Prediction method of S-characteristic curve of pump-turbine based on artificial neural network

Abstract

The present invention discloses a method for predicting the S characteristic curve of a pump-turbine based on an artificial neural network, which belongs to the technical field of hydraulic characteristic prediction of pump-turbines. A large number of first quadrant curves of pump-turbines and required runner characteristic parameters are obtained through existing literature. After sample conversion, sample reduction, building an artificial neural network, training and verification, the response relationship between the characteristic parameters of the pump-turbine and the first quadrant characteristic curve is obtained, thereby realizing the rapid acquisition of the S characteristic curve in the design stage of the pump-turbine. Compared with the prior art, the present invention does not need to make a model machine and perform measurements in the design stage of the pump-turbine, which solves the technical problem that the existing pump-turbine needs to make a model machine and perform measurements to obtain the S characteristic curve of the pump-turbine in the design stage, resulting in high production costs.

Description

Technical Field

[0001] The invention belongs to the technical field of hydraulic characteristic prediction of a water pump turbine, and in particular relates to a water pump turbine S characteristic curve prediction method based on an artificial neural network. Background Art

[0002] As the proportion of pumped-storage units in my country's power system gradually increases and the capacity of single units gradually increases, the reliability of unit operation is becoming more and more important for the safe operation of power stations and even power grid systems. As the core component of pumped-storage units, the operating performance of pump-turbines significantly affects the safe and stable operation of pump-storage units. In the transient process conditions such as turbine startup, load rejection, and pump power failure, the full characteristic curve of the pump-turbine has an obvious "S" shape phenomenon. The same opening corresponds to three different unit flows and torques for a unit speed, and there is a reverse flow. The instantaneous reversal of flow is very unfavorable to the force on the guide vanes, and it will also cause problems such as unit power oscillation, grid connection difficulties, and increased vibration intensity. The unstable operation of the unit caused by the S characteristics of the pump-turbine has become the most important problem faced by the unit design and operation stage. However, in the unit design stage, obtaining the S characteristic curve of the pump-turbine requires making a model machine and measuring it, so this work is time-consuming and expensive. Therefore, developing a method for accurately predicting the S characteristic curve of the pump-turbine can reduce the design cost of the pump-turbine and improve the design efficiency. Summary of the invention

[0003] The purpose of the present invention is to provide a method for predicting the S characteristic curve of a pump-turbine based on an artificial neural network, which solves the technical problem that in the existing pump-turbine design stage, a model machine needs to be made and measured to obtain the S characteristic curve of the pump-turbine, resulting in high production costs.

[0004] The technical solution adopted by the present invention is a method for predicting the S characteristic curve of a pump turbine based on an artificial neural network, which specifically comprises the following steps:

[0005] S1: Determine the characteristic parameters of the pump-turbine related to the S characteristic, and reduce the order of the characteristic parameters to obtain the sample set required for artificial neural network prediction;

[0006] S2: Obtain the first quadrant characteristic curve and corresponding characteristic parameters of the pump-turbine in the sample set;

[0007] S3: Through improvement Suter Transformation, processing the first quadrant characteristic curve of the pump turbine in S2, and deriving Suter Inverse transform to obtain n11-Q11 data points;

[0008] S4: Construct an artificial neural network structure, and combine the characteristic parameters obtained in S2 with those obtained after processing in S3.n 11 - Q 11 The data points are taken as a whole sample set, and then the sample set is divided into a training set and a test set. The training set is used to train the artificial neural network, and the test set is used to verify the generalization ability of the constructed artificial neural network model;

[0009] S5: By mean square error and coefficient of determination R 2 Verify the rationality and accuracy of the artificial neural network structure;

[0010] S6: Use the trained artificial neural network to predict the first quadrant curve of the pump turbine containing the characteristic parameters required for prediction. After the prediction is completed, the prediction results are analyzed. Suter Inverse transformation to obtain the first quadrant of the pump turbine n 11 - Q 11 Characteristic curve.

[0011] The present invention is also characterized in that:

[0012] The characteristic parameters in S1 include: S characteristic and pump turbine operating speed ratio n sT , runner high pressure side diameter D 1 , Runner low pressure side diameter D 2 , Blade high pressure side height b 1 And the blade placement angle β 1b , guide vane opening α , the flow rate at the optimal point Q N and the optimal speed n N , the guide vane opening at the optimal point α N , Unit flow on the characteristic curve Q 11 and unit speed n 11 .

