Method for estimating hydrogen desorption rate of hydrogen storage reactor based on BP neural network

Abstract

The invention discloses a method for estimating the hydrogen desorption rate of a hydrogen storage reactor based on a BP neural network, and the method comprises the steps: obtaining the air flow rate, the air temperature, the heat exchange area between the exterior of the reactor and air, and the hydrogen desorption rate in the state of the air convection heat exchange type hydrogen storage reactor, and constructing a parameter data set, inputting the parameter data set into a BP neural network model for training to obtain a hydrogen desorption rate evaluation model; the air velocity, the air temperature and the air heat exchange area are collected, the collected data are input into the hydrogen desorption rate evaluation model, and an estimation result is obtained; wherein the BP neural network model is obtained by training and testing the parameter data set. According to the method, the model precision can be improved, and the hydrogen supply rate can be better estimated, so that the hydrogen supply rate is controlled to improve the system efficiency.

Description

Technical Field

[0001] The invention relates to the technical field of solid-state hydrogen storage, and in particular to a method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network. Background Art

[0002] Hydrogen energy is a clean, efficient, renewable and zero-carbon emission new energy. The hydrogen energy industry chain includes hydrogen production, storage, transportation and use. Among them, hydrogen storage and transportation is the key link connecting the upstream and downstream of the hydrogen energy industry chain. Vigorously developing hydrogen storage technology is the focus and hotspot of research. However, the storage density of hydrogen is low under normal temperature and pressure conditions. Therefore, improving the hydrogen storage density and safety is the key to enhancing the economy of hydrogen energy. Solid-state hydrogen storage does not require a high-pressure environment or low temperature, and has high hydrogen storage density and good safety. Therefore, solid-state hydrogen storage technology has broad application prospects in vehicle-mounted hydrogen storage, hydrogen refueling stations, peak-shaving power stations and distributed energy supply systems. Metal hydride hydrogen storage reactors are the core components of solid-state hydrogen storage technology. During the reversible absorption and desorption of hydrogen by metal hydrides, a large amount of heat exchange will occur, resulting in uneven temperature distribution inside the reactor, affecting the hydrogen storage performance and service life of the reactor. Therefore, effective thermal management of metal hydride hydrogen storage reactors is one of the key technologies to improve their hydrogen storage efficiency and stability. The hydrogen desorption rate is an important indicator of metal hydride hydrogen storage reactors. Better control of the hydrogen desorption rate plays an important role in the thermal management of the hydrogen storage reactor and the efficiency improvement of the entire system.

[0003] In the currently published patent documents, most of them are about optimizing the structure of hydrogen storage reactors, such as using spiral coil heat exchangers or using multi-layer secondary fin structures to improve the thermal management of hydrogen storage reactors. Such optimization can increase the hydrogen release rate but cannot estimate and control the hydrogen release rate in real time. Summary of the invention

[0004] In order to overcome the above shortcomings of the prior art, the present invention proposes a method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network. The hydrogen supply rate of the hydrogen storage reactor is estimated by the acquired air flow rate, air temperature and the heat exchange area between the reactor exterior and the air, and the hydrogen supply rate of the hydrogen storage reactor is adjusted according to the control of the input parameters of air flow rate and air temperature that are convenient to measure to achieve the optimal state of the entire system and improve the system efficiency.

[0005] To achieve the above object, the present invention provides a method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network, comprising:

[0006] Obtaining the air flow rate, air temperature, heat exchange area between the outside of the reactor and the air, and hydrogen release rate of the air convection heat exchange hydrogen storage reactor, constructing a parameter data set, inputting the parameter data set into a BP neural network model for training, and obtaining a hydrogen release rate evaluation model;

[0007] Collecting air velocity, air temperature and air heat exchange area, and inputting the collected data into the hydrogen release rate evaluation model to obtain an estimation result;

[0008] The BP neural network model is obtained after training and testing through the parameter data set.

[0009] Preferably, the parameter data set is constructed by acquiring data on air flow rate, air temperature, heat exchange area between the outside of the reactor and the air, and hydrogen release rate under n states.

