Energy management method for fuel cell loader

Abstract

The invention belongs to the technical field of hybrid power system energy management, and relates to a fuel cell loader energy management method, which comprises the following steps: step S10, constructing a cyclic working condition database; step S20, constructing an original characteristic matrix X representing the cycle working condition of the fuel cell loader; step S30, performing dimension reduction processing on the original feature matrix # imgabs0 # by using a crossover probability algorithm to obtain a working condition feature matrix # imgabs1 # after dimension reduction; step S40, establishing a working condition identification model of the fuel cell loader; step S50, establishing a joint network communication prediction model according to a working condition identification result of the working condition identification model; step S60, during online application, obtaining a future demand power sequence in a prediction time domain; and S70, realizing environment adaptive optimal distribution of the output power of the fuel cell and the power cell. The method has the advantages that the energy management performance and the system reliability of the fuel cell loader under different environmental conditions are remarkably improved while the efficient operation of equipment is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of hybrid power system energy management, and relates to an energy management method for a fuel cell loader based on hierarchical control and environmental adaptation. Background Art

[0002] Traditional diesel loaders suffer from high energy consumption and emissions, making hydrogen fuel cells an increasingly attractive alternative. However, the complex and variable operating conditions and fluctuating power demands of loaders are incompatible with the slow dynamic response of fuel cells. Therefore, a hybrid power system must be combined with power batteries and optimized power distribution through energy management methods.

[0003] The existing energy management methods for fuel cell loaders have three main deficiencies: first, they lack the ability to accurately identify working conditions and predict power requirements, making it difficult to adapt to the complex operating environment of the loader; second, the optimization goal is single, only considering fuel economy while ignoring the power source life degradation cost; third, they do not consider the impact of environmental factors in different regions on the performance and life of the power source, and lack environmental adaptability.

[0004] With the development of artificial intelligence technology, applying machine learning methods to operating condition identification and power demand prediction has become a research hotspot. Furthermore, loaders operating across regions face significant differences in environmental conditions such as temperature, humidity, and altitude. These environmental factors directly affect the performance degradation characteristics of fuel cells and power batteries. Traditional energy management methods use fixed control parameters and are unable to adapt to complex and changing environmental conditions.

[0005] Therefore, it is urgent to develop an energy management method for fuel cell loaders based on hierarchical control and environmental adaptation. Through environmental perception and adaptive optimization control, the energy management performance and system reliability of the fuel cell loader hybrid system under different environmental conditions can be improved, thereby promoting the green and intelligent transformation of the construction machinery industry. Summary of the Invention

[0006] In view of the above problems, the purpose of the present invention is to provide an energy management method for a fuel cell loader. By accurately identifying different operating conditions, a joint network architecture is used to improve the accuracy of demand power prediction, and environmental correction coefficients are obtained through position perception, the impact of environmental factors on the life of the power source is incorporated into optimization control, and the output power distribution of the fuel cell and the power battery is collaboratively optimized. While ensuring the efficient operation of the equipment, the energy management performance and system reliability of the fuel cell loader under different environmental conditions are significantly improved.

[0007] The present invention provides a fuel cell loader energy management method, comprising the following steps: Step S10, using the onboard sensor system of the fuel cell loader, collecting time series data of power requirements of the loader under typical operating conditions of shoveling, transporting, unloading, and idling, and building a cycle operating condition database; Step S20: Extract features from the cycle operating condition database to construct an original feature matrix representing the cycle operating condition of the fuel cell loader. X ; Step S30, using the crossover probability algorithm to calculate the original feature matrix Perform dimensionality reduction processing to obtain the reduced-dimensional working condition feature matrix ; Step S40, the dimension reduction feature matrix outputted in step S30 Adopting kurtosis jump algorithm to adaptively determine the number of clusters and initial cluster centers to establish a working condition identification model for the fuel cell loader; Step S50: Based on the working condition identification results of the working condition identification model, a joint network communication prediction model is established. For each type of working condition, a corresponding window-shifting neural network is established, and the spiral search algorithm is used to independently optimize the parameters of each network to optimize the power demand prediction accuracy of the fuel cell loader under different working conditions. In step S60, when applied online, the on-board sensor system collects and caches the power demand data of the loader in the most recent time window in real time to form a historical power demand sequence. The system then extracts operating condition characteristics based on the historical power demand sequence, uses the operating condition recognition model to determine the current operating condition type, and then selects the corresponding window-shifting neural network in step S50 to perform power demand prediction, thereby obtaining a future power demand sequence within the prediction time domain. Step S70: Based on the environmental characteristics (temperature, humidity, and altitude) acquired in real time in the current area, a multi-objective optimization function is established that takes into account the impact of environmental factors on the service degradation of fuel cells and power batteries. Combined with the future power demand sequence, the optimal control principle is applied within the forecast time domain to solve the optimal control sequence, thereby achieving environmentally adaptive optimization of the output power of the fuel cell and power battery.