[0013] The specific order reduction processing of characteristic parameters in S1 is as follows:

[0014] The characteristic curve parameters are defined as shown in formula (1):

[0015] (1)

[0016] In formula (1):n 11 Indicates unit speed, unit rpm ; n Indicates the speed, unit rpm ; Q 11 Indicates unit flow, unit m 3 / s ; Q Indicates flow rate, unit m 3 / s ; M 11 Represents unit moment, unit N × m ; M Indicates torque, unit N × m ; D 2 Indicates the diameter of the runner low pressure side, unit m ; H Indicates rated head, unit m ;

[0017] When the optimal point is unknown, the efficiency calculation formula of the rotating machinery shown in formula (2) shows that:

[0018] (2)

[0019] In formula (2): or represents the pump-turbine efficiency; r Indicates density, unit kg / m 3 ; g Indicates the acceleration due to gravity, in m / s 2 ; Q Indicates flow rate, unit m 3 / s ; H Indicates rated head, unit m ; M Indicates torque, unit N × m ; oh Indicates the angular velocity of rotation, in rad / s; M 11 Represents unit moment, unit N × m ; n 11 Indicates unit speed, unit rpm ; Q 11 Indicates unit flow, unit m 3 / s .

[0020] According to formula (2), n 11 - Q 11 and n 11 - M 11 Find the highest efficiency point in the first quadrant curve of Q 11 and n 11 That is Q 11N and n 11N ;

[0021] When the blade placement angle on the high-pressure side is unknown, the guide vane opening at the highest efficiency point is the guide vane opening corresponding to the optimal point. α N ,pass α N and Q 11N and n 11N Calculate the blade placement angle on the high pressure side β 1b As shown in formula (3):

[0022] (3)

[0023] In formula (3): β 1b Indicates the blade placement angle on the high-pressure side, unit: °; β 1N Indicates the blade placement angle on the high pressure side of the optimal point, in degrees; V m1N The high pressure side axial velocity component representing the optimal point, U 1N The high pressure side circumferential speed represents the optimal point, V u1N It represents the circumferential component of the absolute velocity on the high pressure side of the optimum point. The unit of velocity component is m / s. n N Indicates the speed at the optimal point, unit rpm ; Q 11N Represents the unit flow rate at the optimal point, unit m 3 / s ; α N Indicates the guide vane opening at the optimal point, in degrees; b 1Indicates the high pressure side blade height, unit m ; n 11N Indicates the unit speed of the optimal point, unit rpm .

[0024] When the specific speed n s When unknown, as shown in formula (4):

[0025] (4)

[0026] In formula (4): n s Indicates specific speed, unit: m×kW; n N Indicates the optimal speed, unit rpm ; N N Indicates the optimal output, unit kW ; H N Indicates the optimal head, unit m ; n 11N Indicates the optimal unit speed, unit rpm ; r Indicates density, unit kg / m 3 ; g Indicates the acceleration due to gravity, unit m / s 2 ; Q 11N represents the optimal unit flow rate, m 2 / s ; D 2 Indicates the diameter of the runner low pressure side, unit m ; or N Indicates the optimum efficiency.

[0027] The first quadrant characteristic curve of the pump turbine in S2 and the corresponding characteristic parameters include at least D 1 , D 2 , b 1 , n 11 - Q 11 and n 11 - M 11 ; or, the samples obtained in S2 at least include D1 , D 2 , b 1 , n 11 - Q 11 、α N , n 11N and Q 11N .

[0028] S3 specifically: adopt improved Suter The transformation converts the characteristic curve of the first quadrant of the pump-turbine into a single-valued function, that is, x-WH The curve is shown in equation (5) and equation (6):

[0029] (5)

[0030] (6)

[0031] In formula (5) and formula (6): k 1 , C y , C h is a constant; k 1 The value range is 0.5~1.2, C y The value range is 0.1~0.3; C h The value range is 0.4~0.6; n 11 Indicates unit speed, unit rpm ; Q 11 Indicates unit flow rate, m 2 / s ; Q 11N represents the unit flow rate at the optimal point, m 2 / s ; n 11 Indicates the unit speed, rpm ; n 11N represents the unit speed of the optimal point, rpm . α Indicates the guide vane opening, unit: °; α N Indicates the guide vane opening at the optimal point, in degrees.

