Organic Rankine cycle performance prediction method established based on quantitative structure-activity
By using the molecular structure of the working fluid based on quantitative structure effect, an artificial neural network model is constructed using the molecular structure of the working fluid to predict the thermal physical properties and cyclic properties of the organic Rankine cyclic working fluid, solving the problem of low working fluid screening efficiency in the existing technology, and achieving rapid and accurate working fluid screening.
Patent Information
- Application Number
- CN202411563706.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-05
- Publication Date
- 2025-02-25
AI Technical Summary
The existing organic Rankine cycle working fluid screening method relies on the acquisition of thermal physical properties parameters of the working fluid. The data sheet is limited and the experimental cost is high, which affects the accuracy and efficiency of the screening.
Using a quantitative structure-effect method, an artificial neural network model is constructed through the molecular structure of the working fluid, the thermal physical properties and parameters of the working fluid are predicted, and the organic Rankine cycle performance is calculated based on these parameters.
In the case where the thermal physical properties of the working fluid are unknown, the organic Rankine cycle performance is quickly and accurately predicted, which improves the accuracy and efficiency of working fluid screening.
Smart Images
Figure CN119517217A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to an organic Rankine cycle performance prediction method based on quantitative structure-activity relationship, and belongs to the technical field of energy conservation and environmental protection. Background Art
[0002] Organic Rankine cycle is an important technical means for waste heat utilization, and screening suitable circulating working fluids is the key to improving the performance of organic Rankine cycle. At present, the screening method of organic Rankine cycle working fluids is mainly achieved through enumeration. The specific method is to first determine the thermophysical property parameters of the working fluid for the candidate working fluid, and combine it with the temperature conditions of the high-temperature heat source and the low-temperature heat source of the cycle operation, predict its cycle performance through theoretical calculation, and then determine its actual cycle performance through experiments, and finally screen out the best cycle working fluid by comparing it with the cycle performance of other candidate working fluids. Therefore, obtaining the thermophysical property parameters of the working fluid is a prerequisite for predicting its thermophysical properties. The thermophysical property parameters of the working fluid are mainly obtained by checking the data manual or experiment. However, the types of substances in the data manual are limited, and it is impossible to exhaust all substances. The time cost and economic cost of the experiment are very high, which restricts the working fluid screening work to a certain extent and affects the accuracy and efficiency of the working fluid screening. Summary of the invention
[0003] In order to overcome the above-mentioned defects of the prior art, the present invention provides an organic Rankine cycle performance prediction method based on quantitative structure-activity relationship. When the thermophysical property parameters of the working fluid are unknown, the organic Rankine cycle performance can be quickly and accurately predicted based on the molecular structure of the working fluid, thereby improving the accuracy and efficiency of working fluid screening.
[0004] The technical solution of the present invention to achieve the above-mentioned purpose is: an organic Rankine cycle performance prediction method established based on quantitative structure-activity relationship, an artificial neural network model of working fluid thermophysical property parameters is used to predict the thermophysical property parameters of alternative organic working fluids (or organic working fluids to be predicted), and based on the predicted thermophysical property parameters of the alternative organic working fluids, the organic Rankine cycle performance of the alternative organic working fluids is obtained based on thermodynamic calculations (the prediction of organic Rankine cycle performance is achieved based on thermodynamic calculations), the input of the artificial neural network model of working fluid thermophysical property parameters is the molecular descriptor of the organic working fluid, and the output is the thermophysical property parameters of the corresponding organic working fluid.
[0005] The working fluid is screened based on the predicted organic Rankine cycle performance of the alternative organic working fluid, providing a basis for subsequent experiments and selection.
[0006] Preferably, the artificial neural network model of the working fluid thermophysical property parameters includes an organic working fluid critical temperature artificial neural network model, an organic working fluid critical pressure artificial neural network model, an organic working fluid eccentricity factor artificial neural network model and an organic working fluid ideal gas constant-pressure specific heat capacity artificial neural network model.
