A thermoelectric power array reconstruction method suitable for NTD conditions in polar regions

Abstract

The application relates to the technical field of thermoelectric power generation, in particular to a thermoelectric power generation array reconstruction method suitable for NTD conditions in polar regions, which comprises the following steps: step S1: constructing a double-target function as a fitness function of a particle swarm algorithm; step S2: randomly selecting N arrangement sequences as the positions of N particles; step S3: predicting the output power and column voltage standard deviation of the thermoelectric power generation array under the N arrangement sequences; step S4: calculating the fitness of the N arrangement sequences, selecting an arrangement sequence with the highest fitness, and storing the arrangement sequence into a solution set; step S5: judging whether the iteration number reaches a maximum iteration number T; step S6: updating the selected N arrangement sequences as the positions of the N particles after mutation; and step S7: selecting an arrangement sequence with the highest fitness from the solution set as a global optimal arrangement sequence. The method solves the problem that a centralized thermoelectric power generation system cannot be efficiently and stably operated when applied to polar regions, and is suitable for polar regions.

Description

Technical Field

[0001] This invention relates to the field of thermoelectric power generation technology, specifically a method for reconfiguring thermoelectric power generation arrays suitable for NTD conditions in polar regions. Background Technology

[0002] A centralized thermoelectric power generation system is a system that utilizes the temperature difference between a high-temperature heat source and a low-temperature cold source to drive a thermoelectric power generation array (composed of multiple thermoelectric power generation modules, such as...). Figure 1 As shown, large-scale power generation systems have demonstrated broad application potential in many fields. However, when centralized thermoelectric power generation systems are applied in polar regions, the non-uniform distribution of temperature differences (originating from geothermal energy) causes the thermoelectric power generation array to operate under NTD (non-uniform temperature difference) conditions for extended periods. This results in lower output power and greater inter-array voltage imbalance, hindering efficient and stable system operation. Therefore, it is necessary to develop a thermoelectric power generation array reconfiguration method suitable for NTD conditions in polar regions to address the problem of inefficient and unstable operation of centralized thermoelectric power generation systems in polar areas. Summary of the Invention

[0003] To address the problem of centralized thermoelectric power generation systems failing to operate efficiently and stably in polar regions, this invention provides a thermoelectric power generation array reconfiguration method suitable for NTD conditions in polar regions.

[0004] This invention is achieved using the following technical solution:

[0005] A reconfiguration method for thermoelectric power generation arrays suitable for NTD conditions in polar regions, comprising the following steps:

[0006] Step S1: Construct a dual objective function with the objectives of maximizing the output power of the thermoelectric power generation array and minimizing the standard deviation of the inter-row voltage, and use the dual objective function as the fitness function of the particle swarm algorithm;

[0007] Step S2: Initialize the particle swarm algorithm, i.e. set the number of particles N and the maximum number of iterations T, and then use the particle swarm algorithm to randomly select N arrangement sequences from all arrangement sequences of the thermoelectric power generation array, and use the N arrangement sequences as the positions of N particles.

[0008] Step S3: Collect voltage and current data of the thermoelectric power generation array under N different arrangement sequences, and send the collected results into a pre-trained recurrent neural network. The recurrent neural network predicts the output power and inter-row voltage standard deviation of the thermoelectric power generation array under N different arrangement sequences based on the collected results.

[0009] Step S4: Substitute the prediction results into the fitness function of the particle swarm optimization algorithm to calculate the fitness of N arrangement orders. Then select the arrangement order with the highest fitness as the optimal arrangement order for the current iteration and store it in the solution set.

[0010] Step S5: Determine if the number of iterations has reached the maximum number of iterations T;

[0011] If the number of iterations has not reached the maximum number of iterations T, then proceed to step S6;

[0012] If the number of iterations has reached the maximum number of iterations T, then proceed to step S7;

[0013] Step S6: Perform a mutation operation on N particles, then use the particle swarm optimization algorithm to update and select N arrangement orders from all arrangement orders of the thermoelectric power generation array, and use the N arrangement orders as the positions of the N particles after mutation, and then return to step S3.