[0010] Preferably, the process of constructing the parameter data set further includes normalizing the acquired data to form a matrix, specifically:

[0011]

[0012] Where D is the matrix symbol, v a ,T,S,v (h2) are air flow rate, air temperature, air heat exchange area and hydrogen flow rate respectively, and n is the nth group of data.

[0013] Preferably, the parameter data set includes a training set and a test set, wherein 80% of the data are randomly selected as the training set and 20% of the data are randomly selected as the test set.

[0014] Preferably, the BP neural network model includes an input layer, a hidden layer and an output layer, the input layer is composed of three neuron nodes for inputting parameters; the hidden layer is composed of four neuron nodes; the output layer is composed of one neuron node for outputting hydrogen desorption rate parameters; wherein the excitation function between the input layer and the hidden layer is a sigmoid function, and the excitation function between the hidden layer and the output layer is a purelin function.

[0015] Preferably, the forward propagation method of constructing the BP network neural model is:

[0016]

[0017] In the formula, v (h2) is the theoretical output, w and b are the weight and bias of the BP neural network model respectively, Sigmoid is the activation function, Represents the weight value from the first node in the first layer to the nth node in the second layer, Represents the weight value from the second node in the first layer to the nth node in the second layer, Represents the weight value from the third node in the first layer to the nth node in the second layer, Represents the weight value from the nth node in the second layer to the first node in the third layer, Represents the deviation value of the nth node in the second layer, Represents the bias value of the first node in the third layer.

[0018] Preferably, the activation function of the BP network neural model is:

[0019]

[0020] Where f(x) is the activation function of the BP network neural model.

[0021] Preferably, the back propagation method of the BP network neural model is:

[0022]

[0023]

[0024] In the formula, xite is the learning rate of the BP neural network model, E is the mean square error of the BP neural network model, w1 is the weight value updated after back propagation, and w 1_1 is the weight value before back-propagation update, b1 is the bias value updated after back-propagation, and b 1_1 is the bias value before back-propagation update.

[0025] Preferably, the mean square error expression is:

[0026]

[0027] In the formula, v (h2) For theoretical output, is the actual output.

[0028] Preferably, training and testing the BP neural network model based on the training set and the test set includes:

[0029] (a) Calculate the theoretical output v (h2) ;

[0030] (b) Calculate the mean square error E;

[0031] (c) Calculate the weight and deviation change rate;

[0032] (d) adjusting weights and updating weights;

[0033] Based on the training set, repeat (a)(b)(c)(d) operations until the iteration ends;

[0034] The network test is carried out through the test set. If the error between the estimated result of the test set and the expected result is within the preset range, it proves that the BP neural network model meets the requirements.

[0035] Compared with the prior art, the present invention has the following advantages and technical effects:

[0036] The present invention obtains the air flow rate, air temperature, heat exchange area between the outside of the reactor and the air, and the hydrogen release rate in this state of the air convection heat exchange hydrogen storage reactor, and uses the air flow rate, air temperature and air heat exchange area as input parameters of the BP neural network to estimate the hydrogen release rate. Combined with the nonlinear algorithm of the BP neural network, the model accuracy can be improved, the hydrogen supply rate can be better estimated, and the hydrogen supply rate can be controlled to improve the system efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0038] Figure 1 This is a flow chart of a method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network according to an embodiment of the present invention;

[0039] Figure 2 This is a BP neural network structure diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0040] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0041] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0042] The present invention proposes Figure 1 As shown, a method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network includes:

[0043] The air flow rate, air temperature, heat exchange area between the outside of the reactor and the air, and hydrogen release rate of the air convection heat exchange hydrogen storage reactor are obtained, a parameter data set is constructed, and the parameter data set is input into the BP neural network model for training to obtain a hydrogen release rate evaluation model;

[0044] Collect air velocity, air temperature and air heat exchange area, input the collected data into the hydrogen release rate evaluation model to obtain the estimated result;

[0045] The BP neural network model is obtained after training and testing through the parameter data set.

