Compressor performance prediction method and device and electronic equipment

Abstract

The invention provides a compressor performance prediction method and device and electronic equipment, and belongs to the technical field of compressors. The method comprises the following steps: inputting a to-be-tested performance curve parameter and a working condition into a performance prediction model to obtain a performance prediction result output by the performance prediction model; inputting the to-be-tested performance curve parameters and the working conditions into the surge point prediction model to obtain a surge point prediction result output by the surge point prediction model; inputting the to-be-tested performance curve parameters and the working conditions into the blockage point prediction model to obtain a blockage point prediction result output by the blockage point prediction model; the surge point prediction model is obtained by training on the basis of a plurality of sample performance curves and surge point labels corresponding to the sample performance curves; the blockage point prediction model is obtained based on training of a plurality of sample performance curves and blockage point labels corresponding to the sample performance curves. The method is used for solving the problem of how to quickly and accurately predict the performance of the compressor on the basis of limited performance curve data.

Description

Technical Field

[0001] The present invention relates to the technical field of compressors, and in particular to a compressor performance prediction method, a compressor performance prediction device, an electronic device, a machine-readable storage medium and a computer program product. Background Art

[0002] The compressor performance curve is an intuitive expression of the compressor performance. A set of compressor performance curves includes multiple curves, such as flow-pressure ratio curve, flow-power curve, flow-efficiency curve, etc. Each curve is divided into multiple performance curves according to the different working conditions of the compressor. On this basis, there are also special curves such as surge line and blockage line.

[0003] After the compressor is designed and manufactured, it will be tested, and several sets of performance curve data under different working conditions will be obtained and handed over to the user as factory performance curve data. However, when the user uses the compressor, its working conditions are difficult to keep consistent with the working conditions of the factory performance curve. If the compressor performance is predicted and adjusted with reference to the factory performance curve data, it will cause large errors. Even if the working conditions are basically the same, the limited performance curve data is difficult to include all operating points, and the operating points far away from the performance curve are difficult to predict. The current solution is to determine whether the performance of the machine can meet the user's target operating conditions through experiments, but the costs in all aspects are relatively high. The second is to convert the performance curve based on the existing performance curve data under certain similarity assumptions, and then predict the performance of the operating points through interpolation, but the similarity assumption is difficult to fully meet and the accuracy is difficult to guarantee.

[0004] Therefore, in view of the changing target operating conditions and usage environment, how to quickly and accurately predict the compressor performance based on limited performance curve data has become an urgent problem to be solved. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a compressor performance prediction method, a compressor performance prediction device, an electronic device, a machine-readable storage medium and a computer program product to solve the problem of how to quickly and accurately predict the compressor performance based on limited performance curve data.

[0006] In order to achieve the above object, an embodiment of the present invention provides a compressor performance prediction method, comprising:

[0007] Obtaining the performance curve parameters and working conditions to be tested of the compressor;

[0008] Inputting the performance curve parameters to be measured and the working conditions into a performance prediction model to obtain a performance prediction result output by the performance prediction model;

[0009] Inputting the performance curve parameter to be measured and the operating conditions into a surge point prediction model to obtain a surge point prediction result output by the surge point prediction model;

[0010] Inputting the performance curve parameters to be measured and the working conditions into a congestion point prediction model to obtain a congestion point prediction result output by the congestion point prediction model;

[0011] Among them, the surge point prediction model is trained based on multiple sample performance curves and the surge point labels corresponding to each of the sample performance curves; the congestion point prediction model is trained based on multiple sample performance curves and the congestion point labels corresponding to each of the sample performance curves; each of the sample performance curves is constructed based on multiple sample performance curve parameters, sample working conditions and multiple model prediction results input by the trained performance prediction model; the surge point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set surge line; the congestion point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set congestion line.

[0012] Optionally, the surge point prediction model is trained by the following steps:

[0013] Obtaining multiple sample performance curve parameters and multiple sample working conditions;

[0014] Training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters;

[0015] Predicting the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results;

[0016] Constructing a single sample performance curve based on all sample performance curve parameters under each of the sample working conditions and the model prediction results corresponding to each sample performance curve parameter to obtain multiple sample performance curves;

[0017] determining a plurality of target surge points where the plurality of sample performance curves intersect a set surge line as the surge point labels;

[0018] The surge point prediction model is trained based on the sample performance curve parameters, the model prediction results and the surge point labels corresponding to the sample performance curves on each of the sample performance curves.

