Intelligent prediction method and system for power demand of fuel cell hybrid powertrain

Abstract

The present invention relates to the field of fuel cell technology, and provides a method and system for intelligently predicting power demand of a fuel cell hybrid powertrain, including: obtaining power data at the current moment and multiple historical moments before the current moment to obtain a time series sample; based on the time series sample, using a prediction model to predict power data at future moments; wherein the prediction model uses multiple prediction methods to predict power data at future moments respectively, and weights the prediction values ​​obtained by the multiple prediction methods, and the weighted weights are recursively optimized by using a gray wolf and crown porcupine fusion algorithm with the goal of minimizing the total prediction error value, and the gray wolf and crown porcupine fusion algorithm uses the third defense strategy of the crown porcupine algorithm to update the gray wolf position in the hunting operation of the gray wolf algorithm. Reduce the error between the predicted value and the true value, and obtain a more accurate power prediction result.

Description

Technical Field

[0001] The present invention relates to the field of fuel cell technology, and in particular to a method and system for intelligently predicting power demand of a fuel cell hybrid powertrain. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] In the context of energy conservation and emission reduction, hydrogen fuel cells have broad application prospects as a clean and environmentally friendly energy conversion device in mobile equipment, especially in the fields of automobiles, ships, tractors and construction machinery, due to their advantages such as low operating temperature, high conversion efficiency and zero carbon emissions. With the development of the transportation system, road conditions and traffic have become more complicated. Cars often need to experience extreme power such as idling, rapid acceleration and rapid deceleration during operation; ships will experience large speed changes during entry, exit and navigation, which requires the fuel cell hybrid powertrain on cars or ships to have a higher dynamic response.

[0004] However, it takes a certain amount of time to transport fuel inside the fuel cell control system, especially the air supply system, which is prone to air shortage and affects the dynamic response performance.

[0005] Predicting the future power of fuel cells can provide a basis for the fuel cell control system to predict short-term power demand, accelerate the dynamic response of the fuel cell control system and power system, and improve its power performance.

[0006] However, the current method of using a single algorithm to predict power results in a large error between the predicted power value and the actual value. Summary of the invention

[0007] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a method and system for intelligently predicting the power demand of a fuel cell hybrid powertrain, which can quickly predict short-term power, provide a basis for short-term power demand for the fuel cell control system, and enable the control system to make a dynamic response in advance. In addition, each recursive prediction performs weighted fusion on the results of multiple prediction methods, and uses the gray wolf and crowned porcupine fusion algorithm to optimize the weights and balance the prediction methods, thereby reducing the error between the predicted value and the true value and obtaining a more accurate power prediction result.

[0008] In order to achieve the above object, the present invention adopts the following technical solution:

[0009] A first aspect of the present invention provides a method for intelligently predicting power demand of a fuel cell hybrid powertrain, comprising:

[0010] Obtain power data at the current moment and multiple previous historical moments to obtain time series samples;

[0011] Based on the time series samples, a prediction model is used to predict power data at future moments;

[0012] Among them, the prediction model adopts multiple prediction methods to predict the power data at future moments respectively, and weights the prediction values ​​obtained by multiple prediction methods, and the weighted weights are recursively optimized by using the gray wolf and crowned porcupine fusion algorithm with the goal of minimizing the total prediction error value. The gray wolf and crowned porcupine fusion algorithm adopts the third defense strategy of the crowned porcupine algorithm to update the gray wolf position in the hunting operation of the gray wolf algorithm.

[0013] Furthermore, the step of calculating the total prediction error value includes: for each time series sample obtained by dividing the historical data set, using multiple prediction methods to predict the power data respectively, and weighting the predicted values ​​obtained by the multiple prediction methods using the weights of each individual code in the gray wolf population to obtain the predicted value, and calculating the error value between the predicted value and the true value; calculating the sum of the error values ​​between all predicted values ​​and the true value to obtain the total prediction error value.

[0014] Furthermore, a plurality of time series samples are obtained by sliding sampling a historical data set using a time window, wherein the historical data set is composed of a plurality of historical transient powers including a timestamp.