[0032] Combining equation (5) and equation (6), we get n 11 (x, WH) and Q 11 (x, WH), as shown in equations (7), (8) and (9):

[0033] (7)

[0034] In formula (7): n 11 Indicates unit speed, unit rpm ; n 11N Indicates the unit speed of the optimal point, unit rpm ; Q 11 Indicates unit flow rate, m 2 / s ; Q 11N represents the unit flow rate at the optimal point, m 2 / s ; α Indicates the guide vane opening, unit: °; α N Indicates the guide vane opening at the optimal point, in degrees; k 1 , C y , C h is a constant.

[0035] make:

[0036] (8)

[0037] but:

[0038] (9)

[0039] In formula (8) and formula (9): n 11 Indicates unit speed, unit rpm ; Q 11 Indicates unit flow rate, m 2 / s ; n 11N Indicates the unit speed of the optimal point, unit rpm ; Q 11N represents the unit flow rate at the optimal point, m 2 / s .

[0040] The artificial neural network structure constructed in S4 is as follows:

[0041] The artificial neural network structure is set to a three-layer network structure of input layer-hidden layer-output layer, and the number of hidden layer nodes is determined according to formula (10):

[0042] (10)

[0043] In formula (10), N i Represents the number of neurons in the input layer; N o Represents the number of neurons in the output layer; N s represents the number of samples in the training set; δ represents an arbitrary constant between 2 and 10;

[0044] Input parameters include , , and operating parameter characteristics , n 11N , Q 11N , x ,Right now N i =6; the output parameter is c ,Right now N o =1;

[0045] In S4, the training set is set to 80% of the total sample set, and the test set is set to 20% of the total sample set.

[0046] S5 is as follows:

[0047] The coefficient of determination is shown in formula (11) R 2 and the mean square error as shown in formula (12) MSE Evaluate the prediction results of the training set and the test set. Among the evaluation indicators of the model evaluation, MSE The smaller it is, the more accurate the prediction result is; R 2 The closer it is to 1, the closer the predicted trend is to the true value trend; when the result is overfitting or underfitting, modify d Repeat the training process until the error is minimized;

[0048] (11)

[0049] (12)

[0050] In formula (11) and (12): c a，i and c p，i Respectively represent i Data point c Value actual value and predicted value; Represents a sample set c The mean of the values.

[0051] S6 is specifically: after the artificial neural network training is completed, the characteristic parameter pair is input c After the prediction is completed, several x - c Prediction point, several x - c The prediction points are derived from S3 Suter The inverse transformation yields several n 11 - Q 11 Data points, use a smooth curve to n 11 - Q 11 The data points are connected to obtain the first quadrant characteristic curve of the pump-turbine.

[0052] The beneficial effects of the present invention are:

[0053] The present invention obtains a large number of first quadrant curves of pump-turbines and required runner characteristic parameters through existing literature, and obtains the response relationship between the characteristic parameters of the pump-turbine and the first quadrant characteristic curve through sample conversion, sample reduction, building an artificial neural network, training and verification, thereby realizing the rapid acquisition of the S characteristic curve in the design stage of the pump-turbine. Compared with the prior art, the present invention does not need to make a model machine and perform measurements in the design stage of the pump-turbine, which solves the technical problem that the existing pump-turbine needs to make a model machine and perform measurements to obtain the S characteristic curve of the pump-turbine in the design stage, resulting in high production costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 Schematic diagram of the velocity triangle on the high-pressure side of the pump turbine in Embodiment 3 of the present invention;

[0055] Figure 2 1 is a schematic diagram of the first quadrant curve of the pump turbine prediction obtained by using the artificial neural network prediction in Example 8 of the present invention;

[0056] Figure 3 Under the rated guide vane opening in Example 8 of the present invention, the change Schematic diagram of the impact on the first quadrant curve of the pump turbine;