[0019] Optionally, the congestion point prediction model is trained by the following steps:

[0020] Obtaining multiple sample performance curve parameters and multiple sample working conditions;

[0021] Training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters;

[0022] Predicting the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results;

[0023] Constructing a single sample performance curve based on all sample performance curve parameters under each of the sample working conditions and the model prediction results corresponding to each sample performance curve parameter to obtain multiple sample performance curves;

[0024] determining a plurality of target blocking points where the plurality of sample performance curves intersect a set blocking line as the blocking point labels;

[0025] The congestion point prediction model is trained based on the sample performance curve parameters on each of the sample performance curves, the model prediction results and the congestion point labels corresponding to the sample performance curves.

[0026] Optionally, the training of the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions and the performance label corresponding to each of the sample performance curve parameters includes:

[0027] Repeat the following steps until the set number of iterations is reached:

[0028] Input the target sample performance curve parameters and target sample working conditions into the performance prediction model to obtain

[0029] to the model output result of the performance prediction model; the target sample performance curve parameter is any one of the multiple sample performance curve parameters, and the target sample working condition is any one of the multiple sample working conditions;

[0030] Calculate the loss function based on the target sample performance curve parameters, the real performance labels corresponding to the target sample working conditions, and the model output results;

[0031] Model parameters of the loss function are adjusted based on the loss function.

[0032] Optionally, the performance prediction model is constructed using any one of a BP neural network, a generalized regression neural network and a probabilistic neural network.

[0033] Optionally, the surge point prediction model is constructed by using any one of a BP neural network, a generalized regression neural network and a probabilistic neural network.

[0034] Optionally, the congestion point prediction model is constructed by using any one of BP neural network, generalized regression neural network and probabilistic neural network.

[0035] Optionally, the performance curve parameter to be measured includes at least one of pressure, polynomial efficiency, flow rate, power and compression ratio of the compressor.

[0036] Optionally, the operating conditions include at least one of a rotation speed of the compressor, an inlet pressure of the compressor, an inlet temperature of the compressor, and an inlet guide vane angle of the compressor.

[0037] The embodiment of the present invention further provides a compressor performance prediction device, comprising:

[0038] An acquisition module, used to acquire the performance curve parameters and working conditions to be tested of the compressor;

[0039] A first prediction module, used for inputting the performance curve parameters to be measured and the working conditions into a performance prediction model to obtain a performance prediction result output by the performance prediction model;

[0040] A second prediction module, used for inputting the performance curve parameter to be measured and the working condition into a surge point prediction model to obtain a surge point prediction result output by the surge point prediction model;

[0041] A third prediction module, used for inputting the performance curve parameter to be measured and the working condition into a congestion point prediction model to obtain a congestion point prediction result output by the congestion point prediction model;

[0042] The surge point prediction model is based on a plurality of sample performance curves and each of the sample performance curves.

[0043] The surge point label corresponding to the energy curve is trained; the blocking point prediction model is trained based on multiple sample performance curves and the blocking point label corresponding to each sample performance curve; each sample performance curve is constructed based on multiple sample performance curve parameters, sample working conditions and multiple model prediction results inputted from the trained performance prediction model; the surge point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set surge line; the blocking point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set blocking line.

[0044] Optionally, the surge point prediction model is trained by the following steps:

[0045] Obtaining multiple sample performance curve parameters and multiple sample working conditions;

[0046] Training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters;

[0047] Predicting the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results;

[0048] Constructing a single sample performance curve based on all sample performance curve parameters under each of the sample working conditions and the model prediction results corresponding to each sample performance curve parameter to obtain multiple sample performance curves;

[0049] determining a plurality of target surge points where the plurality of sample performance curves intersect a set surge line as the surge point labels;

[0050] The surge point prediction model is trained based on the sample performance curve parameters, the model prediction results and the surge point labels corresponding to the sample performance curves on each of the sample performance curves.