[0015] Furthermore, the update formula of the gray wolf position in the hunting operation is:

[0016]

[0017] in, represents the position of the gray wolf after the tth iteration, represents the position of the i-th individual after the t-th iteration, r1, r2, r3 are random numbers, N is the size of the gray wolf population, t is the number of current iterations, τ3 is a random number, is a parameter used to control the search direction. Used to simulate the three situations in the third defense strategy, is the odor diffusion factor, is the rth wolf in the gray wolf population after the tth iteration i The position of each individual, γ t is the defense factor, A1, A2 and A3 are coefficient vectors, and are the positions of α wolf, β wolf and δ wolf respectively, and are the position of the gray wolf after the tth iteration and the distances between α wolf, β wolf and δ wolf respectively.

[0018] Furthermore, the multiple prediction methods include generalized regression neural network and unscented Kalman filtering.

[0019] Furthermore, the power data is transient power and its first-order derivative including a timestamp.

[0020] A second aspect of the present invention provides a power demand intelligent prediction system for a fuel cell hybrid powertrain, comprising:

[0021] A data acquisition module is configured to: acquire power data at a current moment and multiple previous historical moments to obtain time series samples;

[0022] A power prediction module is configured to: predict power data at a future time by using a prediction model based on the time series samples;

[0023] Among them, the prediction model adopts multiple prediction methods to predict the power data at future moments respectively, and weights the prediction values ​​obtained by multiple prediction methods, and the weighted weights are recursively optimized by using the gray wolf and crowned porcupine fusion algorithm with the goal of minimizing the total prediction error value. The gray wolf and crowned porcupine fusion algorithm adopts the third defense strategy of the crowned porcupine algorithm to update the gray wolf position in the hunting operation of the gray wolf algorithm.

[0024] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the method for intelligently predicting power demand of a fuel cell hybrid powertrain as described above.

[0025] The fourth aspect of the present invention provides a computer device, including a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein when the processor executes the program, the steps in the method for intelligently predicting power demand of a fuel cell hybrid powertrain as described above are implemented.

[0026] A fifth aspect of the present invention provides a computer program product, which is a computer program product containing a computer program. When a processor executes the computer program, the steps in the method for intelligently predicting power demand of a fuel cell hybrid powertrain as described above are implemented.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention can quickly predict short-term power, provide a basis for short-term power demand for a fuel cell control system, and enable the fuel cell control system to make a dynamic response in advance, thereby improving the dynamic response of the fuel cell hybrid powertrain.

[0029] The time window constructed by the present invention selects the appropriate number of data points and time intervals to effectively suppress phase mismatch and meet the requirements of a certain prediction time range. The power demand is predicted by weighted generalized regression neural network and unscented Kalman filter algorithms, and the weights are optimized by using the gray wolf and crowned porcupine fusion algorithm to balance the prediction methods, thereby reducing the error between the predicted value and the true value and obtaining a more accurate power prediction result.

[0030] The present invention adopts the third defense strategy of the crested porcupine to update the position of the gray wolf in the hunting operation of the gray wolf algorithm, so as to avoid the gray wolf algorithm from falling into the local optimum, improve the accuracy of the weight, and further improve the accuracy of the power prediction result. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0032] Figure 1 is a flow chart of a method for intelligently predicting power demand of a fuel cell hybrid powertrain according to a first embodiment of the present invention;

[0033] Figure 2 It is a structural diagram of a computer device according to a fourth embodiment of the present invention. DETAILED DESCRIPTION

[0034] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0035] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0036] Embodiment 1

[0037] This embodiment provides a method for intelligently predicting power demand of a fuel cell hybrid powertrain.

[0038] The present embodiment provides a method for intelligently predicting power demand of a fuel cell hybrid powertrain, which is applied in the fields of automobiles, ships, tractors and engineering machinery.

[0039] This embodiment provides a method for intelligently predicting the power demand of a fuel cell hybrid powertrain. An intelligent algorithm is designed to quickly predict short-term power, provide a basis for short-term power demand for the fuel cell control system, enable the control system to make a dynamic response in advance, reduce the impact of time delay in the fuel cell control system, increase the dynamic response of the fuel cell control system, and improve the power performance of the fuel cell vehicle.