[0057] Figure 4 Under the rated guide vane opening in Example 8 of the present invention, the change Schematic diagram of the impact on the first quadrant curve of a pump-turbine. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0059] Example 1

[0060] The pump-turbine S characteristic curve prediction method based on artificial neural network disclosed in this embodiment specifically includes the following steps:

[0061] S1: The pump-turbine unit speed-flow first quadrant curve and the required runner characteristic parameters obtained from the literature are used to determine the pump-turbine characteristic parameters related to the S characteristic, and the characteristic parameters are reduced to obtain the sample set required for artificial neural network prediction;

[0062] S2: Combined with S1, obtain the first quadrant characteristic curve and corresponding characteristic parameters of the pump-turbine in the sample set;

[0063] S3: Through improvement Suter Transformation is used to process the first quadrant characteristic curve of the pump-turbine in S2 to ensure the single value of the first quadrant characteristic curve, thereby facilitating the fitting of the artificial neural network; at the same time, the Suter Inverse transformation, we get n 11 - Q 11 Data points to make the predictions intuitive;

[0064] S4: Construct an artificial neural network structure, and combine the characteristic parameters obtained in S2 with those obtained after processing in S3. n 11 - Q 11 The data points are taken as a whole sample set, and then the sample set is divided into a training set and a test set. The training set is used to train the artificial neural network, and the test set is used to verify the generalization ability of the constructed artificial neural network model;

[0065] S5: By mean square error and coefficient of determination R 2 Verify the rationality and accuracy of the artificial neural network structure;

[0066] S6: Use the trained artificial neural network to predict the first quadrant curve of the pump turbine containing the characteristic parameters required for prediction. After the prediction is completed, the prediction results are analyzed. Suter Inverse transformation to obtain the first quadrant of the pump turbine n 11 - Q 11 Characteristic curve.

[0067] Example 2

[0068] On the basis of Example 1, according to existing research, the characteristic parameters in S1 include: S characteristic and the turbine operating speed of the pump turbine n sT , runner high pressure side diameter D 1 , Runner low pressure side diameter D 2 , Blade high pressure side height b 1 And the blade placement angle β 1b , guide vane opening α , the flow rate at the optimal point Q N and the optimal speed n N , the guide vane opening at the optimal point α N , Unit speed on the characteristic curve Q 11 and unit flow n 11 , a total of eleven parameters.

[0069] Example 3

[0070] On the basis of Example 2, in the process of obtaining the characteristic curve, there are problems such as unknown optimal point, unknown specific speed, unknown high-pressure side blade placement angle, etc., and at the same time, the above parameters have the problem of cross-correlation. Through parameter processing, data collection and data reduction are facilitated. The characteristic parameter reduction processing in S1 is specifically as follows:

[0071] The characteristic curve parameters are defined as shown in formula (1):

[0072] (1)

[0073] In formula (1): n 11 Indicates unit speed, unit rpm ; n Indicates the speed, unit rpm ;Q 11 Indicates unit flow, unit m 3 / s ; Q Indicates flow rate, unit m 3 / s ; M 11 Represents unit moment, unit N × m ; M Indicates torque, unit N × m ;; D 2 Indicates the diameter of the runner low pressure side, unit m ; H Indicates rated head, unit m ;

[0074] When the optimal point is unknown, the efficiency calculation formula of the rotating machinery shown in formula (2) shows that:

[0075] (2)

[0076] In formula (2): or represents the pump-turbine efficiency; r Indicates density, unit kg / m 3 ; g Indicates the acceleration due to gravity, unit m / s 2 ; Q Indicates flow rate, unit m 3 / s ; H Indicates rated head, unit m ; M Indicates torque, unit N × m ; oh Indicates the angular velocity of rotation, in rad / s; M 11 Represents unit moment, unit N × m ; n 11 Indicates unit speed, unit rpm ; Q 11 Indicates unit flow, unit m 3 / s .