[0051] Optionally, the congestion point prediction model is trained by the following steps:

[0052] Obtaining multiple sample performance curve parameters and multiple sample working conditions;

[0053] Training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters;

[0054] Predicting the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results;

[0055] Constructing a single sample performance curve based on all sample performance curve parameters under each of the sample working conditions and the model prediction results corresponding to each sample performance curve parameter to obtain multiple sample performance curves;

[0056] determining a plurality of target blocking points where the plurality of sample performance curves intersect a set blocking line as the blocking point labels;

[0057] The congestion point prediction model is trained based on the sample performance curve parameters on each of the sample performance curves, the model prediction results and the congestion point labels corresponding to the sample performance curves.

[0058] Optionally, the training of the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions and the performance label corresponding to each of the sample performance curve parameters includes:

[0059] Repeat the following steps until the set number of iterations is reached:

[0060] Inputting a target sample performance curve parameter and a target sample working condition into a performance prediction model to obtain a model output result of the performance prediction model; the target sample performance curve parameter is any one of the multiple sample performance curve parameters, and the target sample working condition is any one of the multiple sample working conditions;

[0061] Calculate the loss function based on the target sample performance curve parameters, the real performance labels corresponding to the target sample working conditions, and the model output results;

[0062] Model parameters of the loss function are adjusted based on the loss function.

[0063] Optionally, the performance prediction model is constructed using any one of a BP neural network, a generalized regression neural network and a probabilistic neural network.

[0064] Optionally, the surge point prediction model is constructed by using any one of a BP neural network, a generalized regression neural network and a probabilistic neural network.

[0065] Optionally, the congestion point prediction model is constructed by using any one of BP neural network, generalized regression neural network and probabilistic neural network.

[0066] Optionally, the performance curve parameters to be measured include pressure, variable efficiency, flow rate,

[0067] At least one of power and compression ratio.

[0068] Optionally, the operating conditions include at least one of a rotation speed of the compressor, an inlet pressure of the compressor, an inlet temperature of the compressor, and an inlet guide vane angle of the compressor.

[0069] On the other hand, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned compressor performance prediction method when executing the program.

[0070] On the other hand, the present invention further provides a machine-readable storage medium having a computer program stored thereon, and the computer program implements the above-mentioned compressor performance prediction method when executed by a processor.

[0071] On the other hand, the present invention further provides a computer program product, comprising a computer program, wherein the computer program implements the above-mentioned compressor performance prediction method when executed by a processor.

[0072] Through the above technical solution, the embodiment of the present invention constructs a sample performance curve based on multiple sample performance curve parameters, sample working conditions and multiple model prediction results input in the performance prediction model training, and then performs surge point and blockage point prediction based on multiple sample performance curves and surge point labels and blockage point labels corresponding to each sample performance curve. The performance of the compressor is predicted based on the performance prediction model. Therefore, the embodiment of the present invention builds a neural network model based on sample performance curve parameters and sample working conditions, which can predict performance curves including surge points and blockage points under different working conditions. Compared with the existing method, it is simple and easy to implement, and can achieve rapid and accurate prediction of compressor performance.

[0073] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0075] Figure 1 It is a schematic flow chart of the compressor performance prediction method provided by the present invention;

[0076] Figure 2 It is a schematic diagram of the training process of the surge point prediction model and the blockage point prediction model provided by the present invention;

[0077] Figure 3 It is a schematic diagram of the structure of the BP neural network used in the present invention;

[0078] Figure 4 is a schematic diagram of error back propagation provided by the present invention;

[0079] Figure 5 It is a schematic diagram of obtaining the surge point and the blockage point provided by the present invention;

[0080] Figure 6 It is a structural schematic diagram of a compressor performance prediction device provided by the present invention;

[0081] Figure 7 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0082] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0083] Method Embodiment

[0084] Please refer to Figure 1 , an embodiment of the present invention provides a compressor performance prediction method, comprising:

[0085] Step 100: Obtain the performance curve parameters and working conditions to be tested of the compressor.