[0040] This embodiment provides a method for intelligently predicting power demand of a fuel cell hybrid powertrain. Figure 1 As shown, the following steps are included:

[0041] Step 1: Obtain a large amount of historical transient power data of the fuel cell hybrid powertrain including time stamps and store them as a historical data set;

[0042] Step 2: Based on the historical data set (i.e., sample set) collected from the fuel cell hybrid powertrain, the time window is divided to obtain time series samples, and the prediction recursive step size is set. The demand power is predicted respectively using the generalized regression neural network and the unscented Kalman filter algorithms according to the time series samples, and the demand power predicted by the two algorithms is weighted c1 and c2 to finally obtain the predicted demand power of the fuel cell system. The corresponding error is obtained based on the predicted demand power and the actual power. The corresponding c1 and c2 weights are obtained based on the fused gray wolf-crowned porcupine algorithm (gray wolf and crested porcupine fusion algorithm) with the minimum total prediction error value, thereby obtaining a weighted combined intelligent prediction model (referred to as the prediction model).

[0043] Step 3: Collect current power information, and obtain the power prediction result at future times based on the weighted combined intelligent prediction model and the recursive prediction step size.

[0044] In this embodiment, the operation process of obtaining the historical data set in step 1 is as follows: the transient current I(t) and voltage U(t) containing timestamps on the DC bus of the fuel cell hybrid powertrain collected by the CAN communication network are read through the OBD-Ⅱ (second generation on-board diagnostic system) interface of the fuel cell vehicle, and the transient power P(t) of the fuel cell system is calculated according to the formula P(t)=U(t)·I(t). The interval between two adjacent transient power points is Δt. si Second.

[0045] In this embodiment, the time window in step 2 selects the transient power of m+1 consecutive data points in the historical data set of the fuel cell hybrid system, and the interval between two adjacent transient power points is Δt si Seconds. According to the formula Calculate the first-order derivative corresponding to each transient power point, and construct the time series sample of the corresponding time window with the transient power and its first-order derivative, such as x = [P0, P1, P1', P2, P2', ..., P i ,P i '] T , where i = 1, 2, ..., m, m + 1 is the size of the time window, P i represents the i-th transient power, P i represents the first-order derivative of the i-th transient power.

[0046] In this embodiment, a suitable recursive prediction step length Δt is selected. hi Seconds, Δt hi =nΔt si , n is an integer greater than zero and less than m, that is, the time interval for each sliding of the time window, and each time the time window advances Δt hi Seconds, n instantaneous powers are recorded from the historical sample set and their first-order derivatives are calculated. Thus, the time window and its time series samples are reconstructed to obtain a series of time windows and their time series samples.

[0047] For example, when constructing a time window, m is 8 and Δt si are all 0.1, and the resulting time window is X = [P0, P1, P2, ..., P8] T , thus we get the time series sample x = [P0, P1, P1', P2, P2', ..., P8, P8'] T n=1, the time window advances by 0.1 second each time, and a transient power point is recorded from the historical sample set and its first-order derivative is calculated.

[0048] In step 3, the weighted values ​​of the generalized regression neural network and the unscented Kalman filter are weighted, and the gray wolf and crested porcupine fusion algorithm is used for recursive optimization with the goal of minimizing the total prediction error value. The calculation steps of the total prediction error value are as follows:

[0049] Step (1) For the first prediction power time series sample x = [P0, P1, P1', P2, P2', ..., P m ,P m '] T (e.g., x = [P0, P1, P1', P2, P2', ..., P8, P8'] T ), the prediction model is obtained by training the generalized regression neural network and the unscented Kalman filter respectively, and the step length of the backward recursion is Δt hi , which is 0.1 at this time, for P m+1 (e.g., P9) is used for prediction. The prediction results of generalized regression neural network and unscented Kalman filter are P (m+1)a and P(m+1)b (e.g., P 9a and P 9b ), the weighted prediction power is P h(m+1) =c1P (m+1)a +c2P (m+1)b (e.g., P h9 =c1P 9a +c2P 9b ), the error between the two is |P m+1 -P h(m+1) |(e.g., |P9-P h9 |);