[0007] Preferably, the method for constructing the artificial neural network model of the working fluid thermophysical property parameters comprises the following steps:
[0008] S1: Based on the critical temperature of the organic working fluid, multiple organic working fluids are selected from the physical property database (e.g., NIST database) to construct a data set of thermophysical property parameters of the organic Rankine cycle working fluids;
[0009] S2: drawing the molecular structure of each organic working substance in the data set;
[0010] S3: Calculate the molecular descriptor of the corresponding organic working fluid according to the molecular structure of each organic working fluid;
[0011] S4: Taking the molecular descriptors of each organic working fluid as input and the thermophysical property parameters of the corresponding organic working fluid as output, an artificial neural network model of the thermophysical property parameters of the working fluid is constructed and trained.
[0012] Preferably, the molecular structure of the organic working fluid in S2 is a molecular structure in which the organic working fluid is in a stable minimum energy state.
[0013] Preferably, in S2, the molecular structure of each organic working fluid is drawn using ChemDraw 20.0 software.
[0014] Furthermore, in S2, Gaussian software is used to optimize the molecular structures of the drawn organic working fluids, so that the molecular structures of the organic working fluids are all molecular structures in which the organic working fluids are in a stable minimum energy state.
[0015] Preferably, in S3, the molecular descriptors of the calculated organic working fluid are preprocessed / screened, any one of the two molecular descriptors with a correlation greater than 90% is removed, and the molecular descriptors in which more than 80% of the data in the molecular descriptor calculation are constants are removed to obtain molecular descriptors for constructing an artificial neural network model.
[0016] Preferably, in S3, AlvaDesc 2.0.6 software is used to calculate the molecular descriptor of the organic working fluid.
[0017] Preferably, the thermophysical property parameters of the organic working fluid in S4 include the critical temperature, critical pressure, eccentricity factor and specific heat capacity of an ideal gas at constant pressure of the organic working fluid.
[0018] Preferably, in S4, a BP neural network is used to construct a neural network regression model of the thermophysical property parameters of the working fluid.
[0019] Preferably, in S4, a genetic algorithm is used to screen (or simplify) the organic working fluid descriptors obtained in S3, and some of the most frequently used molecular descriptors are selected to construct an artificial neural network model of working fluid thermophysical property parameters.
[0020] Preferably, in S4, the data set is randomly divided (according to a preset ratio) into a training set, a validation set and a test set to train a neural network regression model of the thermophysical property parameters of the working fluid.
[0021] The beneficial effect of the present invention is that the present invention constructs an artificial neural network model of the quantitative relationship between the molecular structure and thermophysical property parameters of the working fluid based on the quantitative structure-activity modeling method. By adopting the artificial neural network model, when the thermophysical property parameters of the working fluid are unknown, the performance of the organic Rankine cycle can be accurately predicted based on the molecular structure of the working fluid (through analysis and calculation of the molecular structure of the working fluid), thereby improving the accuracy and efficiency of working fluid screening. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 It is a logic block diagram of the technical implementation of the present invention;
[0023] Figure 2 It is a schematic diagram of the structure of the BP neural network used in constructing the neural network regression model of the thermophysical property parameters of the working fluid of the present invention;
[0024] Figure 3 It is a usage frequency diagram of molecular descriptors (molecular descriptors with higher usage frequency) in the process of using genetic algorithm to screen molecular descriptors when constructing an artificial neural network model of critical temperature of an organic working fluid according to the present invention;
[0025] Figure 4 This is a prediction result diagram of the critical temperature parameters of 85 kinds of organic working fluids using the organic working fluid critical temperature artificial neural network regression model of the present invention;
[0026] Figure 5 It is a usage frequency diagram of molecular descriptors (molecular descriptors with higher usage frequency) in the process of using genetic algorithm to screen molecular descriptors when constructing an artificial neural network model of critical pressure of organic working fluids in the present invention;