[0086] The electronic device obtains the performance curve parameters to be tested and the working conditions of the compressor. Among them, the performance curve parameters to be tested can be independent variables in various compressor performance curves. The performance curve parameters to be tested include at least one of the pressure, variable efficiency, flow, power and compression ratio of the compressor. For example, the performance curve parameters to be tested can be flow parameters in the pressure-flow curve, efficiency-flow curve, power-flow curve and compression ratio-flow curve; or they can be parameters such as time and compression ratio in the temperature change curve. The working conditions may include the operating conditions of the compressor under various working conditions. Because each curve is divided into multiple performance curves according to different compressor speeds or different compressor inlet guide vane angles. When the compressor inlet pressure and inlet temperature change, the performance of the compressor will change, and the performance curve will also change. Therefore, the working conditions include the speed of the compressor, the inlet pressure of the compressor, and the inlet temperature of the compressor.

[0087] and at least one of an inlet guide vane angle of the compressor. In one embodiment, the operating conditions include a rotation speed of the compressor, an inlet pressure of the compressor, an inlet temperature of the compressor, and an inlet guide vane angle of the compressor.

[0088] Step 200: Input the performance curve parameters to be measured and the working conditions into a performance prediction model to obtain a performance prediction result output by the performance prediction model.

[0089] The electronic device inputs the performance curve parameters to be tested and the working conditions into the performance prediction model to obtain the performance prediction results under the working conditions output by the performance prediction model. The performance prediction model can be trained based on sample performance curve parameters, sample working conditions, and performance labels corresponding to the sample performance curve parameters.

[0090] The sample performance curve parameters may be parameters such as flow, time, and compression ratio in the historical operating data or test data of the compressor. For example, it may be the flow data of the compressor when it was working in the past week. The sample operating conditions include the speed of the compressor, the inlet pressure of the compressor, the inlet temperature of the compressor, and the inlet guide vane angle of the compressor. The performance label is the actual performance value corresponding to the sample performance curve parameter. For example, when the performance curve of the compressor is a pressure-flow curve, the sample performance curve parameter is the compressor flow, and the performance label is the actual value of the pressure corresponding to the compressor flow; when the performance curve of the compressor is a compression ratio-flow curve, the sample performance curve parameter is the compressor flow, and the performance label is the actual value of the compression ratio corresponding to the compressor flow.

[0091] In the embodiment of the present invention, the performance prediction model is unsupervisedly trained based on the sample performance curve parameters, the sample working conditions and the performance labels corresponding to the sample performance curve parameters to obtain a trained performance prediction model. For example, the embodiment of the present invention inputs the sample performance curve parameters into the performance prediction model, denormalizes the output value of the output layer of the performance prediction model and compares it with the true performance value of the sample performance curve parameters, uses the root mean square error (RMSE) of all output values ​​as the error function, and adopts a training algorithm to backpropagate from the last layer to the previous layer, and iteratively calculates the weights and thresholds of each layer of neurons. The training is completed when the error function is minimized, and the corresponding performance prediction model is obtained after verification by test data. Among them, the performance prediction model can use BP neural network, generalized regression neural network and probabilistic neural network

[0092] In one embodiment, the performance prediction model in the embodiment of the present invention can be constructed by using a BP neural network.

[0093] Step 300: Input the performance curve parameters to be measured and the operating conditions into a surge point prediction model to obtain a surge point prediction result output by the surge point prediction model.

[0094] Wherein, the surge point prediction model is obtained by training based on multiple sample performance curves and the surge point labels corresponding to each of the sample performance curves. Each of the sample performance curves is constructed based on multiple sample performance curve parameters, sample working conditions and multiple model prediction results input by the trained performance prediction model. Each model prediction result is predicted by the performance prediction model based on each sample performance curve parameter. In the embodiment of the present invention, a sample performance curve is constructed based on multiple sample performance curve parameters under the same sample working condition and the model prediction result of each sample performance curve parameter. For example, when the compressor speed is 2000r / min, a first sample performance curve is constructed based on data composed of several sample performance curve parameters and model prediction results. The leftmost point of the first sample performance curve is the surge point, that is, the surge point label corresponding to the first sample performance curve. Different sample performance curves and corresponding surge point labels can be obtained by changing the sample working conditions. Therefore, the surge point prediction model is trained based on different sample performance curves and corresponding surge point labels, and can predict the surge points corresponding to different performance curve parameters and working conditions to be tested. Wherein, the surge point prediction model can be constructed by any one of the neural networks selected from BP neural network, generalized regression neural network and probabilistic neural network. In one embodiment, the surge point prediction model in the embodiment of the present invention can be constructed by using a BP neural network.

