Thermal working condition graph searching method for thermal management system of thermal chemical heat storage material chip

Abstract

The invention discloses a thermal working condition graph lookup method for a thermal management system of a thermal chemical heat storage material chip, which comprises the following steps of: calculating a thermal working condition of the thermal management system of the thermal chemical heat storage material chip based on a finite element model, and constructing a thermal working condition data set; carrying out dimension reduction processing on the thermal working condition data set based on a power function product form; the thermal working condition data set is subjected to dimension reduction processing, then feature time of each data point is calculated, the data set is divided into a plurality of data sets with mapping relations, and the mapping relations are identified based on a back propagation neural network; and arranging graph lines of the mapping relation to form a query graph. Based on the generated chart search strategy, the working temperature and progress of the heat storage material can be quickly estimated, system parameters and material characteristics are predicted according to temperature requirements, and the learning cost and the calculation time are greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal condition charting, and particularly to a method for charting thermal conditions of a thermal management system for a thermochemical energy storage material chip. Background Art

[0002] When electronic devices operate, heat generation often occurs. In some scenarios, the heat generated by the chip will seriously affect the service life and working efficiency of the device. In addition to common active cooling methods (such as forced air cooling, heat pipe cooling, and liquid cooling), passive cooling using energy storage materials can meet the working conditions of some high-power devices with a closed thermal environment and large space limitations, and has gradually become the preferred cooling method for some sealed electronic devices.

[0003] Among energy storage materials, thermochemical energy storage has a larger energy storage density, far greater than sensible heat energy storage and phase change energy storage. A common thermochemical energy storage material used for electronic device heat dissipation is hydrated salt. Its endothermic process includes sensible heat energy storage and solid-liquid phase change energy storage, and an endothermic thermal decomposition reaction will also occur to remove crystal water molecules. Therefore, its energy storage density is 4-5 times larger than that of common phase change energy storage materials. For example, researchers have developed a sodium acetate trihydrate / expanded graphite thermochemical energy storage material with an energy storage density of 793.4 J / g, while the energy storage density of a typical phase change material is only 180 J / g. With the advantage of high energy storage density, thermochemical energy storage materials have begun to be used in chip thermal management systems.

[0004] The chip thermal management system based on thermochemical energy storage materials has achieved good results in the laboratory stage, which means that a prediction method for thermal management conditions needs to be established to further guide its design in industrial applications. Currently, the finite element method is the most common method. This method aims to couple the kinetic models of thermal decomposition and phase change of thermochemical energy storage materials with the heat conduction model, so as to consider the endothermic effect of the material into the heat conduction process. This approach relies on mature finite element calculation software and sufficient theoretical equations, and exhibits excellent accuracy. However, the disadvantages of this method are also obvious: high learning cost, large workload, and long time consumption.

[0005] Therefore, there is an urgent need for an efficient thermal condition charting technology that integrates machine learning, so as to maintain the accuracy of the training data generated by the finite element method, enable researchers to quickly estimate the energy storage progress and temperature when thermochemical energy storage materials are working, and improve the design efficiency. Summary of the Invention

[0006] In order to overcome the defects and deficiencies of the existing technology, the present invention provides a method for looking up the thermal conditions of a thermal management system for a chip of a thermochemical energy storage material, which solves the problems of high learning cost, large workload and long time consumption in predicting the thermal management conditions of the chip thermal management system of the thermochemical energy storage material. The present invention generates a sufficient amount of training data with a wide range through finite element simulation calculation. Each data point includes information such as temperature, position, progress quantity, thermal physical properties of materials, heat sources, etc., and the result has good reliability. Based on the generated chart lookup strategy, the working temperature and progress of the energy storage material can be quickly estimated, and the system parameters and material characteristics can be predicted according to the temperature requirements, greatly reducing the learning cost and calculation time.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] The present invention provides a method for looking up the thermal conditions of a thermal management system for a chip of a thermochemical energy storage material, including the following steps:

[0009] Calculate the thermal conditions of the thermal management system for the chip of the thermochemical energy storage material based on a finite element model, and construct a thermal condition data set;

[0010] Perform dimensionality reduction processing on the thermal condition data set based on the product form of power functions;

[0011] After the dimensionality reduction processing of the thermal condition data set, calculate the characteristic time of each data point, and split it into data sets with multiple mapping relationships, and identify the mapping relationships based on a backpropagation neural network;

[0012] Arrange the graph lines of the mapping relationships to form a query graph.

[0013] As a preferred technical solution, to calculate the thermal conditions of the thermal management system for the chip of the thermochemical energy storage material based on a finite element model, specifically, the reaction kinetic equation of the thermochemical energy storage material is used to construct the control equation of the finite element model, and the finite element model calculates the working process conditions of the thermal management system for the chip of the thermochemical energy storage material according to the thermal physical property parameters and boundary conditions of the material.