[0027] Figure 6 This is a prediction result diagram of the critical pressure parameters of 85 kinds of organic working fluids predicted by the artificial neural network regression model of the critical pressure of organic working fluids of the present invention;
[0028] Figure 7It is a usage frequency diagram of molecular descriptors (molecular descriptors with higher usage frequency) in the process of using genetic algorithm to screen molecular descriptors when constructing an artificial neural network model of eccentricity factor of organic working fluid in the present invention;
[0029] Figure 8 This is a prediction result diagram of the eccentricity factor parameters of 85 kinds of organic working fluids using the organic working fluid eccentricity factor artificial neural network regression model of the present invention;
[0030] Fig. 9 It is a usage frequency diagram of molecular descriptors (molecular descriptors with higher usage frequency) in the process of using genetic algorithm to screen molecular descriptors when constructing an artificial neural network model of specific heat capacity of ideal gas at constant pressure of an organic working fluid according to the present invention;
[0031] Fig.10 This is a prediction result diagram of the ideal gas constant pressure specific heat capacity parameters of 85 kinds of organic working fluids using the artificial neural network regression model of the ideal gas constant pressure specific heat capacity of organic working fluids of the present invention;
[0032] Fig.11 is a schematic structural diagram of a simple organic Rankine cycle system of the present invention;
[0033] Fig.12 is a temperature entropy diagram of a simple organic Rankine cycle of the present invention;
[0034] Fig.13 This is a simple organic Rankine cycle thermal efficiency diagram calculated using a coupling model based on a BP neural network according to the present invention. DETAILED DESCRIPTION
[0035] The present invention discloses a method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship. The quantitative structure-activity modeling method is applied to the prediction of the thermophysical property parameters of the working fluid and the performance of the organic Rankine cycle. The final performance effect of the organic Rankine cycle operation can be obtained quickly and directly by analyzing and calculating the molecular structure of the working fluid. The logic block diagram of the technical route is as follows: Figure 1 As shown, it includes the following five steps:
[0036] Step 1: Construct a data set of thermophysical properties parameters of the organic Rankine cycle working fluid.
[0037] Alkanes, alkenes, aromatic hydrocarbons and their halides are important sources of organic Rankine cycle working fluids. Aiming at the temperature range of medium and low temperature heat sources (the high temperature heat source temperature is set to 250°C and the low temperature heat source temperature is set to 15°C), 85 relatively suitable organic working fluids are selected from the NIST database of organic working fluids based on the boiling point (the boiling point is generally higher than the low temperature heat source temperature) and critical temperature (the critical temperature is generally lower than the high temperature heat source temperature) of the working fluid to construct the thermophysical property parameter data set of the organic Rankine cycle working fluids. The names, CAS numbers and thermophysical property parameters of these 85 working fluids are shown in Table 1.
[0038] Table 1 Names, CAS numbers and thermophysical properties of 85 organic working fluids screened
[0039]
[0040]
[0041]
[0042]
[0043] Step 2: Draw the molecular structure of each organic working fluid in the data set.
[0044] After constructing the data set, the molecular structures of 85 organic working fluids were drawn using ChemDraw 20.0 software, and the molecular structures of 85 organic working fluids were optimized using Gaussian software to ensure that the molecular structures of the working fluids were not in the transition state of the organic working fluids but in a stable minimum energy state.
[0045] Step 3: Calculate and screen the molecular descriptors of the corresponding organic working fluids based on their molecular structures.
[0046] The optimized molecular structure was imported into AlvaDesc 2.0.6 software to calculate the molecular descriptors of these 85 organic working fluids. AlvaDesc 2.0.6 software can realize the digitization of the molecular structure of substances by calculating 33 molecular descriptors (for example, molecular connectivity index, ring descriptor, topological descriptor, information index and 3D matrix descriptor, etc.). The molecular structure of the organic working fluid was calculated by AlvaDesc 2.0.6 software, and 5666 molecular descriptors were obtained. Subsequently, these 5666 molecular descriptions were preprocessed, and any one molecular descriptor was removed from the two molecular descriptors with a correlation greater than 90%, and the molecular descriptors whose data in the descriptor calculation value was more than 80% were removed as constants. Finally, 538 molecular descriptors were obtained to construct an artificial neural network model of the thermophysical property parameters of the working fluid.