[0095] The compressor performance prediction and adjustment method of the present invention is based on a neural network model, which can realize fast and accurate prediction of compressor performance when the compressor working conditions are different and the target operating point is far away from the existing performance curve data point, and provide guidance for compressor adjustment.

[0096] Step 400: Input the performance curve parameters to be measured and the working conditions into a congestion point prediction model to obtain a congestion point prediction result output by the congestion point prediction model.

[0097] Among them, the congestion point prediction model is trained based on multiple sample performance curves and the congestion point labels corresponding to each sample performance curve; each sample performance curve is constructed based on multiple sample performance curve parameters, sample working conditions and multiple model prediction results input by the trained performance prediction model.

[0098] The blocking point prediction model is obtained by training based on multiple sample performance curves and blocking point labels corresponding to each sample performance curve. Each sample performance curve is constructed based on multiple sample performance curve parameters, sample working conditions and multiple model prediction results input by the trained performance prediction model. Each model prediction result is predicted by the performance prediction model based on each sample performance curve parameter. In the embodiment of the present invention, a sample performance curve is constructed based on multiple sample performance curve parameters under the same sample working condition and the model prediction result of each sample performance curve parameter. For example, when the compressor speed is 2000r / min, a first sample performance curve is constructed based on data composed of several sample performance curve parameters and model prediction results. The rightmost point of the first sample performance curve is the blocking point, that is, the blocking point label corresponding to the first sample performance curve. Different sample performance curves and corresponding blocking point labels can be obtained by changing the sample working conditions. Therefore, the blocking point prediction model is trained based on different sample performance curves and corresponding blocking point labels, and can predict the blocking points corresponding to different performance curve parameters and working conditions to be tested. The blocking point prediction model can be constructed by any one of BP neural network, generalized regression neural network and probabilistic neural network. In one embodiment, the blocking point prediction model in the embodiment of the present invention can be constructed by BP neural network.

[0099] In the embodiment of the present invention, the surge point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set surge line. The blocking point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set blocking line. The performance prediction curve is composed of multiple performance curve parameters to be tested under the same working conditions and the performance prediction results of each performance curve parameter to be tested input into the performance prediction model. That is, multiple (performance curve parameters to be tested, performance prediction results) data pairs constitute the performance prediction curve.

[0100] The compressor performance prediction and adjustment method of the present invention is based on a neural network model, which can realize fast and accurate prediction of compressor performance when the compressor working conditions are different and the target operating point is far away from the existing performance curve data point, and provide guidance for compressor adjustment.

[0101] In the embodiments of the present invention, a sample performance curve is constructed based on a plurality of sample performance curve parameters, sample operating conditions, and a plurality of model prediction results input during the training of the performance prediction model. Then, a surge point prediction model and a choke point prediction model trained based on the plurality of sample performance curves and the surge point label and choke point label corresponding to each sample performance curve are used to predict the surge point and choke point, and the compressor performance is predicted based on the performance prediction model. Therefore, in the embodiments of the present invention, a neural network model is built based on the sample performance curve parameters and sample operating conditions, which can predict the performance curve including the surge point and choke point under different operating conditions. Compared with the existing methods, it is simple and easy to implement, and can quickly and accurately predict the compressor performance.

[0102] In other aspects of the embodiments of the present invention, the surge point prediction model is trained through the following steps:

[0103] Step 11: Obtain a plurality of sample performance curve parameters and a plurality of sample operating conditions.

[0104] The embodiments of the present invention are described by taking a gas storage compressor as an example. The gas storage compressor is used to compress the upstream natural gas and inject it into the gas storage during the gas injection stage. The incoming flow temperature and pressure at the compressor inlet change with the upstream. A certain gas storage compressor has a total of 20 sets of factory performance curves, where the inlet pressure P in ∈[41, 51, 61, 71, 81] (bar), the inlet temperature T in ∈[0, 10, 20, 30] (°C), and the performance curve parameters include the volume flow rate q v , compression ratio ε, polytropic efficiency η, power N, rotational speed n, etc. The embodiments of the present invention are described by taking the working condition point of the flow-compression ratio of this compressor as an example.