[0077] According to formula (2), n11 - Q 11 and n 11 - M 11 Find the highest efficiency point in the first quadrant curve of Q 11 and n 11 That is Q 11N and n 11N ;

[0078] When the blade placement angle on the high-pressure side is unknown, the guide vane opening at the highest efficiency point is the optimal guide vane opening. α N ,like Figure 1 As shown, according to the principle of optimal point without impact on the high pressure side inlet, α N and Q 11N and n 11N The blade placement angle on the high pressure side can be calculated β 1b As shown in formula (3):

[0079] (3)

[0080] In formula (3): β 1b Indicates the blade placement angle on the high-pressure side, unit: °; β 1N Indicates the blade placement angle on the high pressure side of the optimal point, in degrees; V m1N The high pressure side axial velocity component representing the optimal point, U 1N The high pressure side circumferential speed represents the optimal point, V u1N It represents the circumferential component of the absolute velocity on the high pressure side of the optimal point. The units of the velocity components are m / s ; n N Indicates the speed at the optimal point, unit rpm ; Q 11N Represents the unit flow rate at the optimal point, unit m 3 / s ; α N Indicates the guide vane opening at the optimal point, in degrees; b 1 Indicates the high pressure side blade height, unit m; n 11N Indicates the unit speed of the optimal point, unit rpm ;; D 1 Indicates the diameter of the high pressure side of the runner, in units m .

[0081] When the specific speed n s When unknown, as shown in formula (4):

[0082] (4)

[0083] In formula (4): n s Indicates specific speed, unit: m×kW; n N Indicates the optimal speed, unit rpm ; N N Indicates the optimal output, unit kW ; H N Indicates the optimal head, unit m ; n 11N Indicates the optimal unit speed, unit rpm ; r Indicates density, unit kg / m 3 ; g Indicates the acceleration due to gravity, unit m / s 2 ; Q 11N represents the optimal unit flow rate, m 2 / s ; D 2 Indicates the diameter of the runner low pressure side, unit m ; or N Indicates the optimum efficiency.

[0084] Example 4

[0085] Based on Example 3 and the conclusion obtained in S1, the first quadrant characteristic curve of the pump turbine in S2 and the corresponding characteristic parameters include at least D 1 , D 2 , b 1 , n 11 - Q 11 and n11 - M 11 ; or, the samples obtained in S2 at least include D 1 , D 2 , b 1 , n 11 - Q 11 、α N , n 11N and Q 11N , the collection of samples in this embodiment is based on the above conclusions.

[0086] Example 5

[0087] On the basis of Example 4, when the pump turbine is in the S characteristic zone, a unit speed at the same opening will correspond to three different unit flows, which will cause deviations in the artificial neural network when predicting the external characteristics of the S zone.

[0088] S3 is specifically:

[0089] The improved Suter transformation is used to transform the characteristic curve of the first quadrant of the pump-turbine into a single-valued function, that is, x-WH The curve is shown in equation (5) and equation (6):

[0090] (5)

[0091] (6)

[0092] In formula (5) and formula (6): k 1 , C y , C h is a constant, where k 1 and C y It is a constant introduced to prevent large interpolation errors caused by large opening curves being too dense and small opening curves being too sparse. C h It is to prevent WH (x, y) The constant introduced by the division by zero problem in the denominator; k 1 The value range is 0.5~1.2, C y The value range is 0.1~0.3; C hThe value range is 0.4~0.6; n 11 Indicates unit speed, unit rpm ; Q 11 Indicates unit flow rate, m 2 / s ; Q 11N represents the unit flow rate at the optimal point, m 2 / s ; n 11 Indicates the unit speed, rpm ; n 11N represents the unit speed of the optimal point, rpm . α Indicates the guide vane opening, unit: °; α N Indicates the guide vane opening at the optimal point, in degrees.

[0093] Combining equation (5) and equation (6), we get n 11 (x, WH) and Q 11 (x, WH), as shown in equations (7), (8), and (9):

[0094] (7)

[0095] In formula (7): n 11 Indicates unit speed, unit rpm ; Q 11 Indicates unit flow rate, m 2 / s ; α Indicates the guide vane opening, unit: °; n 11N Indicates the unit speed of the optimal point, unit rpm ; Q 11N represents the unit flow rate at the optimal point, m 2 / s ; α Indicates the guide vane opening, unit: °; α N Indicates the guide vane opening at the optimal point, in degrees; k 1 , C y , C h is a constant.