[0050] Step (2) Backward recursion Δt hi Seconds later, that is, 0.1 seconds later, the time series sample is x = [P1, P1', P2, P2', ..., P m ,P' m ,P m+1 ,P' m+1 ] T (e.g., x = [P1, P1', P2, P2', ..., P8, P8', P9, P9'] T ), generalized regression neural network and unscented Kalman filter are used to predict the m+2th transient power respectively, and P (m+2)a and P (m+2)b (e.g., P 10a and P 10b ), the weighted prediction power is P h(m+2) =c1P (m+2)a +c2P (m+2)b (e.g., P h10 =c1P 10a +c2P 10b ), the error between the two is |P (m+2) -P h(m+2) |(e.g., |P 10 -P h10 |);

[0051] Step (3) is analogous to this. Based on the historical data set, all time series samples are constructed and traversed. For the transient power P i The prediction of P is obtained by using generalized regression neural network and unscented Kalman filter. ia and P ib , the weighted prediction power is P hi =c1P ia +c2P ib , the corresponding error is |P i -P hi |;

[0052] Step (4) Calculate all error values ​​|P i-P hi | to get the total prediction error value.

[0053] Among them, the generalized regression neural network is a form of radial basis function neural network deformation. Since the output value is power, its output layer is 1.

[0054] In this embodiment, the generalized regression neural network consists of an input layer, a hidden layer, a summation layer and an output layer. The number of nodes in the input layer is determined by the number of input variables, the number of nodes in the hidden layer is the same as the number of training samples, and the summation layer usually contains two types of nodes, namely, weighted summation neurons S and S. w and simple summation neuron S s .

[0055] The transfer function used by the hidden layer nodes is the Gaussian radial basis function. The output of the i-th hidden layer node after calculation by the transfer function is p i , the formula is:

[0056]

[0057] Among them, p i is the transfer function, σ is the smoothing factor, X is the network input variable, X i is the learning sample of the i-th neuron.

[0058] The summation layer uses two types of neurons for summation: simple summation neurons S s and weighted sum neuron S w , S s The outputs of all pattern layer neurons are summed up, and the connection weight is 1. The transfer function formula is:

[0059]

[0060] Among them, S w The outputs of all pattern layer neurons are weighted and summed, and the number is equal to the number of output variables. The transfer function formula is:

[0061]

[0062] Among them, w i To connect to S w The weight of the i-th neuron.

[0063] The number of neurons in the output layer is equal to the dimension of the output vector Y. The output value is the power or the first-order derivative of the power, so the dimension is 1. The formula for the output value Y of the generalized regression neural network is:

[0064]

[0065] Among them, the unscented Kalman filter is a combination of the unscented transform and the Kalman filter algorithm to form an unscented Kalman filter algorithm. The unscented transform mainly calculates the mean and covariance, and then determines the sample point set to achieve an approximate probability distribution. The unscented transform approximates the Gaussian distribution by selecting a certain number of sample point sets. The principle is: first select 2n+1 sample point sets from the initial distribution so that the mean and covariance of these point sets are the same as those of the initial distribution; then substitute the point set into the nonlinear function, use the result of the operation as the new point set, and finally calculate the mean and covariance of the unscented transform through the new point set.

[0066] Among them, unscented Kalman filter and unscented transformation are the core of the algorithm. Unscented transformation is used to deal with the nonlinear genetic problem of mean and covariance, and then the sample point set (called Sigma point) is determined to achieve approximate probability distribution. The basic principle of unscented transformation is:

[0067] Sample the variable x to generate 2n+1 n-dimensional Sigma points:

[0068]

[0069] Represents the i-th column of the square root of the matrix. and P denote the mean and covariance of the random vector x.

[0070] Calculate the corresponding weights of these sampling points:

[0071]

[0072] The subscript m is the mean of the Sigma point, c is the covariance, the superscript is the sampling point, and the parameter λ = α 2 (n+κ)-n is a scaling parameter, 10 -4 ≤α≤1, the distribution state of the sampling points can be controlled; κ is the secondary scaling factor of the selected parameter and merges the variable β with the variable x. At this time, β≥0, κ=0. When it is in Gaussian distribution, κ=n-3, β=2, which ensures the semi-positive definiteness of the matrix (n+λ)P and obtains the optimal solution.