[0105] Please refer to Figure 2 , Figure 2 which shows the training process of the surge point prediction model and the compressor performance prediction of the choke point in the embodiments of the present invention. The electronic device obtains a plurality of sample performance curve parameters and a plurality of sample operating conditions. The plurality of sample performance curve parameters include the volume flow rate q v , polytropic efficiency η, power N, etc. The plurality of sample operating conditions include the rotational speed of the compressor, the inlet pressure of the compressor, the inlet temperature of the compressor, and the inlet guide vane angle of the compressor, etc. Before using each sample performance curve parameter and each sample operating condition for model training data, preprocessing of the compressor performance curve data is required. For example, there are l groups of compressor performance curves, the inlet temperature of each group of performance curves is T i , the inlet pressure is P i , 1 < i < l, and each group of performance curves is divided into m i performance curves according to different rotational speeds or inlet guide vane angles. Each performance

[0106] The curve corresponds to the speed or inlet guide vane angle Z i,j , 1 <j<m i , each curve contains n i,j data points, the corresponding horizontal and vertical coordinate values ​​of the data points in the two-dimensional coordinate system are X i,j,k , Y i,j,k , 1 <k<n i,j , X represents flow rate (volume flow rate or mass flow rate), Y represents other performance parameters (compression ratio, power, variable efficiency, energy head, exhaust temperature, etc.). Each line contains the leftmost surge point and the rightmost blockage point.

[0107] Normalize the variable data of T (inlet temperature), P (inlet pressure), Z (speed or inlet guide vane angle), X (volume flow or mass flow), and Y (compression ratio) to the interval [c, d] (d>c). Take variable X as an example:

[0108]

[0109] Among them, X i,j,k,norm represents the sample performance curve parameter after normalization, c and d are the lower and upper thresholds respectively. i,j,k represents the sample performance curve parameter that has not been normalized, X min Represents the minimum value of all sample performance curve parameters, X max Indicates the maximum value among all sample performance curve parameters.

[0110] The normalized data is divided into three parts: training data, validation data, and test data.

[0111] Specifically, the inlet pressure P in , inlet temperature T in , the data of compressor speed Z, flow rate X, and performance parameter Y are normalized to the interval [-1,1] to obtain the normalized inlet pressure P in，norm , inlet temperature T in，norm , compressor speed Z norm 、Flow rate X norm 、Performance parameters Y norm . Take the compression ratio ε variable as an example:

[0112]

[0113] Among them, ε norm represents the normalized compression ratio. ε represents the unnormalized compression ratio, min Represents the minimum value of all compression ratios, ε maxIndicates the maximum value of all compression ratios. The normalized data is divided into three parts: 70% training data, 15% validation data, and 15% test data.

[0114] Step 12: training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters.

[0115] Specifically, the training of the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions and the performance label corresponding to each of the sample performance curve parameters includes:

[0116] Repeat the following steps until the set number of iterations is reached:

[0117] Inputting a target sample performance curve parameter and a target sample working condition into a performance prediction model to obtain a model output result of the performance prediction model; the target sample performance curve parameter is any one of the multiple sample performance curve parameters, and the target sample working condition is any one of the multiple sample working conditions;

[0118] Calculate the loss function based on the target sample performance curve parameters, the real performance labels corresponding to the target sample working conditions, and the model output results;

[0119] Model parameters of the loss function are adjusted based on the loss function.

[0120] For example, the embodiment of the present invention trains the performance prediction model based on multiple volume flow rates, multiple compressor speeds, and the compression ratio (i.e., performance label) corresponding to each volume flow rate. The performance prediction model of the embodiment of the present invention uses a BP neural network. Please refer to Figure 3 The BP neural network of the performance prediction model includes an input layer, a hidden layer (or implicit layer), and an output layer. The number of nodes in the input layer can be set to 4 (for example, including the inlet pressure P in，norm , inlet temperature T in，norm , compressor speed Z norm 、Flow rate X norm 、Performance parameters Y norm ). For example, in the embodiment of the present invention, the input layer can be volume flow, compressor speed, inlet pressure and inlet temperature. The number of hidden layers is multiple, the number of nodes in each layer is multiple, and the number of nodes in the output layer is 1. The output layer variable can be flow rate X norm Or performance parameter Y norm Or compressor speed Z norm, and the remaining variables are input layer variables. In the embodiment of the present invention, the output layer variable may be a compression ratio. In other embodiments, the number of input layer nodes may be set to 4 (e.g., inlet pressure, inlet temperature, compressor speed, compression ratio), the number of hidden layers is multiple, the number of nodes in each layer is multiple, and the number of output layer nodes is 1. The output layer variable may be a volume flow rate.