[0096] make:

[0097] (8)

[0098] but:

[0099] (9)

[0100] In formula (8) and formula (9): n 11 Indicates unit speed, unit rpm ; Q 11 Indicates unit flow rate, m 2 / s ; n 11N Indicates the unit speed of the optimal point, unit rpm ; Q 11N represents the unit flow rate at the optimal point, m 2 / s .

[0101] Example 6

[0102] On the basis of Example 5, according to S2 and S3, the geometric similarity principle of the pump turbine is considered at the same time, and the input sample includes the geometric characteristics of the pump turbine. , , and operating parameter characteristics , n 11N , Q 11N , x , the output parameters are WH However, robustness tests show that using WH As the output parameter prediction, under the condition of large opening, a small error is likely to cause c The value is too large to make the inverse transformation impossible. Due to a certain opening curve of a pump turbine to be predicted, after the parameters are determined is a constant, c Only with WH Therefore, c As the output parameter. The artificial neural network structure constructed in S4 is specifically as follows: the artificial neural network structure is set to a three-layer network structure of input layer-hidden layer-output layer, and the number of hidden layer nodes is determined according to formula (10):

[0103] (10)

[0104] In formula (10),N i Represents the number of neurons in the input layer; N o Represents the number of neurons in the output layer; N s Indicates the number of samples in the training set; d represents any constant from 2 to 10;

[0105] Input parameters include , , and operating parameter characteristics , n 11N , Q 11N , x ,Right now N i =6; the output parameter is c ,Right now N o =1;

[0106] The training set is set to 80% of the total number of sample sets, and the training set is used to construct the optimal weights of each neuron in the neural network model; the test set is set to 20% of the total number of sample sets, and is used to evaluate the final performance of the constructed neural network model; after the neural network structure is set, the artificial neural network is trained;

[0107] Example 7

[0108] On the basis of Example 6, S5 is specifically:

[0109] After the training is completed, the determination coefficient is shown in formula (11) R 2 and the mean square error as shown in formula (12) MSE Evaluate the prediction results of the training set and the test set. Among the evaluation indicators of the model evaluation, MSE The smaller it is, the more accurate the prediction result is; R 2 The closer it is to 1, the closer the predicted trend is to the true value trend. When the result is overfitting or underfitting, modify d Repeat the training process until the error is minimized;

[0110] (11)

[0111] (12)

[0112] In formula (11) and (12): c a，i and c p，i Respectively represent iData point c Value actual value and predicted value; Represents a sample set c The mean of the values.

[0113] Example 8

[0114] On the basis of Example 7, S6 specifically comprises: after the artificial neural network training is completed, input the characteristic parameter pair c After the prediction is completed, several x - c Prediction points, several x - c The prediction points are derived from S3 Suter The inverse transformation yields several n 11 -Q 11 Data points, use a smooth curve to n 11 -Q 11 The data points are connected to obtain Figure 2 The first quadrant characteristic curve of the pump turbine is shown. Figure 2 In this pump turbine =0.65, =0.137, n 11N =51.2 rpm , Q 11N =650 L / s .

[0115] In addition, by using the method of the present invention, the input parameters of the same group of pump-turbine samples are changed, and the response relationship between the input parameters and the first quadrant characteristic curve of the pump-turbine can be obtained, so as to predict the following: Figure 3 and Figure 4 The correlation between the pump turbine S characteristics and the input parameters is shown in Fig. It significantly affects the characteristic parameters of the first quadrant of the pump-turbine. As this parameter increases, the unit flow rate at the same unit speed of the turbine moves downward, and the curvature of the S-zone curve is weakened. The smaller impeller high-pressure side height makes the characteristic curve of the first quadrant of the pump-turbine move upward, but has little effect on the curvature of the S-zone curve.