[0073] Implementation of the Unscented Kalman Filter algorithm:

[0074] First, select the initial value required for filtering:

[0075]

[0076] Untraceable transformation, generating 2n+1 n-dimensional sampling points and calculating the corresponding weights:

[0077]

[0078] Using the state variable of the previous moment, the variable of the current moment is updated, and its covariance P is further calculated. f(x) is the nonlinear state equation function, χ i is the selected 2n+1 Sigma point set, u k Refers to the control variable in the system at the kth sampling moment selected.

[0079] χ i,k|k-1 =f(X i,k-1 ,u k-1 )

[0080]

[0081] Q k To utilize the process noise w k The calculated covariance is, w k Obey Gaussian distribution w k ~N(0,Q k ).

[0082] The estimated value will be output renew:

[0083] y i,k|k-1 =g(X i,k-1 ,u k-1 )

[0084]

[0085] g(x) is the nonlinear observation equation function.

[0086] Change the cross-covariance of state estimates and output estimates, and the covariance of output estimates:

[0087]

[0088] R k-1 is the observation noise v k-1 The covariance matrix of .

[0089] The unscented Kalman filter gain is expressed as:

[0090]

[0091] After solving all the above physical quantities, the error and covariance are updated, and the final result of the unscented Kalman filter algorithm is as follows:

[0092]

[0093]

[0094] In this embodiment, step 2 trains the time series samples using a generalized regression neural network and an unscented Kalman filter to predict the power of the fuel cell hybrid powertrain at the corresponding moment. The prediction value of the generalized regression neural network is P ia , the prediction value using the unscented Kalman filter is P ib , the weighted final prediction power is P hi =c1P ia +c2P ib The fused gray wolf-crested porcupine algorithm is used to train c1 and c2 to predict P hi The minimum power error corresponding to the corresponding moment in the historical data is used as the objective function, and the optimal c1 and c2 are obtained, thereby obtaining a weighted combined intelligent prediction algorithm.

[0095] In this embodiment, in the gray wolf algorithm in step 2, the standard gray wolf algorithm is ranked from high to low as α wolf, β wolf, δ wolf and ω wolf, and this embodiment only uses α wolf, β wolf and δ wolf. The gray wolf algorithm mainly includes four processes, namely, encirclement, hunting, development and exploration.

[0096] (1) Encirclement process:

[0097] The main formula for the process of surrounding prey is:

[0098]

[0099] Where t is the current iteration number, and is the coefficient vector, is the position vector of the prey, is the position vector of the gray wolf, is the distance between the gray wolf and its prey.

[0100] Coefficient vector and The calculation formula is:

[0101]

[0102] In the formula, is a random value between [0,1], is a random value between [0,1], is the convergence factor.

[0103] Convergence Factor The linear decreasing expression from 2 to 0 as the number of iterations decreases is as follows:

[0104]

[0105] Among them, T maxIndicates the maximum number of iterations of the fused grey wolf-crested porcupine algorithm.

[0106] (2) Hunting process:

[0107] When the gray wolves identify the location of the prey, the leader will direct the movement of the other wolves.

[0108] First, calculate the distance between the current search agent and wolf α, wolf β, and wolf δ:

[0109]

[0110] In the formula, and are the current positions of α wolf, β wolf and δ wolf respectively, and is the coefficient vector, and The current search agent The distance between α wolf, β wolf and δ wolf.

[0111] The search agent updates its position under the leadership of α wolf, β wolf, and δ wolf. They represent the update of the search agent position based on α wolf, β wolf, and δ wolf respectively, and the formula is:

[0112]

[0113] Among them, A1, A2 and A3 are coefficient vectors.

[0114] The position update formula for the t+1th gray wolf update is:

[0115]

[0116] Among them, the crown porcupine algorithm first starts the search process from the initial individual set (candidate solution):

[0117]

[0118] Where N' represents the population size, is the i-th candidate solution in the search space, and are the lower and upper limits of the search range, respectively. is a random number between 0 and 1.

