the charge / discharge switch for lithium batteries (positive for charging, negative for discharging). yes The output current of the lithium battery at all times. It is the open-circuit voltage of the supercapacitor. It is the output current of the supercapacitor. It is the internal resistance of the supercapacitor. It is the initial SoC for supercapacitors. This is the maximum charge capacity of a supercapacitor. This is the initial charge of the supercapacitor. yes The instantaneous voltage of the supercapacitor at any given moment.

5. The energy management method for a hybrid self-propelled plant protection machine according to claim 1, characterized in that: The specific process of step S2 is as follows: (1) Determine the characteristic parameters of the plant protection machine's operating conditions, including speed, acceleration, load, battery state (SOC), nozzle pressure, and steering angle. Collect the characteristic parameters in the actual operating environment through sensors as operating condition samples. (2) Construct a BP neural network model to divide the plant protection machine operation conditions into uniform spraying operation conditions, field turning operation conditions and boundary turning operation conditions. The BP neural network model outputs the recognition results as follows: uniform spraying operation condition outputs 1, field turning operation condition outputs 2, and boundary turning operation condition outputs 3. (3) Based on the identified operating conditions, calculate the total power demand under different operating conditions. The power required for the spraying system is: in, For the spray flow rate, The pressure difference of the nozzle; The total power requirement of the plant protection machine is: in, For air resistance, For rolling resistance, It is the slope resistance. To increase resistance, This is for steering resistance.

6. The energy management method for a hybrid self-propelled plant protection machine according to claim 1, characterized in that: The specific process in step S3 is as follows: (1) By using the integral method and the formula Calculate the weighted SoC of lithium battery and supercapacitor; (2) Input the obtained power requirement of the plant protection machine and the weighted SoC into the fuzzy filter, and finally obtain the required adjustment frequency after fuzzy processing. ; Input demand power The fuzzy sets represent the power ranges as {negative large region (NB), negative medium region (NM), negative small region (NS), zero region (ZE), positive small region (PS), positive medium region (PM), positive large region (PB)}, and the weighted SoC fuzzy set represents the ranges as {small region (S), relatively small region (RS), medium region (M), relatively large region (RB), large region (B)}; output adjustment frequency The fuzzy sets represent the ranges of {small region (S), relatively small region (RS), medium region (M), relatively large region (RB), and large region (B)} respectively. (3) The obtained adjustment frequency and required power The input is fed into an adaptive low-pass filter, where the required power is divided by frequency and allocated appropriately. (4) Define the state space of deep reinforcement learning: Set a time window with a width of t, and continuously collect driving data within a specific time range so that the system can obtain a continuous data sequence containing various information such as acceleration, deceleration, and constant speed in each detection cycle. The sampling sequence at time t is shown below: in, yes The status of the selected agricultural machinery vehicle at any given time. yes The status sequence of the agricultural drone at any given time; (5) Design the reward function for deep reinforcement learning: in, It represents the total hydrogen consumption at time t during the operation of the agricultural machinery. The hydrogen consumption of the fuel cell during the operation of the agricultural drone at time t. and These represent the equivalent hydrogen consumed by the supercapacitor and the lithium battery at time t, respectively. The losses are due to the frequent power fluctuations in fuel cells and lithium batteries. It is a penalty factor for fuel cells. and These are the equivalent factors for supercapacitors and lithium batteries, respectively. (6) The designed reward function is incorporated into the deep reinforcement learning algorithm. Then, the deep reinforcement learning algorithm is trained with real working condition data to obtain the optimal energy management strategy for the plant protection machine. Finally, the trained deep reinforcement learning algorithm is used to perform real-time energy management on the hybrid self-propelled plant protection machine.

Images