[0047] Step 4: Using the molecular descriptors of each organic working fluid finally obtained in step 3 as input and the thermophysical property parameters of the corresponding organic working fluid as output, an artificial neural network model of the thermophysical property parameters of the working fluid is constructed and trained.
[0048] The thermophysical property parameters of the organic working fluid include the critical temperature, critical pressure, eccentricity factor and specific heat capacity of ideal gas at constant pressure of the organic working fluid. Accordingly, the artificial neural network models of the thermophysical property parameters of the working fluid constructed include an artificial neural network model of the critical temperature of the organic working fluid, an artificial neural network model of the critical pressure of the organic working fluid, an artificial neural network model of the eccentricity factor of the organic working fluid and an artificial neural network model of the specific heat capacity of ideal gas at constant pressure of the organic working fluid.
[0049] In this step, a BP neural network is used to construct a neural network regression model of the working fluid thermophysical property parameters. The structural diagram of the BP neural network is shown in Figure 2 shown.
[0050] (1) Establishment of artificial neural network model of critical temperature of organic working fluid
[0051] The 538 molecular descriptors obtained in step three are used as initial variables, and the critical temperature of the organic working fluid is used as the target variable to construct an artificial neural network regression model of the critical temperature of the organic working fluid. In order to simplify the construction of the model, a genetic algorithm is used to streamline (or screen) the 538 molecular descriptors. The genetic algorithm parameters are set as follows: the average absolute error of the artificial neural network regression model for predicting the critical temperature of the working fluid is used as the fitness function, the population size is set to 2000, the termination generation is set to 500, and the mutation rate is set to 0.2. After the molecular descriptors are screened by the genetic algorithm, the usage frequency of each molecular descriptor is counted, and the usage frequency distribution is as follows: Figure 3As shown in the figure, it can be seen that there are 43 molecular descriptors with a usage frequency greater than 2%. The molecular descriptors are sorted from high to low according to the usage frequency. The top 9 molecular descriptors are MeanDD, RTp, GATS1e, P_VSA_ppp_L, P_VSA_i_3, MaxDD, P_VSA_LogP_7, P_VSA_v_3, B02[CF], and the usage frequencies are 39.58%, 31.25%, 29.17%, 27.08%, 25%, 22.92%, 10.42%, 10.42%, and 8.33%, respectively. With the above 9 molecular descriptors as input variables and the critical temperature of the working fluid as the output variable, a double hidden layer BP neural network QSPR (quantitative structure-activity relationship) model is constructed. The data set of the critical temperature of 85 organic working fluids is randomly divided into a training set, a validation set, and a test set in a ratio of 0.7:0.15:0.15 to train the BP neural network regression model. After training, when the learning rate of the regression model is set to 0.5, the number of neurons in the hidden layer is set to 35, and the training function is set to the trainlm function, the BP neural network model can achieve the best prediction performance. Figure 4 This is the prediction result diagram of the critical temperature of 85 working fluids obtained by using this BP neural network regression model. It can be seen from the figure that the prediction error range of the critical temperature of the working fluid is -6.02%-6.76%, and the average relative error of the prediction is about 0.73%.