[0014] As a preferred technical solution, the control equation is expressed as:

[0015]

[0016] Among them, T represents the temperature quantity, T s represents the solidus temperature, T l represents the liquidus temperature, t represents time, c represents the volume heat capacity of the energy storage material, λ represents the thermal conductivity of the energy storage material, Q pc represents the phase change heat absorption of the energy storage material, Q r represents the heat absorption of the thermal reaction of the energy storage material, ΔH pc represents the enthalpy change per unit volume of phase change, ΔH rrepresents the enthalpy change per unit volume of the thermal reaction, β is the percentage progress of the material phase change, α is the percentage progress of the material thermal reaction, and E a is the apparent activation energy of the material thermal reaction, and R is the thermodynamic constant.

[0017] As a preferred technical solution, a thermal condition dataset is constructed, specifically including:

[0018] Based on the Monte Carlo method, randomly select model parameters for finite element calculation, change the material physical property parameters and boundary conditions, calculate the change of the working conditions, and construct the thermal condition dataset.

[0019] As a preferred technical solution, based on the product form of power functions, the thermal condition dataset is reduced in dimension, specifically including:

[0020] Through the finite element model, when the material physical properties, material quantity, and heat source power are known, the heat storage process and temperature of the material at a certain position at a certain moment are deduced. The thermal condition dataset records the influence relationship of temperature with time, position, external heat source, and material physical properties, expressed as:

[0021] T(x,t,λ,c,q,L,ΔH)

[0022] where x is the position, t is the time, λ is the thermal conductivity of the heat storage material, c is the volume heat capacity of the heat storage material, q represents the heat flux density of the heat source chip, L represents the material volume, and ΔH represents the enthalpy change per unit volume;

[0023] For the relationship T = f(X1,X2) determined by the binary function, calculate the product of the power functions of the two variables, combine the two independent variables into one, and find the unary function describing the original binary relationship, that is, for any X1, X2, there exists a scaling coefficient n such that:

[0024]

[0025] Calculate the minimum value of the loss function to obtain the scaling coefficient n. The loss function is expressed as:

[0026]

[0027] where Abs represents taking the absolute value, taking the temperature T as the dependent variable, and m represents the number of dependent variables.

[0028] As a preferred technical solution, when the loss function takes the minimum value, the value of the loss function represents the average relative error after being described by the unary function, and t, λ, c, q, L, ΔH are combined successively;

[0029] Reduce the parameters affecting the temperature to two, expressed as:

[0030] T(x,tch )

[0031] Among them, t ch represents the characteristic time;

[0032] Reduce the parameters affecting the heat storage process quantity of the material to two, expressed as:

[0033] α(x, t ch ), β(x, t ch ), γ(x, t ch )

[0034] Among them, β is the material phase change percentage process, α is the material thermal reaction percentage process, and γ represents the process quantity.

[0035] As a preferred technical solution, after the heat condition data set is dimensionally reduced, calculate the characteristic time of each data point and split it into data sets with multiple mapping relationships, specifically split into (t ch , α), (t ch , γ), (t ch , β), (α, T), (γ, T), (β, T) six groups of data sets with mapping relationships.

[0036] As a preferred technical solution, the process quantity γ is expressed as:

[0037]

[0038] Among them, t b,r represents the time point when the thermochemical reaction starts, and t e,pc represents the time point when the solid-liquid phase change ends.

[0039] As a preferred technical solution, arrange the graph lines of the mapping relationship to form a query graph, and find the temperature and energy storage progress of a certain position of the heat storage material at time t according to the query graph, specifically including:

[0040] Calculate the characteristic time t ch ;

[0041] Find the calculated characteristic time t ch at the abscissa, map the graph line to the process quantity, and find out whether the heat storage material is in the phase change stage or the endothermic reaction stage, and the corresponding progress percentage;

[0042] Map the process quantity back to the other section of the abscissa to find the temperature of the heat storage material.

[0043] As a preferred technical solution, arrange the graph lines of the mapping relationship to form a query graph, and estimate the numerical range that the thermophysical properties of the material and the heat source power conform to according to the query graph, specifically including:

[0044] Locate the known temperature on the temperature axis and correspond it to the ordinate of the process quantity, thereby determining whether the heat storage material is in the phase change stage or the endothermic reaction stage, and the corresponding progress percentage;

[0045] According to the process quantity and the known position, correspond to the characteristic time t in the corresponding graph line ch ;

[0046] After obtaining the characteristic time t ch Based on the characteristic time t ch Calculate the numerical ranges that the thermal conductivity, heat capacity, enthalpy value, and volume heat source power of the heat storage material conform to according to the calculation formula of t.

[0047] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0048] (1) The present invention generates a sufficient amount and wide range of training data through finite element simulation calculations. Each data point includes information such as temperature, position, process quantity, thermal physical properties of materials, heat sources, etc., and the results have good reliability.