[0008] As a preferred embodiment of the present invention, when extracting features from the cycle operating condition database in step S20, each complete working cycle is defined as a working condition sample, and the average power, minimum power, maximum power, and power standard deviation in the characteristic parameters are extracted to construct the original feature matrix characterizing the cycle operating condition of the fuel cell loader. .

[0009] As a preferred embodiment of the present invention, step S30 includes the following steps: Step S301: original feature matrix Perform Z-score standardization to obtain a standardized feature matrix : ; Where, is the original feature matrix No. The first working condition sample eigenvalues; is the original feature matrix No. Mean of column features; is the original feature matrix No. Standard deviation of column characteristics; is the standardized feature matrix The corresponding elements in ; Step S302: construct a K-nearest neighbor graph and calculate the conditional probability of two working condition sample points in the high-dimensional space and the corresponding conditional probability in low-dimensional space , and then minimize the cross entropy loss function CE by gradient descent: ; Where, is a logarithmic function; 、 The working condition sample number; Iteratively optimize the low-dimensional coordinates of each working condition sample in the low-dimensional space until the loss function CE converges or reaches the preset number of iterations, and obtain the working condition feature matrix after dimensionality reduction by the cross probability algorithm .

[0010] As a preferred embodiment of the present invention, step S40 includes the following steps: Step S401: Based on the kurtosis jump algorithm, the decision value of each working condition sample after dimension reduction is calculated. Adaptively determine cluster centers: ; Where, , is the total number of working condition samples; For working condition samples The local density of For working condition samples to all local densities greater than The minimum distance of the working condition sample; Step S402: The decision value of each working condition sample Arrange in descending order to get a new sequence ,in ; Calculate the descending gradient of adjacent decision values ​​after sorting: ; Where, is the subscript of the decision value after descending order, ; After sorting The decision value and The relative descent gradient between decision values; 、 are the first and decision values; is the maximum decision value after sorting; When a certain descent gradient When the set jump threshold is exceeded, the number of clusters is determined , that is, select the first one with the largest decision value The working condition samples are used as the initial clustering centers to initialize the clustering parameters and complete the construction of the working condition identification model.

[0011] As a preferred embodiment of the present invention, step S50 is constructed by The independent windows shift the neural network one by one and optimize their parameters respectively to form a joint network communication prediction model, which specifically includes the following steps: Step S501: for the Types of working conditions, design the joint network communication prediction model architecture, the joint network communication prediction model consists of A window-shifted neural network is composed of: ; Where, For the The window shift neural network output of each operating condition type is the future predicted power sequence; is the input vector, i.e. the historical power sequence; For the The number of hidden nodes in the network under different working conditions; For the The connection weights from hidden layer nodes to the output layer; For the The center vector of the basis functions; For the The width parameter of the basis function; is an exponential function; Number the hidden layer nodes; Step S502: The parameters to be optimized of the window shift neural network corresponding to the working condition type are encoded as the position vector in the spiral search algorithm : ; Step S503: construct a fitness function based on the accuracy of future demand power series prediction: ; Where, For the The root mean square error of the working condition type is used as the fitness function of the spiral search algorithm; For the The number of training working condition samples of each working condition type; To predict the time domain length; 、 For the The working condition samples will be The power prediction value and actual power demand value of each time step; Step S504, using a spiral search algorithm to iteratively optimize the network parameters for each operating condition type; In the initial optimization stage, the search individual moves closer to the current optimal position: ; Where, l is the number of iterations; 、 For the Search individual in the first 、 The position vector of the iteration; For the The current optimal individual position under the working condition type; is the adaptive factor; for Random numbers in the interval; is the maximum number of iterations; is a natural constant; In the exploration phase, the individual searches near the optimal position through spiral motion: ; Where, To find the optimal distance; is a sine function, is pi; Step S505: Repeat steps S503 to S504 for each operating condition type until the convergence condition is met or the maximum number of iterations is reached. When , the optimal position of each working condition type is output The corresponding window-by-window neural optimization parameters are shifted to complete the construction of the joint network communication prediction model.