[0119] This example uses the third defense strategy of the crested porcupine, which is characterized by an odor strategy. The crested porcupine will secrete a foul chemical odor and spread it around its surrounding area to prevent predators from approaching. This strategy can usually successfully resist predators. The formula that mathematically simulates this behavior is:

[0120]

[0121] In the formula, represents the position of the i-th individual after t+1 iterations, r1, r2, r3 are random numbers between [1, N], N is the population size, t is the number of current iterations, and τ3 is a random number between 0 and 1. It is a parameter used to control the search direction, and its definition equation is:

[0122]

[0123] is a vector containing randomly generated values ​​between 0 and 1, γ t is the defense factor, and its definition equation is:

[0124]

[0125] Among them, t max is the maximum number of iterations, is the odor diffusion factor responsible for the rate of odor diffusion during the optimization process, various values ​​of this factor are between 0.3 and 2.6. If the value of this factor is small, the diffusion rate is very small, so solutions close to random selection are exploited, on the contrary, if these values ​​are high, the odor diffusion rate is high, so solutions far from random selection are exploited. Its definition equation is:

[0126]

[0127] In the formula, represents the objective function, and ε represents a small value to avoid division by zero.

[0128] Used to simulate three possible situations in this strategy:

[0129] (1) When When it is equal to 0, the crested porcupine will stop diffusing its scent because the predator will stop moving due to fear of the crested porcupine, so the distance between the predator and the crested porcupine remains unchanged;

[0130] (2) When When it is equal to 1, the crested porcupine will emit a significant scent due to the presence of a predator nearby;

[0131] (3) When Between 0 and 1, predators keep a safe distance from crested porcupines, so there is no need to significantly release their scent.

[0132] The third defense strategy of the crown porcupine is used to improve the gray wolf algorithm, forming a fusion gray wolf-crowned porcupine algorithm to prevent the gray wolf algorithm from falling into the local optimum. Specifically, the third defense strategy of the crown porcupine is used to update the search agent position with the β wolf and δ wolf of the gray wolf algorithm. and Weighted to avoid the entire gray wolf algorithm falling into the local optimum, the weighted formula of β wolf is:

[0133]

[0134] The weighted formula of δwolf is:

[0135]

[0136] The final position update formula for the t+1th gray wolf update is:

[0137]

[0138] This results in the fused grey wolf-crested porcupine algorithm.

[0139] Embodiment 2

[0140] This embodiment provides a power demand intelligent prediction system for a fuel cell hybrid powertrain, which specifically includes:

[0141] A data acquisition module is configured to: acquire power data at a current moment and multiple previous historical moments to obtain time series samples;

[0142] A power prediction module is configured to: predict power data at a future time by using a prediction model based on the time series samples;

[0143] Among them, the prediction model adopts multiple prediction methods to predict the power data at future moments respectively, and weights the prediction values ​​obtained by multiple prediction methods, and the weighted weights are recursively optimized by using the gray wolf and crowned porcupine fusion algorithm with the goal of minimizing the total prediction error value. The gray wolf and crowned porcupine fusion algorithm adopts the third defense strategy of the crowned porcupine algorithm to update the gray wolf position in the hunting operation of the gray wolf algorithm.

[0144] It should be noted here that each module in this embodiment corresponds to each step in Example 1 one by one, and the specific implementation process is the same, which will not be repeated here.

[0145] Embodiment 3

[0146] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps in the method for intelligently predicting power demand of a fuel cell hybrid powertrain as described in the first embodiment above are implemented.

[0147] Embodiment 4

[0148] This embodiment provides a computer device, such as Figure 2 As shown, it includes a computer-readable storage medium 1003, a processor 1001, a communication interface 1002, and a computer program stored on the computer-readable storage medium 1003 and executable on the processor 1001, wherein the processor 1001, the communication interface 1002, and the computer-readable storage medium 1003 can be connected via a bus or other means. The communication interface 1002 is used to receive and send data, and when the processor 1001 executes the program, the steps in the method for intelligently predicting power demand of a fuel cell hybrid powertrain as described in the first embodiment above are implemented.

[0149] Embodiment 5

[0150] This embodiment provides a computer program product, which is a computer program product containing a computer program. When a processor executes the computer program, the steps in the method for intelligently predicting power demand of a fuel cell hybrid powertrain as described in the first embodiment above are implemented.