[0052] (2) Establishment of artificial neural network model of critical pressure of organic working fluid
[0053] The 538 molecular descriptors obtained in step 3 were used as initial variables, and the critical pressure of the working fluid was used as the target variable to construct an artificial neural network regression model for the critical pressure of organic working fluids. In order to simplify the construction of the model, a genetic algorithm was used to streamline the 538 molecular descriptors. The genetic algorithm parameters were set as follows: the average absolute error of the artificial neural network regression model for predicting the critical pressure of the working fluid was used as the fitness function, the population size was set to 2000, the termination generation was set to 500, and the mutation rate was set to 0.2. After the molecular descriptors were screened by the genetic algorithm, the usage frequency of each molecular descriptor was counted, and the usage frequency distribution is shown in the figure below. Figure 5As shown in the figure, it can be seen that there are 41 molecular descriptors with a usage frequency greater than 2%. The molecular descriptors are sorted from high to low according to the usage frequency. The top 9 molecular descriptors are JGT, Mor31v, CIC0, GATS1e, ATS3e, GATS1m, IDM, H2e, and PHI, and the usage frequencies are 68%, 42%, 28%, 16%, 12%, 12%, 10%, 10%, and 8%, respectively. The above 9 molecular descriptors are used as input variables and the critical pressure of the working fluid is used as the output variable to construct a double hidden layer BP neural network model. The data set of the critical pressure of 85 organic working fluids is randomly divided into a training set, a validation set, and a test set in a ratio of 0.7:0.15:0.15 to train the BP neural network regression model. After training, when the learning rate of the regression model is set to 0.4, the number of hidden layer neurons is set to 35, and the training function is set to the trainlm function, the BP neural network model can obtain the best prediction performance. Figure 6 This is the prediction result diagram of the critical pressure of 85 working fluids obtained by using this BP neural network regression model. It can be seen from the figure that the prediction error range of the critical pressure of the working fluid is -7.32%-6.6%, and the average relative error of the prediction is about 0.89%.
[0054] (3) Establishment of artificial neural network model of eccentricity factor of organic working fluid
[0055] The 538 molecular descriptors obtained in step three were used as initial variables, and the working fluid eccentricity factor was used as the target variable to construct an artificial neural network regression model of the organic working fluid eccentricity factor. In order to simplify the construction of the model, a genetic algorithm was used to streamline the 538 molecular descriptors. The genetic algorithm parameters were set as follows: the average absolute error of the artificial neural network regression model for the working fluid eccentricity factor was used as the fitness function, the population size was set to 2000, the termination generation was set to 500, and the mutation rate was set to 0.2. After the molecular descriptors were screened by the genetic algorithm, the usage frequency of each molecular descriptor was counted. The usage frequency is as follows: Figure 7As shown in the figure, it can be seen that there are 40 molecular descriptors with a usage frequency greater than 2%. The molecular descriptors are sorted from high to low according to the usage frequency. The top 8 molecular descriptors are IDE, ATSC3e, ATSC6v, SpMax2_Bh(m), SpMax2_Bh(p), MaxDD, MATS1p and Dm, and the usage frequencies are 66%, 54%, 22%, 14%, 10%, 8%, 8%, 8%. The double hidden layer BP neural network model is constructed with the above 8 molecular descriptors as input variables and the working fluid eccentricity factor as the output variable. The data set of 85 organic working fluid eccentricity factors is randomly divided into training set, validation set and test set in the ratio of 0.7:0.15:0.15 to train the BP neural network regression model. After training, when the learning rate of the regression model is set to 0.4, the number of hidden layer neurons is set to 50, and the training function is set to the trainlm function, the BP neural network model can obtain the best prediction performance. Figure 8 This is the prediction result diagram of 85 kinds of working fluid eccentricity factors obtained by using this BP neural network regression model. It can be seen from the figure that the prediction error range of the working fluid eccentricity factor is -14.87%-8.42%, and the average relative error of the prediction is about 1.68%.