[0014] Step S7: Select the arrangement order with the highest fitness from the T optimal arrangement orders contained in the solution set as the global optimal arrangement order. Then, according to the global optimal arrangement order, use the switching matrix to rearrange several thermoelectric power generation modules of the thermoelectric power generation array under the reconstruction constraint, thereby realizing the reconstruction of the thermoelectric power generation array.

[0015] Furthermore, the dual objective function is expressed as:

[0016] ;

[0017] ;

[0018] ;

[0019] ;

[0020] In the formula: This indicates the output power of the thermoelectric generator array; This represents the standard deviation of the inter-row voltage of the thermoelectric array; This indicates the total number of rows in the thermoelectric power generation array; This indicates the total number of columns in the thermoelectric power generation array; Indicates the first Line number The voltage generated by the corresponding module is listed below; Indicates the first Line number List the internal resistance of the corresponding module; Indicates the first Column voltage; This represents the average value of the voltages in each column.

[0021] Further, the recurrent neural network comprises sequentially connected input layer, first hidden layer, second hidden layer, third hidden layer and output layer; the number of neurons of the first hidden layer is 64, and the first hidden layer adopts ReLU function as the activation function; the number of neurons of the second hidden layer is 32, and the second hidden layer adopts L2 regularization for weight decay; the number of neurons of the third hidden layer is 16.

[0022] Further, the ReLU function standardizes and integrates the input data to [0, 1] to accelerate convergence, which is expressed as:

[0023] ;

[0024] In the formula: represents the input data.

[0025] Further, the L2 regularization is expressed as:

[0026] ;

[0027] In the formula: represents the total loss of the recurrent neural network; represents the mean square error loss of the recurrent neural network; represents the normalization coefficient, which is set to 0.0001; represents the i-th connection weight parameter of the recurrent neural network; represents the total number of connection weight parameters of the recurrent neural network.

[0028] Further, in the step S6, the mutation operation process is as follows: N particles are divided into three parts: the first part of particles does not mutate; the second part of particles is uniformly mutated, that is, the position of the particles in this part is updated and regenerated in the search space; the third part of particles is non-uniformly mutated, and the mutation degree gradually decreases with the increase of the number of iterations.

[0029] Further, the mutation degree calculation formula is as follows:

[0030] ;

[0031] In the formula: represents the mutation degree; represents the initial mutation probability, which is set to 0.8; represents the termination mutation probability, which is set to 0.4; represents the current iteration number; represents the maximum iteration number.

[0032] ​Further, the step S7, the reconstruction constraint condition is: when the rearrangement operation is performed, the modules in the first row remain the original arrangement order unchanged, and the modules in the remaining rows are only exchanged with another module in the same row.

[0033] The application is aimed at the NTD conditions in polar regions, realizes the reconstruction of the thermoelectric array by constructing a double-objective function and combining the particle swarm algorithm and the recurrent neural network, thereby effectively increasing the output power of the thermoelectric array and effectively reducing the inter-column voltage imbalance degree of the thermoelectric array, so as to ensure that the centralized thermoelectric system can operate efficiently and stably when applied in polar regions.

[0034] In order to verify the beneficial effects of the application, the thermoelectric array under four common NTD conditions (column NTD condition, row NTD condition, internal NTD condition and external NTD condition) in polar regions is simulated and reconstructed by Matlab / Simulink software, and the thermoelectric array distribution diagrams before and after the reconstruction (as shown in Figure 3 、 Figure 4 、 Figure 7 、 Figure 8 、 Figure 11 、 Figure 12 、 Figure 15 、 Figure 16 indicated), voltage-power curve comparison diagrams (as shown in Figure 5 、 Figure 9 、 Figure 13 、 Figure 17 indicated), voltage-current curve comparison diagrams (as shown in Figure 6 、 Figure 10 、 Figure 14 、 Figure 18 indicated), a maximum output power comparison table (as shown in Table 1) and a minimum inter-column voltage standard deviation comparison table (as shown in Table 2) are obtained.

[0035] Table 1 is as follows:

[0036] .

[0037] Table 2 is as follows:

[0038] .