[0151] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0152] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0153] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0154] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for intelligently predicting power demand of a fuel cell hybrid powertrain, characterized in that: include: Obtain power data at the current moment and multiple previous historical moments to obtain time series samples; Based on the time series samples, a prediction model is used to predict power data at future moments; The prediction model uses a variety of prediction methods to predict the power data at future moments respectively, and weights the prediction values ​​obtained by the multiple prediction methods, and the weighted weights are recursively optimized by using the gray wolf and crown porcupine fusion algorithm with the goal of minimizing the total prediction error value. The gray wolf and crown porcupine fusion algorithm uses the third defense strategy of the crown porcupine algorithm to update the gray wolf position in the hunting operation of the gray wolf algorithm; The multiple prediction methods include generalized regression neural network and unscented Kalman filter.

2. The method for intelligently predicting power demand of a fuel cell hybrid powertrain according to claim 1, characterized in that: The steps for calculating the total prediction error value include: for each time series sample obtained by dividing the historical data set, using multiple prediction methods to predict the power data respectively, and weighting the predicted values ​​obtained by the multiple prediction methods using the weights of each individual code in the gray wolf population to obtain the predicted value, and calculating the error value between the predicted value and the true value; calculating the sum of the error values ​​between all the predicted values ​​and the true value to obtain the total prediction error value.

3. The method for intelligently predicting power demand of a fuel cell hybrid powertrain according to claim 2, characterized in that: A plurality of time series samples are obtained by sliding sampling a historical data set using a time window, wherein the historical data set is composed of a plurality of historical transient powers including a timestamp.

4. The method for intelligently predicting power demand of a fuel cell hybrid powertrain according to claim 1, characterized in that: The update formula of the gray wolf position in the hunting operation is: in, represents the position of the gray wolf after the t+1th iteration, represents the position of the i-th individual after the t-th iteration, is a random number between [1, N], N is the size of the gray wolf population, t is the number of current iterations, is a random number, is a parameter used to control the search direction. Used to simulate the three situations in the third defense strategy, is the odor diffusion factor, is the number of gray wolf population after the tth iteration r i The location of an individual, is the defense factor, A1, A2 and A3 are coefficient vectors, , and They are Wolf, Wolf and The location of the wolf, , and are the gray wolf position and Wolf, Wolf and The distance between wolves, The weighted updated position of the third defense strategy of the beta wolf and crested porcupine algorithm, The weighted updated position of the third defense strategy of the Delta Wolf and Crested Porcupine algorithm, , and According to the Wolf, Wolf and Wolf updates the search agent's location, is the position of the i-th individual after t+1 iterations.

5. The method for intelligently predicting power demand of a fuel cell hybrid powertrain according to claim 1, characterized in that: The power data is the instantaneous power and its first-order derivative including a timestamp.

6. A power demand intelligent prediction system for a fuel cell hybrid powertrain, characterized in that: include: A data acquisition module is configured to: acquire power data at a current moment and multiple previous historical moments to obtain time series samples; A power prediction module is configured to: predict power data at a future time by using a prediction model based on the time series samples; The prediction model uses a variety of prediction methods to predict the power data at future moments respectively, and weights the prediction values ​​obtained by the multiple prediction methods, and the weighted weights are recursively optimized by using the gray wolf and crown porcupine fusion algorithm with the goal of minimizing the total prediction error value. The gray wolf and crown porcupine fusion algorithm uses the third defense strategy of the crown porcupine algorithm to update the gray wolf position in the hunting operation of the gray wolf algorithm; The multiple prediction methods include generalized regression neural network and unscented Kalman filter.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the method for intelligently predicting power demand of a fuel cell hybrid powertrain as described in any one of claims 1 to 5 are implemented.

8. A computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, characterized in that: When the processor executes the program, the steps in the method for intelligently predicting power demand of a fuel cell hybrid powertrain as described in any one of claims 1-5 are implemented.

9. A computer program product, which is a computer program product containing a computer program, characterized in that When the processor executes the computer program, the steps in the method for intelligently predicting power demand of a fuel cell hybrid powertrain as described in any one of claims 1 to 5 are implemented.

Images