[0046] Specifically:

[0047] Step 1: Obtain the air flow rate, air temperature, heat exchange area between the outside of the reactor and the air, and hydrogen release rate of the air convection heat exchange hydrogen storage reactor to form an initial parameter data set;

[0048] Step 2: Preprocess the initial parameter data set;

[0049] Step 3: Divide the initial parameter data set into a training set and a test set (parameter data set), and reserve the test data (parameter data set to be tested);

[0050] Step 4: Establish a BP neural network model and obtain the BP neural network model structure;

[0051] Step 5: Import the training set into the initial BP neural network model for training, and use the test set to test the trained BP neural network model;

[0052] Step 6: Estimate the hydrogen desorption rate using the hydrogen desorption rate evaluation model based on the effects of air velocity, air temperature and air heat exchange area on the hydrogen desorption rate.

[0053] Furthermore, step 2 includes normalizing the obtained air flow rate, temperature, heat exchange area and hydrogen release rate, and the different parameter data are expressed as v a ,T,S,v (h2) , and combine different sets of parameters into one matrix:

[0054]

[0055] Where n represents the nth group of data, for example, T n Indicates the value of the air temperature T of the nth set of data.

[0056] In this embodiment, the parameter data set is divided into a training set and a test set, and 80% of the data is randomly selected as the training set and 20% of the data is randomly selected as the test set.

[0057] like Figure 2 The figure shows the structure of the BP neural network model. The model consists of three layers, namely the input layer FA, the hidden layer FB and the output layer FC. The FA layer has three neuron nodes for inputting the parameter va , T and S. The FB layer has four neuron nodes, connecting the FA layer and the FC layer. The FC layer has one neuron node, which is used to output the parameter v (h2) The excitation function between the FA layer and the FB layer is the sigmoid function, and the excitation function between the FB layer and the FC layer is the purelin function.

[0058] The forward propagation mathematical expression of the constructed BP network neural model is as follows:

[0059]

[0060] In the formula, v a , T, S are the input parameters of the BP neural network model, v (h2) is the theoretical output, w and b are the weight and bias of the BP neural network model respectively, and sigmoid is the activation function.

[0061] The mathematical expression of back propagation of BP network neural model is as follows:

[0062]

[0063] In the formula, xite is the learning rate of the BP neural network model; E is the mean square error of the BP neural network model, which is used to determine the direction of error decrease, so that the error decreases along the direction of w and b, to correct w and b to improve the model accuracy; w1 is the weight value updated after back propagation, w 1_1 is the weight value before back-propagation update, b1 is the bias value updated after back-propagation, and b 1_1 is the bias value before back-propagation update.

[0064] The mean square error expression of the BP network neural model is as follows:

[0065]

[0066] Where E is the mean square error, is the actual output.

[0067] The construction of BP neural network includes initializing BP neural network. The weight w and bias b in the forward propagation formula are randomly assigned, and the weights of each layer of network are set to Set the learning rate xite.

[0068] BP neural network training, according to the initial weight calculation theory output v (h2) The actual output of the training set And perform error correction, specifically:

[0069] (a) Calculate the output v (h2) ;

[0070] (b) Calculate the mean square error E;

[0071] (c) Calculate the weight and deviation change rate,

[0072] (d) Adjust weights and update weights.

[0073] The network is trained using 80% of the training set, and operations (a), (b), (c), and (d) are repeated each time until the iteration ends.

[0074] BP neural network test, after the BP neural network training is completed, the network test is performed through the remaining 20% ​​test set. If the error between the estimated result of the test set and the expected result is within a certain range, it proves that the BP neural network model meets the requirements. Otherwise, the training set is used again to train the BP neural network model to reduce the error.

[0075] The resulting BP neural network model (i.e., hydrogen release rate evaluation model) can be used according to the input parameter v a , T, S, predict the hydrogen supply rate in the hydrogen storage reactor, and improve the efficiency of the system by controlling the hydrogen supply reaction rate.