[0116] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0117] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0118] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the S-characteristic curve of a pump-turbine based on an artificial neural network, characterized in that: The specific steps include: S1: Determine the characteristic parameters of the pump-turbine related to the S characteristic, and reduce the order of the characteristic parameters to obtain the sample set required for artificial neural network prediction; S2: Obtain the first quadrant characteristic curve and corresponding characteristic parameters of the pump-turbine in the sample set; S3: Through improvement Suter Transformation, processing the first quadrant characteristic curve of the pump turbine in S2, and deriving Suter Inverse transformation, we get n 11 - Q 11 Data points; S4: Construct an artificial neural network structure, and combine the characteristic parameters obtained in S2 with those obtained after processing in S3. n 11 -Q 11 The data points are taken as a whole sample set, and then the sample set is divided into a training set and a test set. The training set is used to train the artificial neural network, and the test set is used to verify the generalization ability of the constructed artificial neural network model; S5: By mean square error and coefficient of determination R 2 Verify the rationality and accuracy of the artificial neural network structure; S6: Use the trained artificial neural network to predict the first quadrant curve of the pump turbine containing the characteristic parameters required for prediction. After the prediction is completed, the prediction results are analyzed. Suter Inverse transformation to obtain the first quadrant of the pump turbine n 11 - Q 11 Characteristic curve; The characteristic parameters in S1 include: S characteristic and the specific speed of the turbine operating condition of the pump turbine n sT , runner high pressure side diameter D 1. Runner low pressure side diameter D 2. Blade high pressure side height b 1 and blade placement angle β 1b , guide vane opening α , the flow rate at the optimal point Q N and the optimal speed n N , the guide vane opening at the optimal point α N , Unit flow on the characteristic curve Q 11 and unit speed n 11 .

2. The method for predicting the S characteristic curve of a pump-turbine based on an artificial neural network according to claim 1, characterized in that: The specific reduction process of the characteristic parameters in S1 is as follows: The characteristic curve parameters are defined as shown in formula (1): （1） In formula (1): n 11 Indicates unit speed, unit r / min ; n Indicates the speed, unit r / min ; Q 11 Indicates unit flow, unit m 3 / s ; Q Indicates flow rate, unit m 3 / s ; M 11 Represents unit moment, unit N × m ; M Indicates torque, unit N × m ; D 2 represents the diameter of the low-pressure side of the runner, unit m ; H Indicates rated head, unit m ; When the optimal point is unknown, the efficiency calculation formula of the rotating machinery shown in formula (2) shows that: （2） In formula (2): η represents the pump-turbine efficiency; ρ Indicates density, unit kg / m 3 ; g Indicates the acceleration due to gravity, unit m / s 2 ; Q Indicates flow rate, unit m 3 / s ; H Indicates rated head, unit m ; M Indicates torque, unit N × m ; ω Indicates the angular velocity of rotation, in rad / s; M 11 Represents unit moment, unit N × m ; n 11 Indicates unit speed, unit r / min ; Q 11 Indicates unit flow, unit m 3 / s ; According to formula (2), n 11 - Q 11 and n 11 - M 11 Find the highest efficiency point in the first quadrant curve of Q 11 and n 11 That is Q 11N and n 11N ; When the blade placement angle on the high-pressure side is unknown, the guide vane opening at the highest efficiency point is the optimal guide vane opening. α N ,pass α N and Q 11N and n 11N Calculate the blade placement angle on the high pressure side β 1b As shown in formula (3): （3） In formula (3): β 1b Indicates the blade placement angle on the high-pressure side, unit: °; β 1N Indicates the blade placement angle on the high pressure side of the optimal point, in degrees; V m1N The high pressure side axial velocity component representing the optimal point, U 1N The high pressure side circumferential speed represents the optimal point, V u1N It represents the circumferential component of the absolute velocity on the high pressure side of the optimal point. The units of the velocity components are m / s ; n N Indicates the speed at the optimal point, unit r / min ; Q 11N Represents the unit flow rate at the optimal point, unit m 3 / s ; α N Indicates the guide vane opening at the optimal point, in degrees; b 1 indicates the blade height on the high pressure side, unit m ; n 11N Indicates the unit speed of the optimal point, unit r / min ; D 1 represents the diameter of the high pressure side of the runner, unit m ; When the specific speed n s When unknown, as shown in formula (4): （4） In formula (4): n s Indicates specific speed, unit: m×kW; n N Indicates the optimal speed, unit r / min ; N N Indicates the optimal output, unit kW ; H N Indicates the optimal head, unit m ; n 11N Indicates the optimal unit speed, unit r / min ; ρ Indicates density, unit kg / m 3 ; g Indicates the acceleration due to gravity, unit m / s 2 ; Q 11N represents the optimal unit flow rate, m 2 / s ; D 2 represents the diameter of the low-pressure side of the runner, unit m ; η N Indicates the optimum efficiency.