[0049] (2) The present invention uses the form of the product of power functions to combine variables to achieve dimensionality reduction. This dimensionality reduction is equivalent to eliminating the influence of heat capacity, thermal conductivity, heat source power, and material volume on the rate difference of the heat transfer process, which helps to extract the principal components for visualizing the machine learning results.

[0050] (3) The present invention uses a backpropagation neural network to identify the relationship between the characteristic time, energy storage process, and temperature from the training dataset, and integrates and plots them in a graph, which can display the training results in a visual manner, enabling researchers to intuitively understand how the temperature of the heat storage material and the energy storage process change over time are affected by the thermal physical properties of the material and the external heat source.

[0051] (4) Based on the generated chart lookup strategy, the present invention can estimate the working temperature and progress of the heat storage material within a few minutes using a scientific calculator and a chart, and predict the system parameters and appropriate material properties according to the temperature requirements, rather than relying on hours of simulation calculations, which greatly reduces the learning cost and calculation time. Description of the Drawings

[0052] Figure 1 is a schematic flow chart of the thermal condition chart lookup method for the thermochemical heat storage material chip thermal management system of the present invention;

[0053] Figure 2 is a schematic diagram of the mapping relationship in the chart lookup strategy of the present invention;

[0054] Figure 3 is a schematic diagram of the chart lookup steps of the present invention;

[0055] Figure 4This is a schematic diagram of the computational domain for calculating the thermal conditions of SAT / EG materials in the chip thermal management system using finite element calculations in the present invention;

[0056] Figure 5 (a) Schematic diagram showing the variation of Loss(n) with n when the variable investigated in the present invention is c;

[0057] Figure 5 (b) Schematic diagram showing the variation of Loss(n) with n when the variable investigated in the present invention is λ;

[0058] Figure 5 (c) Schematic diagram showing the variation of Loss(n) with n when the variable investigated in the present invention is q;

[0059] Figure 5 (d) Schematic diagram showing the variation of Loss(n) with n when the variable investigated in the present invention is L;

[0060] Figure 5 (e) Schematic diagram showing the variation of Loss(n) with n when the variable investigated in the present invention is ΔH;

[0061] Figure 6 (a) Regression diagram of the mapping of neural network training (t ch , α) in the present invention;

[0062] Figure 6 (b) Error distribution diagram of the mapping of neural network training (t ch , α) in the present invention;

[0063] Figure 6 (c) Regression diagram of the mapping of neural network training (t ch , β) in the present invention;

[0064] Figure 6 (d) Error distribution diagram of the mapping of neural network training (t ch , β) in the present invention;

[0065] Figure 6 (e) Regression diagram of the mapping of neural network training (α, Τ) in the present invention;

[0066] Figure 6 (f) Error distribution diagram of the mapping of neural network training (α, Τ) in the present invention;

[0067] Figure 6 (g) Regression diagram of the mapping of neural network training (γ, Τ) in the present invention;

[0068] Figure 6 (h) Error distribution diagram of the mapping of neural network training (γ, Τ) in the present invention;

[0069] Figure 7 The operating condition query diagram of the chip thermal management system for SAT / EG formed by a neural network. Specific implementation manners

[0070] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0071] Embodiment 1

[0072] As Figure 1 shown, this embodiment provides a method for querying the thermal condition diagram of a chip thermal management system for a thermochemical heat storage material, including the following steps:

[0073] S1: Calculating the thermal condition of the chip thermal management system for the thermochemical heat storage material based on a finite element model. The control equations involved are:

[0074]

[0075] wherein, T represents the temperature quantity, T s represents the solidus temperature, T l represents the liquidus temperature, and the units are both °C. t is the time, and the unit is s. c is the volume heat capacity of the heat storage material, and the unit is J / (m 3 ·K). λ is the thermal conductivity of the heat storage material, and the unit is W / (m·K). Q pc is the phase change heat absorption of the heat storage material, and Q r represents the heat absorption of the heat reaction of the heat storage material, and the units are both W / m 3 . This is the heat conduction equation of the solid material with two additional heat source terms to consider the heat absorption effect of the heat storage material. ΔH pc represents the enthalpy change per unit volume of phase change, and ΔH r represents the enthalpy change per unit volume of heat reaction, and the units are both J / m 3 . β is the phase change percentage process of the material, and α is the heat reaction percentage process of the material; E a is the apparent activation energy of the heat reaction of the material, and R is the thermodynamic constant.