[0039] Through Figure 3 、 Figure 4 、 Figure 7 、 Figure 8 、 Figure 11 、 Figure 12 、 Figure 15 、 Figure 16It can be seen that the thermoelectric power generation array was reconfigured by rearranging several thermoelectric power generation modules (in the figure, the reference temperature difference is set to 70℃, the white modules represent thermoelectric power generation modules that have not deviated from the reference temperature difference, and the gray and black modules represent thermoelectric power generation modules that have deviated from the reference temperature difference).

[0040] pass Figure 5 , Figure 9 , Figure 13 , Figure 17 As can be seen from Tables 1 and 2, the maximum output power of the reconstructed thermoelectric generator array is significantly improved, the minimum inter-column voltage standard deviation is significantly reduced, and the voltage-power curve is a single-peak curve (the original multi-peak phenomenon is eliminated).

[0041] pass Figure 6 , Figure 10 , Figure 14 , Figure 18 It can be seen that the voltage-current curve of the reconstructed thermoelectric generator array is a linear curve (the original multiple inflection points have disappeared).

[0042] This invention effectively solves the problem of centralized thermoelectric power generation systems failing to operate efficiently and stably in polar regions, and is suitable for polar regions. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the thermoelectric power generation array.

[0044] Figure 2 This is a flowchart illustrating the present invention.

[0045] Figure 3 This is a temperature difference distribution diagram before reconfiguration of the thermoelectric power generation array under NTD conditions.

[0046] Figure 4 This is a temperature difference distribution diagram after the thermoelectric power generation array is reconstructed under NTD conditions.

[0047] Figure 5 This is a comparison chart of voltage-power curves before and after reconfiguration of the thermoelectric power generation array under NTD conditions.

[0048] Figure 6 This is a comparison chart of voltage-current curves before and after reconfiguration of the thermoelectric generator array under NTD conditions.

[0049] Figure 7 This is a temperature difference distribution diagram before reconfiguration of the thermoelectric power generation array under NTD conditions.

[0050] Figure 8 This is a temperature difference distribution diagram after the thermoelectric power generation array is reconstructed under NTD conditions.

[0051] Figure 9 This is a comparison chart of voltage-power curves before and after reconfiguration of the thermoelectric generator array under NTD conditions.

[0052] Figure 10 This is a comparison of the voltage-current curves before and after the reconstruction of the thermoelectric generator array under NTD conditions.

[0053] Figure 11 This is a temperature difference distribution diagram before the thermoelectric power generation array is reconfigured under external NTD conditions.

[0054] Figure 12 This is a temperature difference distribution diagram after the thermoelectric power generation array is reconstructed under external NTD conditions.

[0055] Figure 13 This is a comparison chart of voltage-power curves before and after reconfiguration of the thermoelectric power generation array under external NTD conditions.

[0056] Figure 14 This is a comparison chart of voltage-current curves before and after reconfiguration of the thermoelectric generator array under external NTD conditions.

[0057] Figure 15 This is a temperature distribution diagram before the thermoelectric power generation array is reconfigured under internal NTD conditions.

[0058] Figure 16 This is a temperature difference distribution diagram after the thermoelectric power generation array is reconstructed under internal NTD conditions.

[0059] Figure 17 This is a comparison chart of voltage-power curves before and after reconfiguration of the thermoelectric power generation array under internal NTD conditions.

[0060] Figure 18 This is a comparison chart of voltage-current curves before and after reconfiguration of the thermoelectric power generation array under internal NTD conditions. Detailed Implementation

[0061] A reconfiguration method for thermoelectric power generation arrays suitable for NTD conditions in polar regions, comprising the following steps:

[0062] Step S1: Construct a dual objective function with the objectives of maximizing the output power of the thermoelectric power generation array and minimizing the standard deviation of the inter-row voltage, and use the dual objective function as the fitness function of the particle swarm algorithm;

[0063] Step S2: Initialize the particle swarm algorithm, i.e. set the number of particles N and the maximum number of iterations T, and then use the particle swarm algorithm to randomly select N arrangement sequences from all arrangement sequences of the thermoelectric power generation array, and use the N arrangement sequences as the positions of N particles.