[0121] Except for the input layer, the neuron nodes in other layers have their thresholds b. The input value received by each neuron node has its weight ω. The value of the neuron is obtained by adding the threshold to all the weighted input values. After being converted by its own activation function, it can be output and passed to the next layer. The output value y of the neuron node with M inputs is calculated as follows:

[0122]

[0123] Where f is the activation function, 1 <m<M。

[0124] The output value ε of the output layer neuron is calculated as follows:

[0125]

[0126] in is the weight from the tth node in the input layer to the sth node in the hidden layer. is the weight from the tth node in the hidden layer to the sth node in the output layer. is the threshold of the tth node in the hidden layer. is the threshold of the tth node in the output layer.

[0127] By denormalizing the output value of the output layer and comparing it with the true value of the performance label, the root mean square error (RMSE) of all sample output values ​​is used as the error function, and the training algorithm is used to backpropagate from the last layer to the previous layer, and the weight ω and threshold b of each layer of neurons are iteratively calculated. When the error function is minimized, the training is completed, and the corresponding BP neural network model, that is, the weight ω and threshold b of each neuron, is obtained after verification by the test data.

[0128] Please refer to Figure 4 , where the calculation process of back propagation is as follows:

[0129] E is the calculated total error, o=ω1*x1+ω2*x2+ω3*x3+b, y=f(o), taking ω1 as an example, the partial derivative of the total error E with respect to ω1 (chain rule) can tell the degree of influence of ω1 on the error:

[0130]

[0131] According to the obtained partial derivative update, ω′1 is obtained, and λ is the learning rate, which takes a certain fixed value.

[0132]

[0133] All weights can be updated according to the total error through the back propagation algorithm for the next round of iteration. Back propagation uses the chain rule to calculate the partial derivatives of each parameter from the output layer to the input layer in sequence. The partial derivatives calculated in the middle can be reused without repeated calculation, which is more efficient than forward propagation.

[0134] Step 13: predict the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results.

[0135] Step 14: construct a single sample performance curve based on all sample performance curve parameters under each sample working condition and the model prediction result corresponding to each sample performance curve parameter to obtain multiple sample performance curves.

[0136] The embodiment of the present invention can use the trained performance prediction model to predict P in ∈[41,51,61,71,81](bar), inlet temperature T in ∈[0,10,20,30](℃) performance curves at different compressor speeds n. Please refer to Figure 5 Taking the performance curves at different compressor speeds n as an example, the embodiment of the present invention first predicts multiple volume flow rates when the compressor speed is 2000r / min through the performance prediction model to obtain multiple compressor prediction results. Thus, a sample performance curve (i.e. Figure 5 The predicted performance curve in red). By changing the compressor speed, for example, firstly predicting multiple volume flow rates when the compressor speed is 2100 r / min through the performance prediction model, multiple compressor prediction results when the compressor speed is 2100 r / min are obtained, and then constructing a sample performance curve at the compressor speed of 2100 r / min based on the multiple volume flow rates and the multiple compressor prediction results. Thus, the embodiment of the present invention can obtain multiple sample performance curves (i.e., Figure 5 There are multiple prediction performance curves from Z1 to Z5 in ).

[0137] Step 15: Determine a plurality of target surge points where the plurality of sample performance curves intersect a set surge line as the surge point labels.