[0076] In this embodiment, β is described by a common melting-solidification model for the solid-liquid phase change process. This model is widely used in the simulation of phase change materials. The differential equation describing α is the process of the decomposition reaction, and its form is obtained by measuring the thermochemical reaction kinetics. There is a sensible heat storage stage in the heat storage process of the thermochemical heat storage material where the phase change is completed but the thermal decomposition reaction has not started. The process quantity of this stage is defined by the time ratio:

[0077]

[0078] Among them, t b,r represents the time point when the thermochemical reaction starts, and t e,pc represents the time point when the solid-liquid phase change ends. The above equation is used to calculate the temperature and energy storage process when the thermochemical energy storage material is used for chip thermal management. The calculation domain can be simplified to a one-dimensional single-sided heat source case (the heat flux density of the heat source chip is q), and the fact that the material works in a closed space means that the heat exchange between the energy storage system and the environment is very small, which can be regarded as an adiabatic boundary condition. The initial conditions are set as the global initial temperature of 298.15 K (25 °C), and the progress variables α and β are both 0;

[0079] The control equation of this embodiment is based on heat conduction, adds heat source terms for the phase change and endothermic reaction of the energy storage material, further expands the heat source terms respectively, and introduces their respective percentage progress to describe. The established finite element model can calculate the working conditions of the thermochemical energy storage material chip thermal management system, such as temperature and progress variables, according to the thermal physical properties of the material and the boundary conditions.

[0080] In this embodiment, the reaction kinetic equation of the thermochemical energy storage material is used to construct the control equation of the finite element model, and calculations are carried out to generate training data, so that the final obtained chart lookup method still retains the advantage of high accuracy of the finite element simulation method;

[0081] Specifically, after the finite element model is built, the Monte Carlo method is used to randomly select model parameters for finite element calculations. The changeable model parameters include the material thermal conductivity λ, the material heat capacity c, the material enthalpy value ΔH, the material volume L, and the heat flux density q of the heat source chip. This process is repeated multiple times until enough training data sets are collected;

[0082] S2: Compress the high-dimensional data obtained from the finite element calculation to low-dimensional data based on the product form of power functions, and introduce a dimensionality reduction method to merge many variables into one, which helps to extract the main components for visualizing the graph lines. Specifically, it includes:

[0083] Through the finite element model, it is possible to deduce the energy storage process (i.e., α and β) and temperature T of the material at a certain position and time when the material physical properties, material quantity, and heat source power are known. Therefore, by continuously changing the material physical property parameters and boundary conditions, and calculating the changes in the working conditions, a sufficient number of high-dimensional data sets are formed. The data sets record the influence relationship of temperature with time, position, external heat source, and material physical properties, which is expressed as:

[0084] T(x, t, λ, c, q, L, ΔH)

[0085] Among them, T is the material temperature, x is the position, t is the time, λ is the thermal conductivity of the energy storage material, c is the volume heat capacity of the energy storage material, q represents the heat flux density of the heat source chip, L represents the material volume, and ΔH represents the enthalpy change per unit volume;

[0086] Due to the excessive number of affected variables, the results cannot be presented in the form of a planar graph line. Therefore, a dimensionality reduction method needs to be introduced to combine numerous variables into one while trying to retain the original information. In this embodiment, the power function product form is used to combine variables. For the relationship T = f(X1, X2) determined by a binary function, find a unary function to describe it. Thus, the originally two independent variables are combined into one, and its form is the power function product of the original two variables. It is expected that the unary function can describe the original binary relationship as correctly as possible, that is, for any X1, X2, there exists n such that:

[0087]

[0088] Therefore, a number of groups of X1, X2 (where i represents the i-th group) that make the function values all equal to T are collected from the finite element calculation, denoted as:

[0089]

[0090] After being combined into a unary function, there should be:

[0091]

[0092] In this thermal management calculation, the temperature T is monotonically increasing. Then, one dependent variable in the unary function can only be mapped to one independent variable. Therefore, the independent variables in the above formula should all be equal, that is:

[0093]

[0094] Due to the loss of information, the equal sign in the above formula cannot be achieved, but the optimal n value can be found by continuously adjusting n to make them as equal as possible. The scaling coefficient n is determined by finding the minimum value of the following loss function:

[0095]

[0096] The construction method of the loss function in the above formula is similar to finding the variance, which measures the equality degree of the independent variables. Abs represents taking the absolute value. Regarding the temperature T as the dependent variable y here, m y values are selected, and s groups of (yX1, y X1) (the summation indices i and j are written at the lower left corner) that make the function values all equal to y are collected for calculation. Thus, by selecting a value of n, the corresponding loss function Loss(n) can be calculated. By traversing the values of n, the n that makes Loss(n) the smallest can be regarded as the final result.

[0097] Thus, the scaling coefficient n corresponding to the product of power functions by combining two variables is found. After multiple pairwise combinations of high-dimensional data, finally, t, λ, c, q, L, and H are successively combined into one variable, which is called the characteristic time t. ch ;

[0098] When Loss(n) reaches the minimum value, the value of n at this time has the least information loss during the combination process, and the value of the loss function represents the average relative error after being described by a unary function. In the present invention, t, λ, c, q, L, and H are successively combined by this method, and 5 n values will be generated in sequence, which are numbered as n1, n2, n3, n4, and n5 respectively. After combination, it becomes one variable: the characteristic time t. ch , and it is stipulated that it is dimensionless, that is:

[0099]

[0100] Therefore, the parameters affecting the temperature are reduced to two, that is:

[0101] T(x,t ch )

[0102] Similarly, the heat storage process quantity of the material is also affected by two parameters, that is, α(x,t ch ), β(x,t ch ), γ(x,t ch ).