[0012] As a preferred embodiment of the present invention, step S70 includes the following steps: Step S701, time interval The current geographical location of the fuel cell loader is obtained through GPS positioning, and the environmental characteristics of the current area are identified based on the geographical location information to obtain the environmental correction coefficient: ; Where, and are the current latitude and longitude coordinates respectively; and are the position decay correction coefficients of the fuel cell and power battery respectively; 、 are the position decay correction functions of the fuel cell and the power battery respectively; Step S702: Establish a comprehensive objective function including hydrogen consumption rate, fuel cell service degradation rate and power battery service degradation rate : ; Where, 、 is the start time and end time; is the hydrogen consumption rate; Fuel cell service degradation rate; For power battery service decline; is the hydrogen consumption cost weight; Weighting fuel cell service decay costs; The weight of the service degradation cost of the power battery; is the time element; Step S703: Constructing a state function : ; Where, The state of charge of the power battery changes; is a co-variable; is the system state equation; Output power for the fuel cell; According to the optimal control principle, combined with the future demand power sequence obtained in step S60, the optimal output power sequence of the fuel cell is solved: ; Where, is the minimum operator; is the optimal output power sequence of the fuel cell; is the optimal co-state variable; is the state of charge of the power battery under the optimal trajectory; After the solution is completed, the optimal output power of the fuel cell and power battery at the current moment is executed, and the optimization solution is re-performed in the next control cycle according to the predicted future demand power sequence to achieve rolling time domain optimization control.

[0013] The beneficial effects of the present invention are as follows: 1. This paper proposes a crossover probability algorithm and applies it to the processing of fuel cell loader operating condition features, achieving effective dimensionality reduction of complex operating condition features and providing high-quality low-dimensional feature representation for subsequent cluster analysis and operating condition recognition model construction.

[0014] 2. The present invention proposes an adaptive initialization method guided by kurtosis jump clustering, which overcomes the defects of traditional clustering algorithms that are sensitive to initial values ​​and require a preset number of clusters. It adaptively adjusts the optimal number of clusters and initial centers through kurtosis changes, thereby improving the accuracy and stability of working condition identification.

[0015] 3. The present invention proposes a parameter optimization method that combines a spiral search algorithm with a window-shifting neural network, which significantly improves the demand power prediction accuracy of the fuel cell loader.

[0016] 4. The present invention proposes a fuel cell loader joint network communication prediction model, which independently trains dedicated demand power prediction subnetworks for different working conditions, and can effectively improve the adaptability of the fuel cell loader demand power prediction model.

[0017] 5. The present invention introduces a location-based environmental correction coefficient to dynamically adjust the calculation of the power source decay rate, thereby achieving environmental adaptation of the energy management strategy and significantly improving the adaptability and operational performance of the system in different geographical environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By referring to the following description in conjunction with the accompanying drawings, and with a more complete understanding of the present invention, other objects and results of the present invention will become more clear and easy to understand. In the accompanying drawings: Figure 1 This is a flow chart of the fuel cell loader energy management method according to an embodiment of the present invention; Figure 2 This is a flow chart of a spiral search algorithm for optimizing a window-shifting neural network according to an embodiment of the present invention; Figure 3 This is an optimization framework for the fuel cell loader energy management method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] See Figure 1-3 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] An embodiment of the present invention provides an energy management method for a fuel cell loader, comprising the following steps: Step S10, using the onboard sensor system of the fuel cell loader, collecting time series data of power requirements of the loader under typical operating conditions of shoveling, transporting, unloading, and idling, and building a cycle operating condition database; Step S20: Extract features from the cycle operating condition database to construct an original feature matrix representing the cycle operating condition of the fuel cell loader. X ; Among them, when extracting features from the cycle working condition database, each complete working cycle is defined as a working condition sample, and the average power, minimum power, maximum power, and power standard deviation in the characteristic parameters are extracted to construct the original feature matrix that characterizes the cycle working condition of the fuel cell loader. .