[0056] (4) Establishment of an artificial neural network model for the specific heat capacity of an ideal gas at constant pressure for organic working fluids
[0057] The 538 molecular descriptors obtained in step three were used as initial variables, and the ideal gas specific heat capacity at constant pressure of the working fluid was used as the target variable to construct a neural network regression model for the ideal gas specific heat capacity at constant pressure of the organic working fluid. In order to simplify the construction of the model, a genetic algorithm was used to streamline the 538 molecular descriptors, and the genetic algorithm parameters were set as follows: the average absolute error of the artificial neural network regression model predicting the ideal gas specific heat capacity at constant pressure of the working fluid was used as the fitness function, the population size was set to 2000, the termination generation was set to 500, and the mutation rate was set to 0.2. After the molecular descriptors were screened by the genetic algorithm, the usage frequency of each molecular descriptor was counted, and the usage frequency was as follows: Fig. 9As shown in the figure, it can be seen that there are 41 molecular descriptors with a usage frequency greater than 2%. The molecular descriptors are sorted from high to low according to the usage frequency. The top 9 molecular descriptors are HTe, RBN, TDB03p, GGI1, Av, VAR, P_VSA_ppp_L, P_VSA_v_3, Mor04v, and the usage frequencies are 75.51%, 38.78%, 34.69%, 18.37%, 12.25%, 8.16%, 8.16%, 6.12%, and 6.12%, respectively. The above 9 molecular descriptors are used as input variables and the ideal gas constant pressure specific heat capacity of the working fluid is used as the output variable to construct a double hidden layer BP neural network model. The data set of the ideal gas constant pressure specific heat capacity of 85 organic working fluids is randomly divided into a training set, a validation set, and a test set according to the ratio of 0.7:0.15:0.15 to train the BP neural network regression model. After training, when the learning rate of the regression model is set to 0.9, the number of neurons in the hidden layer is set to 35, and the training function is set to the trainlm function, the BP neural network model can achieve the best prediction performance. Fig.10 This is a graph showing the prediction results of the constant-pressure specific heat capacity of ideal gas for 85 working fluids obtained using the BP neural network regression model. It can be seen from the figure that the prediction error range of the constant-pressure specific heat capacity of ideal gas for working fluids is -6.57%-5.90%, and the average relative error of the prediction is about 1.20%.
[0058] Step 5: The artificial neural network model of the thermophysical property parameters of the working fluid is used to predict the thermophysical property parameters of the candidate organic working fluid (or the organic working fluid to be predicted), and the prediction of the performance of the organic Rankine cycle is realized based on thermodynamic calculations.
[0059] The final performance of the organic Rankine cycle is an important criterion for selecting working fluids, and the thermal efficiency of the organic Rankine cycle is the most commonly used cycle performance evaluation indicator. Taking a simple organic Rankine cycle as an example, the coupling model established based on the BP neural network is verified from the organic Rankine cycle system level. Fig.11 is the structural diagram of a simple organic Rankine cycle, and the corresponding temperature entropy diagram is as follows Fig.12 As shown. The whole system of the simple organic Rankine cycle is mainly composed of four parts: evaporator 1, expander 2, condenser 3 and working fluid pump 4. The thermodynamic process of the simple organic Rankine cycle is as follows: the organic working fluid is compressed into a high-pressure liquid in the working fluid pump. Subsequently, the high-pressure working fluid enters the evaporator to absorb heat and turns into high-temperature and high-pressure steam. The high-temperature and high-pressure steam enters the expander to expand and do work. The low-pressure steam after doing work is condensed into a low-pressure liquid working fluid through the condenser. Then the working fluid enters the working fluid pump again to start a new thermodynamic cycle. The following assumptions are made for the operating conditions of the simple organic Rankine cycle:
[0060] (1) The simple organic Rankine cycle system is in a stable operating state;
[0061] (2) Ignore the pressure loss and heat loss in the system pipelines and components;
[0062] (3) The isentropic efficiencies of the working fluid pump and expander in the simple organic Rankine cycle system are set to 0.7 and 0.65, respectively;
[0063] (4) The evaporation temperature of the working fluid in the cycle is set to 0.9 times the critical temperature of the working fluid, the superheat of the working fluid is set to 5K, and the condensation temperature of the working fluid is set to 0.9 times the boiling point of the working fluid.