[0103] S3: Extract the key mappings of low-dimensional data. After the high-dimensional data is dimensionally reduced, calculate the characteristic time t of each data point ch , and split it into six groups of datasets with mapping relationships: (t ch ,α), (t ch ,γ), (t ch ,β), (α,T), (γ,T), (β,T). Use a backpropagation neural network to identify these mapping relationships, and finally arrange the graph lines of the mapping relationships in one graph to form a chart method. The formed graph can integrate thermal condition information such as the thermal properties of the heat storage material, heat source heat flow, operating time, temperature, and heat storage process. By calculating the characteristic time, the thermal condition information of the material can be known according to the graph line. As Figure 2 shown, the graph divides the abscissa into two segments: t ch and temperature, and divides the ordinate into three segments to present three process quantities, so that the six groups of mapping relationships can be plotted in it.

[0104] In this embodiment, the backpropagation neural network is a multi-layer feedforward neural network with self-learning ability and is commonly used for problems such as classification and regression. Its core idea is to gradually adjust the weights and biases of the network through the processes of forward propagation and backpropagation, thereby minimizing the error. It has an input layer, a hidden layer, and an output layer. Briefly speaking, data is operated on in each layer, and the results of the previous layer are passed to the next layer for operation until the prediction result is obtained at the output layer. Specifically, in the present invention, the hidden layer uses the Tan-Sig transfer function, and the output layer uses the Purelin transfer function;

[0105] S4: Given the physical properties of the material and the heat source power, such as Figure 3 As shown, according to the drawn graph, the temperature and energy storage progress of a certain position of the heat storage material at time t can be found according to the following steps:

[0106] (1) Calculate the characteristic time t ch ;

[0107] (2) Find the calculated characteristic time t on the abscissa ch , and map the graph line to the progress quantity, then it can be found whether the heat storage material is in the phase change stage or the endothermic reaction stage, and the corresponding progress percentage;

[0108] (3) Map the progress quantity back to the other section of the abscissa to find the temperature of the heat storage material.

[0109] At the same time, this graph can also be used to estimate the numerical range that the thermal physical properties of the material and the heat source power conform to under the condition of knowing the temperature of the material at a certain moment. The search process is opposite to the above, which are respectively:

[0110] (1) Find the known temperature on the temperature coordinate axis, map it to the progress quantity ordinate, and thus judge whether the heat storage material is in the phase change stage or the endothermic reaction stage, and the corresponding progress percentage;

[0111] (2) According to the progress quantity and the known position, map it to the characteristic time t in the corresponding graph line ch ;

[0112] (3) After knowing the characteristic time t ch , according to the calculation formula of the characteristic time t ch , the numerical range that the thermal conductivity, heat capacity, enthalpy value, and volume heat source power of the heat storage material conform to can be calculated, so as to select materials under specific temperature requirements.

[0113] Embodiment 2

[0114] As Figure 4 shown, this embodiment realizes the collection of finite element calculation data of the sodium acetate trihydrate / expanded graphite thermochemical heat storage material in the chip thermal management system and the dimensionality reduction processing of the data;

[0115] In this embodiment, the sodium acetate trihydrate / expanded graphite material (hereinafter abbreviated as SAT / EG in English) used is a composite hydrated salt thermochemical energy storage material. Since the compounding ratio of the materials can be changed, the thermophysical properties of SAT / EG can be regulated within a certain range. However, regarding the problem of which thermophysical properties of SAT / EG can achieve good enough effects in the chip thermal management system, the previous solution was to substitute multiple sets of thermophysical parameters into the finite element model for calculation one by one to find the optimal material thermophysical properties, which would take a large amount of calculation time. This embodiment will establish a chart lookup method to solve this problem.

[0116] First, establish a finite element model to calculate the thermal conditions of SAT / EG in the chip thermal management system to generate the required training data. The control equations of the model are as follows:

[0117]

[0118]

[0119] The calculation domain of the chip thermal management system is simplified to a one-dimensional single-sided heat source case. The relevant calculation settings adopted in this embodiment for finite element calculation are shown in Table 1 below:

[0120] Table 1 Relevant settings table for finite element calculation

[0121]

[0122] Among the model parameters adopted in this embodiment for finite element calculation, the quantities related to the thermophysical properties of the materials (such as thermal conductivity λ, material heat capacity c, material enthalpy value ΔH, material volume V, heat flux density q of the heat source chip) will change their values for calculation multiple times, with the aim of widely collecting thermal condition data under different material physical properties; while the parameters related to the material reaction kinetic equation (such as pre-exponential factor A, apparent activation energy E a , phase change solidus line T s , phase change liquidus line T l ) remain unchanged throughout the data generation process because the phase change and thermal decomposition kinetic properties of SAT / EG do not change with the compounding ratio and are completely determined by the properties of the sodium acetate trihydrate molecules. The values of these parameters are shown in Table 2. It should also be noted that when calculating, the x-axis in the one-dimensional direction has a unit, with a range from 0 - L, and when processing the calculation results, the position x with a unit is rewritten in dimensionless form, that is, the percentage within the range of 0 - L.