3. The method for predicting the S characteristic curve of a pump-turbine based on an artificial neural network according to claim 2, characterized in that: The first quadrant characteristic curve of the pump turbine in S2 and the corresponding characteristic parameters at least include D 1. D 2. b 1. n 11 - Q 11 and n 11 - M 11 ; or, the samples obtained in S2 at least include D 1. D 2. b 1. n 11 - Q 11 、α N , n 11N and Q 11N .

4. The method for predicting the S characteristic curve of a pump-turbine based on an artificial neural network according to claim 3 is characterized in that: The S3 is specifically: adopting improved Suter The transformation converts the characteristic curve of the first quadrant of the pump-turbine into a single-valued function, that is, x-WH The curve is shown in equation (5) and equation (6): （5） （6） In formula (5) and formula (6): k 1. C y , C h is a constant; k 1The value range is 0.5~1.2, C y The value range is 0.1~0.3; C h The value range is 0.4~0.6; n 11 Indicates unit speed, unit r / min ; Q 11 Indicates unit flow rate, m 2 / s ; Q 11N represents the unit flow rate at the optimal point, m 2 / s ; n 11 Indicates the unit speed, r / min ; n 11N represents the unit speed of the optimal point, r / min, α Indicates the guide vane opening, unit: °; α N Indicates the guide vane opening at the optimal point, in degrees; Combining equation (5) and equation (6), we get n 11 ( x , WH )and Q 11 ( x , WH ), as shown in equations (7), (8) and (9): （7） In formula (7): n 11 Indicates unit speed, unit r / min ; n 11N Indicates the unit speed of the optimal point, unit r / min ; Q 11 Indicates unit flow rate, m 2 / s ; Q 11N represents the unit flow rate at the optimal point, m 2 / s ; α Indicates the guide vane opening, unit: °; α N Indicates the guide vane opening at the optimal point, in degrees; k 1. C y , C h is a constant; make: but: In formula (8) and formula (9): n 11 Indicates unit speed, unit r / min ; n 11N Indicates the unit speed of the optimal point, unit r / min ; Q 11 Indicates unit flow rate, m 2 / s ; Q 11N represents the unit flow rate at the optimal point, m 2 / s .

5. The method for predicting the S characteristic curve of a pump-turbine based on an artificial neural network according to claim 4, characterized in that: The artificial neural network structure constructed in S4 is specifically: the artificial neural network structure is set to a three-layer network structure of input layer-hidden layer-output layer, and the number of hidden layer nodes is determined according to formula (10): （10） In formula (10), N i Represents the number of neurons in the input layer; N o Represents the number of neurons in the output layer; N s represents the number of samples in the training set; δ represents an arbitrary constant between 2 and 10; Input parameters include , , and operating parameter characteristics , n 11N , Q 11N , x ,Right now N i =6; the output parameter is c ,Right now N o =1; in S4, the training set is set to 80% of the total number of sample sets, and the test set is set to 20% of the total number of sample sets.

6. The method for predicting the S characteristic curve of a pump-turbine based on an artificial neural network according to claim 5, characterized in that: The S5 is specifically: The coefficient of determination is shown in formula (11) R 2 And the mean square error as shown in formula (12) MSE Evaluate the prediction results of the training set and the test set. Among the evaluation indicators of the model evaluation, MSE The smaller it is, the more accurate the prediction result is; R 2 The closer it is to 1, the closer the predicted trend is to the true value trend; When the result is overfitting or underfitting, modify δ Repeat the training process until the error is minimized; （11） （12） In formula (11) and (12): c a，i and c p，i Respectively represent i Data point c Value actual value and predicted value; Represents a sample set c The mean of the values.

7. The method for predicting the S characteristic curve of a pump-turbine based on an artificial neural network according to claim 6, characterized in that: The S6 is specifically: After the artificial neural network training is completed, the input feature parameters are c After the prediction is completed, several x - c Prediction point, several x - c The predicted points are derived according to S3 Suter The inverse transformation yields several n 11 - Q 11 Data points, use a smooth curve to n 11 - Q 11 The data points are connected to obtain the first quadrant characteristic curve of the pump-turbine.

Images