[0021] Step S30, using the crossover probability algorithm to calculate the original feature matrix Perform dimensionality reduction processing to obtain the reduced-dimensional working condition feature matrix ; Step S301: original feature matrix Perform Z-score standardization to obtain a standardized feature matrix : ; Where, is the original feature matrix No. The first working condition sample eigenvalues; is the original feature matrix No. Mean of column features; is the original feature matrix No. Standard deviation of column characteristics; is the standardized feature matrix The corresponding elements in ; Step S302: construct a K-nearest neighbor graph and calculate the conditional probability of two working condition sample points in the high-dimensional space and the corresponding conditional probability in low-dimensional space , and then minimize the cross entropy loss function CE by gradient descent: ; Where, is a logarithmic function; 、 The working condition sample number; Iteratively optimize the low-dimensional coordinates of each working condition sample in the low-dimensional space until the loss function CE converges or reaches the preset number of iterations, and obtain the working condition feature matrix after dimensionality reduction by the cross probability algorithm .

[0022] Step S40, the dimension reduction feature matrix outputted in step S30 Adopting kurtosis jump algorithm to adaptively determine the number of clusters and initial cluster centers to establish a working condition identification model for the fuel cell loader; Step S401: Based on the kurtosis jump algorithm, the decision value of each working condition sample after dimension reduction is calculated. Adaptively determine cluster centers: ; Where, , is the total number of working condition samples; For working condition samples The local density of For working condition samples to all local densities greater than The minimum distance of the working condition sample; Step S402: The decision value of each working condition sample Arrange in descending order to get a new sequence ,in ; Calculate the descending gradient of adjacent decision values ​​after sorting: ; Where, is the subscript of the decision value after descending order, ; After sorting The decision value and The relative descent gradient between decision values; 、 are the first and decision values; is the maximum decision value after sorting; When a certain descent gradient When the set jump threshold is exceeded, the number of clusters is determined , that is, select the first one with the largest decision value The working condition samples are used as the initial clustering centers to initialize the clustering parameters and complete the construction of the working condition identification model.

[0023] Step S50: Based on the working condition identification results of the working condition identification model, a joint network communication prediction model is established. For each type of working condition, a corresponding window-shifting neural network is established, and the spiral search algorithm is used to independently optimize the parameters of each network to optimize the power demand prediction accuracy of the fuel cell loader under different working conditions. By building The independent windows shift the neural network one by one and optimize their parameters respectively to form a joint network communication prediction model, which specifically includes the following steps: Step S501: for the Types of working conditions, design the joint network communication prediction model architecture, the joint network communication prediction model consists of A window-shifted neural network is composed of: ; Where, For the The window shift neural network output of each operating condition type is the future predicted power sequence; is the input vector, i.e. the historical power sequence; For the The number of hidden nodes in the network under different working conditions; For the The connection weights from hidden layer nodes to the output layer; For the The center vector of the basis functions; For the The width parameter of the basis function; is an exponential function; Number the hidden layer nodes; Step S502: The parameters to be optimized of the window shift neural network corresponding to the working condition type are encoded as the position vector in the spiral search algorithm : ; Step S503: construct a fitness function based on the accuracy of future demand power series prediction: ; Where, For the The root mean square error of the working condition type is used as the fitness function of the spiral search algorithm; For the The number of training working condition samples of each working condition type; To predict the time domain length; 、 For the The working condition samples will be The power prediction value and actual power demand value of each time step; Step S504, using a spiral search algorithm to iteratively optimize the network parameters for each operating condition type; In the initial optimization stage, the search individual moves closer to the current optimal position: ; Where, l is the number of iterations; 、 For the Search individual in the first 、 The position vector of the iteration; For the The current optimal individual position under the working condition type; is the adaptive factor; for Random numbers in the interval; is the maximum number of iterations; is a natural constant; In the exploration phase, the individual searches near the optimal position through spiral motion: ; Where, To find the optimal distance; is a sine function, is pi; Step S505: Repeat steps S503 to S504 for each operating condition type until the convergence condition is met or the maximum number of iterations is reached. When , the optimal position of each working condition type is output The corresponding window-by-window neural optimization parameters are shifted to complete the construction of the joint network communication prediction model.

[0024] In step S60, when applied online, the on-board sensor system collects and caches the power demand data of the loader in the most recent time window in real time to form a historical power demand sequence. The system then extracts operating condition characteristics based on the historical power demand sequence, uses the operating condition recognition model to determine the current operating condition type, and then selects the corresponding window-shifting neural network in step S50 to perform power demand prediction, thereby obtaining a future power demand sequence within the prediction time domain. Step S70: Based on the environmental characteristics (temperature, humidity, and altitude) acquired in real time in the current area, a multi-objective optimization function is established that takes into account the impact of environmental factors on the service degradation of fuel cells and power batteries. Combined with the future power demand sequence, the optimal control principle is applied within the forecast time domain to solve the optimal control sequence, thereby achieving environmentally adaptive optimization of the output power of the fuel cell and power battery.