[0123] Table 2 Model parameter table for finite element calculation

[0124]

[0125] The training data of the neural network is generated by a finite element model. Using the Monte Carlo method, the model parameters are randomly selected for 100 times of finite element calculations (the random value ranges of each parameter are shown in Table 2). The time range of each calculation is 0 - 3000 s, and these calculation results are statistically analyzed;

[0126] Among the various calculation results generated by the finite element model, the heat storage process (i.e., α and β) and temperature T of the SAT / EG material at a certain position at a certain moment are affected by material physical properties, material quantity, and heat source power. It is necessary to introduce a dimensionality reduction method to combine numerous variables into one and retain the original information as much as possible. In this embodiment, the power function product form is used to combine variables. For the relationship T = f(X1, X2) determined by a binary function, a unary function is found to describe it. Thus, the originally two independent variables are combined into one, and its form is the power function product of the original two variables. The scaling coefficient n is determined by finding the minimum value of the following loss function:

[0127]

[0128] Thus, the scaling coefficient n corresponding to combining two variables into a power function product is found. Based on t, after multiple pairwise combinations of high-dimensional data, finally t, λ, c, q, L, ΔH are successively combined into one variable called the characteristic time t ch . As Figure 5 (a)- Figure 5 (e) shows, it can be seen the variation of Loss(n) with n (at different x positions) for different test variables. The n value when Loss(n) is the smallest is taken as the exponent during the combination of this variable, as shown in Table 3:

[0129] Table 3 The n values and relative errors obtained for each variable during dimensionality reduction

[0130]

[0131] The n value that maximally retains information when combining each variable in the form of a power function product is obtained. Therefore, the combination result is as follows:

[0132] t ch = tc -0.4 λ 0.3 qL -1 ΔH -0.6

[0133] It should be noted that calculations must be carried out in the units given in Table 3. This dimensionality reduction is equivalent to eliminating the influence of heat capacity, thermal conductivity, heat source power, and material volume on the rate difference of the heat transfer process, enabling different physical property parameters to be converted to the same standard for reference. Therefore, the combined variable is called the characteristic time t ch, it is stipulated that it is dimensionless. Thus, the heat storage process of the material (i.e., α and β) and the temperature T are only affected by two independent variables, namely the position x and the characteristic time t ch .

[0134] Example 3

[0135] In this example, a mapping strategy for the sodium acetate trihydrate / expanded graphite thermochemical heat storage material in the chip thermal management system is established using a machine learning method;

[0136] In this example, each data point in the finite element calculation results includes information such as temperature, position, process quantity, material thermal properties, heat source, etc. In the example, the heat storage process of the material (i.e., α and β) and the temperature T have been made to be affected only by two independent variables through a high-dimensional data reduction method in the form of a product of power functions. Therefore, after neural network training, the thermal condition data can be plotted in the form of a graph for reference;

[0137] Calculate the characteristic time t of each data point ch , and split it into six groups of mapping relationship data sets: (t ch ,α), (t ch ,γ), (t ch ,β), (α,T), (γ,T), (β,T). A backpropagation neural network is used to identify these mapping relationships. The backpropagation neural network is a multi-layer feedforward neural network trained according to the error backpropagation algorithm. It has an input layer, a hidden layer, and an output layer. Simply put, data is operated on in each layer, and the result of the previous layer is passed to the next layer for operation until the prediction result is obtained at the output layer. Specifically, the hidden layer uses the Tan-Sig transfer function, the output layer uses the Purelin transfer function, and the number of hidden layers is 11. Essentially, the backpropagation neural network algorithm is to compare the error between the training data result and the prediction result, and continuously adjust the calculation parameters within each layer to make the error as small as possible. This objective function for measuring the error is the mean square error (MSE). The Levenberg-Marquardt training algorithm is used to adjust the calculation parameters within the layer to make the MSE quickly converge to the minimum value. The final fitting effect is judged by the R value, as follows:

[0138]

[0139] where N is the number of data points in the data set, X exp(i) is the i-th data in the data set, and X pred(i) is the i-th data predicted by the neural network;

[0140] As Figure 6 (a)- Figure 6 (h) shows, the backpropagation neural network is used to train (t ch, α) and (t ch , β) two sets of mapping relationships, and the training results are shown in Figure 6 (a) to Figure 6 (d). The R values are 0.99988 and 0.99955 respectively, indicating that good training results have been achieved. Since the mappings (α, T) and (γ, T) do not involve dimensionality reduction processing, Sobol sensitivity analysis is used before neural network training to evaluate the influence degree of other parameters on the mapping relationships. The results show that the influence of q needs to be considered when training the mapping (α, T) of the neural network, and the influence of x needs to be considered when training the mapping (γ, T). Therefore, in fact, the relationships of (α, T, q) and (γ, T, x) are trained separately using a BP neural network, and the training results are shown in Figure 6 (e) to Figure 6 (h), and the R values of the two trainings are 0.99983 and 0.99956 respectively;