[0076] The present invention obtains the air flow rate, air temperature, heat exchange area between the outside of the reactor and the air, and the hydrogen release rate in this state of the air convection heat exchange hydrogen storage reactor, and uses the air flow rate, air temperature and air heat exchange area as input parameters of the BP neural network to estimate the hydrogen release rate. Combined with the nonlinear algorithm of the BP neural network, the model accuracy can be improved, the hydrogen supply rate can be better estimated, and the hydrogen supply rate can be controlled to improve the system efficiency.

[0077] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network, characterized in that: include: Obtaining the air flow rate, air temperature, heat exchange area between the outside of the reactor and the air, and hydrogen release rate of the air convection heat exchange hydrogen storage reactor, constructing a parameter data set, inputting the parameter data set into a BP neural network model for training, and obtaining a hydrogen release rate evaluation model; Collecting air velocity, air temperature and air heat exchange area, and inputting the collected data into the hydrogen release rate evaluation model to obtain an estimation result; The BP neural network model is obtained after training and testing through the parameter data set.

2. The method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network according to claim 1, characterized in that: The parameter data set is constructed by acquiring data on air flow rate, air temperature, heat exchange area between the outside of the reactor and the air, and hydrogen release rate under n states.

3. The method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network according to claim 2, characterized in that: The process of constructing the parameter data set also includes normalizing the acquired data to form a matrix, specifically: Where D is the matrix symbol, v a ,T,S,v (h2) are air flow rate, air temperature, air heat exchange area and hydrogen flow rate respectively, and n is the nth group of data.

4. The method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network according to claim 1, characterized in that: The parameter data set includes a training set and a test set, wherein 80% of the data are randomly selected as the training set and 20% of the data are randomly selected as the test set.

5. The method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network according to claim 1, characterized in that: The BP neural network model includes an input layer, a hidden layer and an output layer, wherein the input layer is composed of three neuron nodes for inputting parameters; the hidden layer is composed of four neuron nodes; the output layer is composed of one neuron node for outputting hydrogen release rate parameters; wherein the excitation function between the input layer and the hidden layer is a Sigmoid function, and the excitation function between the hidden layer and the output layer is a purelin function.

6. The method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network according to claim 5, characterized in that: The method for constructing the forward propagation of the BP network neural model is: In the formula, v (h2) is the theoretical output, w and b are the weight and bias of the BP neural network model respectively, Sigmoid is the activation function, Represents the weight value from the first node in the first layer to the nth node in the second layer, Represents the weight value from the second node in the first layer to the nth node in the second layer, Represents the weight value from the third node in the first layer to the nth node in the second layer, Represents the weight value from the nth node in the second layer to the first node in the third layer, represents the deviation value of the nth node in the second layer, Represents the bias value of the first node in the third layer.

7. The method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network according to claim 6, characterized in that: The activation function of the BP network neural model is: Where f(x) is the activation function of the BP network neural model.

8. The method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network according to claim 7, characterized in that: The back propagation method of the BP network neural model is: In the formula, xite is the learning rate of the BP neural network model, E is the mean square error of the BP neural network model, w1 is the weight value updated after back propagation, and w 1_1 is the weight value before back-propagation update, b1 is the bias value updated after back-propagation, and b 1_1 is the bias value before back-propagation update.

9. The method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network according to claim 8, characterized in that: The mean square error expression is: In the formula, v (h2) For theoretical output, is the actual output.

10. The method for estimating the hydrogen release rate of a hydrogen storage reactor based on a BP neural network according to claim 9, characterized in that: The BP neural network model is trained and tested based on the training set and the test set, including: (a) Calculate the theoretical output v (h2) ; (b) Calculate the mean square error E; (c) Calculate the weight and deviation change rate; (d) adjusting weights and updating weights; Based on the training set, repeat (a)(b)(c)(d) operations until the iteration ends; The network test is carried out through the test set. If the error between the estimated result of the test set and the expected result is within the preset range, it proves that the BP neural network model meets the requirements.

Images