[0064] Step S3: Collecting the voltage data and current data of the thermoelectric generator array under N arrangement sequences, and sending the collection results into the pre-trained recurrent neural network, and the recurrent neural network predicts the output power and inter-column voltage standard deviation of the thermoelectric generator array under N arrangement sequences according to the collection results;

[0065] Step S4: Substituting the prediction results into the fitness function of the particle swarm algorithm, thereby calculating the fitness of N arrangement sequences, then selecting the arrangement sequence with the highest fitness as the optimal arrangement sequence of the current iteration round, and storing it into the solution set;

[0066] Step S5: Determining whether the iteration number reaches the maximum iteration number T;

[0067] If the iteration number does not reach the maximum iteration number T, step S6 is performed;

[0068] If the iteration number reaches the maximum iteration number T, step S7 is performed;

[0069] Step S6: Performing mutation operation on N particles, then updating and selecting N arrangement sequences from all arrangement sequences of the thermoelectric generator array by using the particle swarm algorithm, and corresponding to the positions of the mutated N particles, then returning to step S3;

[0070] Step S7: Selecting the arrangement sequence with the highest fitness from the T optimal arrangement sequences contained in the solution set as the global optimal arrangement sequence, then performing re-layout operation on the thermoelectric generator modules of the thermoelectric generator array by using the switch matrix under the reconstruction constraint according to the global optimal arrangement sequence, thereby realizing the reconstruction of the thermoelectric generator array.

[0071] The double-objective function is represented as:

[0072] ;

[0073] ;

[0074] ;

[0075] ;

[0076] In the formula: represents the output power of the thermoelectric generator array; represents the inter-column voltage standard deviation of the thermoelectric generator array; represents the total number of rows of the thermoelectric generator array; represents the total number of columns of the thermoelectric generator array; represents the voltage generated by the module corresponding to the i-th row and the j-th column; represents the voltage generated by the module corresponding to the i-th row and the j-th column;​ represents the first row and the column corresponding module resistance; represents the first column of column voltage; represents the average value of each column voltage.

[0077] The recurrent neural network comprises sequentially connected input layer, first hidden layer, second hidden layer, third hidden layer and output layer; the number of neurons of the first hidden layer is 64, and the first hidden layer adopts ReLU function as the activation function; the number of neurons of the second hidden layer is 32, and the second hidden layer adopts L2 regularization for weight attenuation; the number of neurons of the third hidden layer is 16.

[0078] The ReLU function standardizes and integrates the input data to [0, 1] to accelerate convergence, and is expressed as:

[0079] ;

[0080] In the formula: represents the input data.

[0081] The L2 regularization is expressed as:

[0082] ;

[0083] In the formula: represents the total loss of the recurrent neural network; represents the mean square error loss of the recurrent neural network; represents the normalization coefficient, which is set to 0.0001; represents the first connection weight parameter of the recurrent neural network; represents the total number of connection weight parameters of the recurrent neural network.

[0084] In the step S6, the mutation operation process is as follows: N particles are divided into three parts: the first part of particles does not mutate; the second part of particles is uniformly mutated, that is, the position of the particles in this part is updated and changed, and the position is regenerated in the search space; the third part of particles is non-uniformly mutated, and the mutation degree gradually decreases with the increase of the number of iterations.

[0085] The mutation degree calculation formula is as follows:

[0086] ;

[0087] In the formula: represents the mutation degree; represents the initial mutation probability, which is set to 0.8; represents the termination mutation probability, which is set to 0.4; represents the current iteration number; represents the maximum iteration number.

[0088] In the step S7, the constraint condition is reconstructed as follows: when the rearrangement operation is performed, the modules in the first row will keep the original arrangement order unchanged, and each module in the remaining rows will only exchange positions with another module in the same row.

[0089] Although the specific embodiments of the present application are described above, those skilled in the art should understand that these are only illustrative, and the protection scope of the present application is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of the present application, and these changes and modifications all fall within the protection scope of the present application.