[0141] In this embodiment, a working condition query diagram for the chip thermal management system for SAT / EG is drawn using the training results of the neural network, as shown in Figure 7 . The right half of the diagram is the process quantity diagram, from which the heat storage stage of the material and the percentage completed can be found through the characteristic time. The left half of the diagram is the temperature diagram, which can correspond the percentage process quantity of the stage to the temperature.

[0142] Example 4

[0143] In this embodiment, the use of the query diagram strategy for the thermochemical energy storage material of sodium acetate trihydrate / expanded graphite in the chip thermal management system is realized. Consider the following design scenario: In a working condition that can be simplified to a one-dimensional single-sided heat source, the thermophysical properties of the SAT / EG material and the heat flux density of the heat source are given as shown in Table 4 below. Use the constructed query diagram to find the temperature at the center of the material at t = 1200 s.

[0144] Table 4 Properties of SAT / EG Material and Heat Source

[0145]

[0146] The steps to solve this problem using the query diagram strategy are as follows:

[0147] (1) Calculate the characteristic time t ch . In order to substitute into the formula to calculate the characteristic time, first, the physical quantities need to be converted to the units specified in Table 4 according to the known conditions. The following demonstrates one by one:

[0148] The c in the formula is the heat capacity per unit volume, so c = 1810 * 1100 J·m -3 ·K -1 = 1.991 * 10 6 J·m -3 ·K-1 ;

[0149] In the formula, λ has the same meaning and unit as the thermal conductivity in Table 4, λ = 7.6 W·m -1 ·K -1 ;

[0150] In the formula, q is the heat flux density of the heat source, that is, the heating power divided by the cross-sectional area, q = 45 / 12.56 W / cm 2 = 3.582*10 4 W·m -2 ;

[0151] In the formula, L is the length of the material in the heat transfer direction, L = 75.36 / 12.56 cm = 0.06 m.

[0152] In the formula, ΔH is the total enthalpy value (per unit volume) of the heat storage material, ΔH = (199.2 + 492)*1000*1100 J·m -3 = 7.603*10 8 J·m -3 .

[0153] So substituting into the formula gives:

[0154] t ch = tc -0.4 λ 0.3 qL -1 ΔH -0.6 = 18.67

[0155] (2) Locate the calculated characteristic time t ch = 18.67 on the right half of the abscissa. Since the position investigated in this embodiment is the center of the material, find the curve of x = 0.5 and look up the corresponding progress variable, that is, α = 0.21. This means that at this moment, the center of the material is in the thermal decomposition reaction stage and 21% has proceeded.

[0156] (3) Correlate the progress variable back to the left half of the abscissa to find the temperature of the heat storage material. Since in this embodiment q = 3.582*10 4 W·m -2 , first find the temperature by looking up the curve of q = 3E4, and then find another temperature by looking up the curve of q = 4E4. Interpolation can be used to know that the temperature required in this embodiment is 114°C.

[0157] The use of the above chart lookup strategy is to find the temperature under known material and heat source conditions. If the temperature is known, this chart lookup strategy can also find suitable material properties and heat source conditions, just reverse the above process. For example, if the temperature at the center of the SAT / EG material at t = 1200 s is known to be 114°C, the steps can be reversed to find the characteristic time tch = 18.67, then it is possible to further evaluate which combinations of material properties meet this temperature requirement.

[0158] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited by the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement methods and are all included in the protection scope of the present invention.

Claims

1. A method for thermal condition mapping of a thermal management system for a thermochemical energy storage material chip, characterized in that, It includes the following steps: Based on the finite element model, calculate the thermal conditions of the thermal management system of the thermochemical energy storage material chip, and construct a thermal condition data set; Reduce the dimensionality of the thermal condition data set based on the product form of power functions; After the dimensionality reduction of the thermal condition data set, calculate the characteristic time of each data point, and split it into data sets of multiple mapping relationships. Identify the mapping relationships based on the backpropagation neural network; Arrange the graph lines of the mapping relationships to form a query graph.

2. The method for checking the thermal condition diagram of the thermal management system of the thermochemical energy storage material chip according to claim 1, wherein, Based on the finite element model, calculate the thermal conditions of the thermal management system of the thermochemical energy storage material chip. Specifically, use the reaction kinetic equation of the thermochemical energy storage material to construct the control equation of the finite element model. The finite element model calculates the working conditions of the thermal management system of the thermochemical energy storage material chip according to the thermal physical properties and boundary conditions of the material.