[0064] Based on the first law of thermodynamics, a thermodynamic model of a simple organic Rankine cycle is established. For the thermodynamic model of a simple organic Rankine cycle, the coupling model established based on the BP neural network is introduced, and the calculated thermal efficiency of the simple organic Rankine cycle is compared with the results calculated by the Refprop software under the same conditions. The deviation of the thermal efficiency of the simple organic Rankine cycle calculated by the coupling model established by the BP neural network and the Refprop software is compared. Fig.13 As shown. Analysis of the deviation between the coupling model based on the BP neural network and the calculation results of Refprop shows that the deviation range of the coupling model based on the BP neural network in calculating the thermal efficiency of the simple organic Rankine cycle is -5.49%-9.2%, and the average absolute deviation is 2.11%. It can be seen that the artificial neural network model of the thermophysical property parameters of the working fluid (the molecular structure-thermophysical properties-cycle performance evaluation coupling model of the working fluid) is effective in the organic Rankine cycle, and can realize a reliable evaluation of the thermodynamic performance of the organic Rankine cycle according to the molecular structure of the working fluid when the thermophysical properties of the working fluid are unknown.
[0065] Unless otherwise specified or one preferred or optional technical means is a further limitation of another technical means, all preferred and optional technical means disclosed in the present invention can be arbitrarily combined to form several different technical solutions.
Claims
1. A method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship, characterized in that An artificial neural network model of the thermophysical property parameters of the working fluid is used to predict the thermophysical property parameters of the alternative organic working fluid. Based on the predicted thermophysical property parameters of the alternative organic working fluid, the organic Rankine cycle performance of the alternative organic working fluid is obtained based on thermodynamic calculation. The input of the artificial neural network model of the thermophysical property parameters of the working fluid is the molecular descriptor of the organic working fluid, and the output is the thermophysical property parameters of the corresponding organic working fluid.
2. The method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship according to claim 1, characterized in that The artificial neural network models of the working fluid thermophysical property parameters include an organic working fluid critical temperature artificial neural network model, an organic working fluid critical pressure artificial neural network model, an organic working fluid eccentricity factor artificial neural network model and an organic working fluid ideal gas constant pressure specific heat capacity artificial neural network model.
3. The method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship according to claim 1, characterized in that The method for constructing the artificial neural network model of the working fluid thermophysical property parameters comprises the following steps: S1: Based on the critical temperature of organic working fluids, multiple organic working fluids are selected from the physical property database to construct a data set of thermophysical property parameters of organic Rankine cycle working fluids; S2: drawing the molecular structure of each organic working substance in the data set; S3: Calculating the molecular descriptor of the corresponding organic working fluid according to the molecular structure of each organic working fluid; S4: Taking the molecular descriptors of each organic working fluid as input and the thermophysical property parameters of the corresponding organic working fluid as output, an artificial neural network model of the thermophysical property parameters of the working fluid is constructed and trained.
4. The method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship according to claim 3, characterized in that The molecular structure of the organic working fluid in S2 is a molecular structure in which the organic working fluid is in a stable minimum energy state.
5. The method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship according to claim 4, characterized in that In S2, the molecular structure of each organic working fluid is drawn using ChemDraw 20.0 software.
6. The method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship according to claim 5, characterized in that In S2, Gaussian software is used to optimize the molecular structures of the drawn organic working fluids, so that the molecular structures of the organic working fluids are all molecular structures in which the organic working fluids are in a stable minimum energy state.
7. The method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship according to claim 3, characterized in that In S3, the molecular descriptors of the calculated organic working fluid are preprocessed to remove any one of the two molecular descriptors with a correlation greater than 90%, and to remove molecular descriptors in which more than 80% of the data in the molecular descriptor calculation are constants, so as to obtain molecular descriptors for constructing an artificial neural network model of the thermophysical property parameters of the working fluid.
8. The method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship according to claim 3, characterized in that In S3, the AlvaDesc 2.0.6 software is used to calculate the molecular descriptors of the organic working fluid.
9. The method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship according to claim 3, characterized in that In S4, a BP neural network is used to construct a neural network regression model of the thermophysical property parameters of the working fluid.
10. The method for predicting the performance of an organic Rankine cycle based on quantitative structure-activity relationship according to claim 9, characterized in that In S4, a genetic algorithm is used to screen the organic working fluid descriptors obtained in S3, and some of the molecular descriptors with the highest usage frequency are selected to construct an artificial neural network model of the working fluid thermophysical property parameters.