[0025] Step S701, time interval The current geographical location of the fuel cell loader is obtained through GPS positioning, and the environmental characteristics of the current area are identified based on the geographical location information to obtain the environmental correction coefficient: ; Where, and are the current latitude and longitude coordinates respectively; and are the position decay correction coefficients of the fuel cell and power battery respectively; 、 are the position decay correction functions of the fuel cell and the power battery respectively; Step S702: Establish a comprehensive objective function including hydrogen consumption rate, fuel cell service degradation rate and power battery service degradation rate : ; Where, 、 is the start time and end time; is the hydrogen consumption rate; Fuel cell service degradation rate; For power battery service decline; is the hydrogen consumption cost weight; Weighting fuel cell service decay costs; The weight of the service degradation cost of the power battery; is the time element; Step S703: Constructing a state function : ; Where, The state of charge of the power battery changes; is a co-variable; is the system state equation; Output power for the fuel cell; According to the optimal control principle, combined with the future demand power sequence obtained in step S60, the optimal output power sequence of the fuel cell is solved: ; Where, is the minimum operator; is the optimal output power sequence of the fuel cell; is the optimal co-state variable; is the state of charge of the power battery under the optimal trajectory; After the solution is completed, the optimal output power of the fuel cell and power battery at the current moment is executed, and the optimization solution is re-performed in the next control cycle according to the predicted future demand power sequence to achieve rolling time domain optimization control.

[0026] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A fuel cell loader energy management method, characterized in that: The following steps are involved: Step S10, using the onboard sensor system of the fuel cell loader, collecting time series data of power requirements of the loader under typical operating conditions of shoveling, transporting, unloading, and idling, and building a cycle operating condition database; Step S20: Extract features from the cycle operating condition database to construct an original feature matrix representing the cycle operating condition of the fuel cell loader. X ; Step S30, using the crossover probability algorithm to calculate the original feature matrix Perform dimensionality reduction processing to obtain the reduced-dimensional working condition feature matrix ; Step S40, the dimension reduction feature matrix outputted in step S30 Adopting kurtosis jump algorithm to adaptively determine the number of clusters and initial cluster centers to establish a working condition identification model for the fuel cell loader; Step S50: Based on the working condition identification results of the working condition identification model, a joint network communication prediction model is established. For each type of working condition, a corresponding window-shifting neural network is established, and the spiral search algorithm is used to independently optimize the parameters of each network to optimize the power demand prediction accuracy of the fuel cell loader under different working conditions. In step S60, when applied online, the on-board sensor system collects and caches the power demand data of the loader in the most recent time window in real time to form a historical power demand sequence. The system then extracts operating condition characteristics based on the historical power demand sequence, uses the operating condition recognition model to determine the current operating condition type, and then selects the corresponding window-shifting neural network in step S50 to perform power demand prediction, thereby obtaining a future power demand sequence within the prediction time domain. Step S70: Based on the environmental characteristics of the current area acquired in real time, a multi-objective optimization function is established that takes into account the impact of environmental factors on the service degradation of fuel cells and power batteries. Combined with the future demand power sequence, the optimal control principle is applied in the prediction time domain to solve the optimal control sequence, thereby achieving environmentally adaptive optimization allocation of the output power of the fuel cell and power battery.

2. The fuel cell loader energy management method according to claim 1, characterized in that: When extracting features from the cycle operating condition database in step S20, each complete working cycle is defined as a working condition sample, and the average power, minimum power, maximum power, and power standard deviation in the characteristic parameters are extracted to construct the original feature matrix characterizing the cycle operating condition of the fuel cell loader. .

3. The fuel cell loader energy management method according to claim 1, characterized in that: Step S30 includes the following steps: Step S301: original feature matrix Perform Z-score standardization to obtain a standardized feature matrix : ； Where, is the original feature matrix No. The first working condition sample eigenvalues; is the original feature matrix No. Mean of column features; is the original feature matrix No. Standard deviation of column characteristics; is the standardized feature matrix The corresponding elements in ; Step S302: construct a K-nearest neighbor graph and calculate the conditional probability of two working condition sample points in the high-dimensional space and the corresponding conditional probability in low-dimensional space , and then minimize the cross entropy loss function CE by gradient descent: ； Where, is a logarithmic function; 、 The working condition sample number; Iteratively optimize the low-dimensional coordinates of each working condition sample in the low-dimensional space until the loss function CE converges or reaches the preset number of iterations, and obtain the working condition feature matrix after dimensionality reduction by the cross probability algorithm .