3. The method for checking the thermal condition of the thermochemical energy storage material chip thermal management system according to claim 2, wherein The control equation is expressed as: Among them, T represents the temperature quantity, and T s represents the solidus temperature, T l represents the liquidus temperature, t represents time, c represents the volume heat capacity of the heat storage material, λ represents the thermal conductivity of the heat storage material, and Q pc represents the heat absorption of phase change of the heat storage material, and Q r represents the heat absorption of the thermal reaction of the heat storage material, and ΔH pc represents the enthalpy change per unit volume of phase change, and ΔH r represents the enthalpy change per unit volume of thermal reaction. β is the percentage progress of the phase change of the material, α is the percentage progress of the thermal reaction of the material, and E a is the apparent activation energy of the thermal reaction of the material, and R is the thermodynamic constant.

4. The method for checking the thermal condition of the thermochemical energy storage material chip thermal management system according to claim 1, characterized in that, Construct a thermal condition data set, specifically including: Based on the Monte Carlo method, randomly select model parameters for finite element calculation, change the material physical properties and boundary conditions, calculate the changes in working conditions, and construct a thermal condition data set.

5. The method for checking the thermal condition of the thermochemical energy storage material chip thermal management system according to claim 1, characterized in that Reduce the dimensionality of the thermal condition data set based on the product form of power functions, specifically including: Through the finite element model, when the material physical properties, material quantity, and heat source power are known, deduce the heat storage process and temperature of the material at a certain position and time. The thermal condition data set records the influence relationship of temperature with time, position, external heat source, and material physical properties, expressed as: T(x,t,λ,c,q,L,ΔH) Among them, x is the position, t is the time, λ is the thermal conductivity of the heat storage material, c is the volume heat capacity of the heat storage material, q represents the heat flux density of the heat source chip, L represents the material volume, and ΔH represents the enthalpy change per unit volume; For the relationship T = f(X1, X2) determined by a binary function, calculate the product of the power functions of the two variables, combine the two independent variables into one, and find a unary function Describe the original binary relationship, that is, for any X1 and X2, there exists a scaling coefficient n such that: Calculate the minimum value of the loss function to obtain the scaling coefficient n. The loss function is expressed as: Among them, Abs represents taking the absolute value. Take the temperature T as the dependent variable, and m represents the number of dependent variables.

6. The method for checking the thermal condition of the thermochemical energy storage material chip thermal management system according to claim 5, characterized in that, When the loss function takes the minimum value, the value of the loss function represents the average relative error after being described by a unary function. Merge t, λ, c, q, L, ΔH successively; Reduce the parameters affecting the temperature to two, expressed as: T(x,t ch ) where t ch represents the characteristic time; Reduce the parameters affecting the heat storage process quantity of the material to two, expressed as: α(x,t ch )、β(x,t ch )、γ(x,t ch ) Among them, β is the percentage process of material phase change, α is the percentage process of material thermal reaction, and γ represents the process quantity.

7. The method for thermo-condition charting of the thermochemical energy storage material chip thermal management system according to claim 6, wherein After the thermal condition data set is dimensionally reduced, the characteristic time of each data point is calculated and split into data sets with multiple mapping relationships, specifically split into (t ch , α), (t ch , γ), (t ch , β), (α, T), (γ, T), (β, T) six groups of data sets with mapping relationships.

8. The method for checking the thermal condition of the thermochemical energy storage material chip thermal management system according to claim 7, characterized in that The process quantity γ is expressed as: Among them, t b,r represents the time point when the thermochemical reaction starts, and t e,pc represents the time point when the solid-liquid phase change ends.

9. The method for thermally operating condition charting of the thermochemical energy storage material chip thermal management system according to claim 1, wherein Arrange the graph lines of the mapping relationships to form a query graph. According to the query graph, find the temperature and energy storage progress of a certain position of the heat storage material at time t, specifically including: Calculate the characteristic time t ch ; Find the calculated characteristic time t at the abscissa ch , map the graph line to the process quantity, and find out whether the heat storage material is in the phase change stage or the endothermic reaction stage, as well as the corresponding progress percentage; Correspond the process quantity back to another section of the abscissa to find the temperature of the heat storage material.

10. The method for checking the thermal condition diagram of the thermal management system of the thermochemical energy storage material chip according to claim 1, characterized in that Arrange the graph lines of the mapping relationships to form a query graph. According to the query graph, estimate the numerical range that the material thermal physical properties and heat source power conform to, specifically including: Find the known temperature on the temperature coordinate axis, correspond it to the ordinate of the process quantity, and thus judge whether the heat storage material is in the phase change stage or the endothermic reaction stage, and the corresponding progress percentage; Corresponding to the characteristic time t in the corresponding graph line according to the process quantity and the known position ch ; After obtaining the characteristic time t ch according to the calculation formula of the characteristic time t ch calculate the numerical ranges that the thermal conductivity, heat capacity, enthalpy value, and volume heat source power of the heat storage material conform to.

Images