Claims

1. A thermoelectric array reconfiguration method suitable for polar NTD conditions, characterized by: The method comprises the following steps: Step S1: A double-objective function is constructed with the maximum output power and the minimum standard deviation of inter-column voltage of the thermoelectric generation array as the target, and the double-objective function is taken as the fitness function of the particle swarm algorithm; Step S2: The particle swarm algorithm is initialized, that is, the number of particles N and the maximum number of iterations T are set, then N arrangement sequences are randomly selected from all arrangement sequences of the thermoelectric generation array by using the particle swarm algorithm, and the N arrangement sequences are taken as the positions of the N particles in a one-to-one correspondence; Step S3: The voltage data and current data of the thermoelectric generation array under the N arrangement sequences are collected, and the collection results are sent to the pre-trained recurrent neural network, and the recurrent neural network predicts the output power and the standard deviation of inter-column voltage of the thermoelectric generation array under the N arrangement sequences according to the collection results; Step S4: The prediction results are substituted into the fitness function of the particle swarm algorithm, and the fitness of the N arrangement sequences is calculated, then the arrangement sequence with the highest fitness is selected as the optimal arrangement sequence of the current iteration round, and the arrangement sequence is stored in the solution set; Step S5: It is judged whether the number of iterations reaches the maximum number of iterations T; If the number of iterations does not reach the maximum number of iterations T, step S6 is performed; If the number of iterations has reached the maximum number of iterations T, step S7 is performed; Step S6: Mutation operation is performed on the N particles, then N arrangement sequences are updated and selected from all arrangement sequences of the thermoelectric generation array by using the particle swarm algorithm, and the N arrangement sequences are taken as the positions of the mutated N particles in a one-to-one correspondence, and then the step S3 is returned; Step S7: The arrangement sequence with the highest fitness is selected from the T optimal arrangement sequences contained in the solution set as the global optimal arrangement sequence, then the global optimal arrangement sequence is used to perform the rearrangement operation on the thermoelectric generation modules of the thermoelectric generation array under the reconstruction constraint condition by using the switch matrix, so that the reconstruction of the thermoelectric generation array is realized; The double-objective function is represented as: ； ； ； ； In the formulae: represents the output power of the thermoelectric array; represents the inter-column voltage standard deviation of the thermoelectric array; represents the total number of rows of the thermoelectric array; represents the total number of columns of the thermoelectric array; represents the voltage generated by the module corresponding to the row and the column; represents the internal resistance of the module corresponding to the row and the column; represents the column voltage of the column; represents the average value of the column voltages.

2. The thermoelectric array reconstruction method for NTD conditions in polar regions according to claim 1, characterized in that: The recurrent neural network comprises sequentially connected input layer, first hidden layer, second hidden layer, third hidden layer and output layer; the number of neurons of the first hidden layer is 64, and the first hidden layer adopts ReLU function as the activation function; the number of neurons of the second hidden layer is 32, and the second hidden layer adopts L2 regularization for weight decay; the number of neurons of the third hidden layer is 16.

3. The thermoelectric array reconstruction method for NTD conditions in polar regions according to claim 2, characterized in that: The ReLU function standardizes and integrates the input data to [0, 1] to accelerate convergence, and is represented as: ； In the formulae: represents input data.

4. The thermoelectric array reconstruction method for NTD conditions in polar regions according to claim 2, characterized in that: The L2 regularization is represented as: ； In the formula: This represents the total loss of the recurrent neural network; This represents the mean squared error loss of a recurrent neural network. This represents the normalization coefficient, set to 0.0001; Represents the first recurrent neural network. Each connection weight parameter; This represents the total number of connection weight parameters in a recurrent neural network.

5. The thermoelectric array reconstruction method for NTD conditions in polar regions according to claim 1, characterized in that: In the step S6, the mutation operation process is as follows: the N particles are divided into three parts: the first part of particles does not mutate; the second part of particles is subjected to uniform mutation, that is, the positions of the particles in the second part are updated and changed, and the positions are generated again in the search space; the third part of particles is subjected to non-uniform mutation, and the mutation degree gradually decreases with the increase of the number of iterations.

6. The thermoelectric array reconstruction method for NTD conditions in polar regions according to claim 5, characterized in that: The mutation degree calculation formula is as follows: ； In the formula: represents the mutation degree; represents the initial mutation probability, and is set to 0.8; represents the termination mutation probability, and is set to 0.4; represents the current iteration number; represents the maximum iteration number.

7. The thermoelectric array reconstruction method for NTD conditions in polar regions according to claim 1, characterized in that: In the step S7, the reconstruction constraint condition is that when the rearrangement operation is performed, the modules in the first row will keep the original arrangement sequence unchanged, and the modules in the remaining rows will only exchange positions with another module in the same row.

Images