4. The fuel cell loader energy management method according to claim 1, characterized in that: Step S40 includes the following steps: Step S401: Based on the kurtosis jump algorithm, the decision value of each working condition sample after dimension reduction is calculated. Adaptively determine cluster centers: ； Where, , is the total number of working condition samples; For working condition samples The local density of For working condition samples to all local densities greater than The minimum distance of the working condition sample; Step S402: The decision value of each working condition sample Arrange in descending order to get a new sequence ,in ; Calculate the descending gradient of adjacent decision values ​​after sorting: ； Where, is the subscript of the decision value after descending order, ; After sorting The decision value and The relative descent gradient between decision values; 、 are the first and decision values; is the maximum decision value after sorting; When a certain descent gradient When the set jump threshold is exceeded, the number of clusters is determined , that is, select the first one with the largest decision value The working condition samples are used as the initial clustering centers to initialize the clustering parameters and complete the construction of the working condition identification model.

5. The fuel cell loader energy management method according to claim 1, characterized in that: Step S50 includes the following steps: Step S501: for the Types of working conditions, design the joint network communication prediction model architecture, the joint network communication prediction model consists of A window-shifted neural network is composed of: ； Where, For the The window shift neural network output of each operating condition type is the future predicted power sequence; is the input vector, i.e. the historical power sequence; For the The number of hidden nodes in the network under different working conditions; For the The connection weights from hidden layer nodes to the output layer; For the The center vector of the basis functions; For the The width parameter of the basis function; is an exponential function; Number the hidden layer nodes; Step S502: The parameters to be optimized of the window shift neural network corresponding to the working condition type are encoded as the position vector in the spiral search algorithm : ； Step S503: construct a fitness function based on the accuracy of future demand power series prediction: ； Where, For the The root mean square error of the working condition type is used as the fitness function of the spiral search algorithm; For the The number of training working condition samples of each working condition type; To predict the time domain length; 、 For the The working condition samples will be The power prediction value and actual power demand value of each time step; Step S504, using a spiral search algorithm to iteratively optimize the network parameters for each operating condition type; In the initial optimization stage, the search individual moves closer to the current optimal position: ； Where, l is the number of iterations; 、 For the Search individual in the first 、 The position vector of the iteration; For the The current optimal individual position under the working condition type; is the adaptive factor; for Random numbers in the interval; is the maximum number of iterations; is a natural constant; In the exploration phase, the individual searches near the optimal position through spiral motion: ； Where, To find the optimal distance; is a sine function, is pi; Step S505: Repeat steps S503 to S504 for each operating condition type until the convergence condition is met or the maximum number of iterations is reached. When , the optimal position of each working condition type is output The corresponding window-by-window neural optimization parameters are shifted to complete the construction of the joint network communication prediction model.

6. The fuel cell loader energy management method according to claim 1, characterized in that: Step S 70 comprises the following steps: Step S701, time interval The current geographical location of the fuel cell loader is obtained through GPS positioning, and the environmental characteristics of the current area are identified based on the geographical location information to obtain the environmental correction coefficient: ； Where, and are the current latitude and longitude coordinates respectively; and are the position decay correction coefficients of the fuel cell and power battery respectively; 、 are the position decay correction functions of the fuel cell and the power battery respectively; Step S702: Establish a comprehensive objective function including hydrogen consumption rate, fuel cell service degradation rate and power battery service degradation rate : ； Where, 、 is the start time and end time; is the hydrogen consumption rate; Fuel cell service degradation rate; For power battery service decline; is the hydrogen consumption cost weight; Weighting fuel cell service decay costs; The weight of the service degradation cost of the power battery; is the time element; Step S703: Constructing a state function : ； Where, The state of charge of the power battery changes; is a co-variable; is the system state equation; Output power for the fuel cell; According to the optimal control principle, combined with the future demand power sequence obtained in step S60, the optimal output power sequence of the fuel cell is solved: ； Where, is the minimum operator; is the optimal output power sequence of the fuel cell; is the optimal co-state variable; is the state of charge of the power battery under the optimal trajectory; After the solution is completed, the optimal output power of the fuel cell and power battery at the current moment is executed, and the optimization solution is re-performed in the next control cycle according to the predicted future demand power sequence to achieve rolling time domain optimization control.

Images