[0138] The embodiment of the present invention uses a trained performance prediction model to predict performance curves at different speeds or inlet guide vane angles under different inlet conditions. For example, the embodiment of the present invention uses a trained performance prediction model to predict performance curves at different compressor speeds. However, the location of the surge point or blockage point of each curve has not yet been determined. Therefore, model prediction is performed using l groups of sample performance curve parameters (each group of sample performance curve parameters is a parameter at the same compressor speed), the speed is encrypted, and the intersection of the l performance curves and the surge line is the supplementary surge point, for example Figure 5 The intersection of the multiple predicted performance curves from Z1 to Z5 and the surge line is the supplemented surge point. The supplemented surge points in the l group of performance curves are also normalized, and the data of the two are divided into three parts: training data, verification data, and test data.

[0139] For example, the embodiment of the present invention uses the trained performance prediction model to predict P in ∈[41,51,61,71,81](bar), inlet temperature T in ∈[0,10,20,30](℃) Performance curves at different speeds n, at the maximum speed n max With minimum speed n min The performance curve clusters corresponding to each speed are encrypted, and the intersection of the encrypted performance curve and the surge line is the supplemented surge point. The supplemented surge point data is also normalized, and the two data are divided into 70% training data, 15% verification data, and 15% test data.

[0140] Step 16: training the surge point prediction model based on the sample performance curve parameters, the model prediction results and the surge point labels corresponding to the sample performance curves on each of the sample performance curves.

[0141] The embodiment of the present invention constructs and trains the surge point data obtained in step 15 using a BP neural network. The neural network includes an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is 3. The input layer data are sample performance curve parameters and sample working conditions, for example, the input layer data are inlet pressure, inlet temperature, compressor speed, and volume flow rate. There are multiple hidden layers, multiple nodes in each layer, and 2 output layer nodes. The output layer variable is the surge point. The remaining parameter settings and calculation process of the performance prediction model constructed based on the BP neural network are the same as step 12, and will not be repeated here.

[0142] The activation function of the hidden layer neurons is selected as the tansig function, so that the network has nonlinear fitting capabilities. The output layer neuron activation function uses an identity mapping to directly output the neuron value. The output value of the output layer is denormalized and compared with the true value in the verification data. The root mean square error of all sample output values ​​is used as the error function. The gradient descent training algorithm is used to backpropagate from the last layer to the previous layer, and the weight ω and threshold b of each layer of neurons are iteratively calculated. When the error function is minimized, the training is completed, and the corresponding BP neural network model, that is, the weight ω and threshold b of each neuron, is obtained after verification by the test data.

[0143] Therefore, the embodiment of the present invention can predict the performance curve including the surge point at different speeds or inlet guide vane angles under different inlet conditions based on the trained performance prediction model and surge point prediction model.

[0144] In other aspects of the embodiments of the present invention, the congestion point prediction model is trained by the following steps:

[0145] Step 21: Obtain multiple sample performance curve parameters and multiple sample working conditions.

[0146] In the training of the congestion point prediction model, the steps of obtaining multiple sample performance curve parameters and multiple sample working conditions, and the steps of data preprocessing of multiple sample performance curve parameters and multiple sample working conditions are the same as step 11. Please refer to the relevant description of step 11 and will not be repeated here.

[0147] Step 22: training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters.

[0148] Specifically, the training of the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions and the performance label corresponding to each of the sample performance curve parameters includes:

[0149] Repeat the following steps until the set number of iterations is reached:

[0150] Inputting a target sample performance curve parameter and a target sample working condition into a performance prediction model to obtain a model output result of the performance prediction model; the target sample performance curve parameter is any one of the multiple sample performance curve parameters, and the target sample working condition is any one of the multiple sample working conditions;

[0151] Calculate the loss function based on the target sample performance curve parameters, the real performance labels corresponding to the target sample working conditions, and the model output results;

[0152] Model parameters of the loss function are adjusted based on the loss function.

[0153] Similarly, in the training of the congestion point prediction model, the training steps of the performance prediction model are the same as step 12. Please refer to the relevant description of step 12, which will not be repeated here.

[0154] Step 23: predict the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results.

[0155] Step 24: construct a single sample performance curve based on all sample performance curve parameters under each sample working condition and the model prediction result corresponding to each sample performance curve parameter to obtain multiple sample performance curves.

[0156] Similarly, in the training of the congestion point prediction model, the steps for generating multiple sample performance curves are the same as steps 13 and 14. Please refer to the relevant descriptions of steps 13 and 14, which will not be repeated here.

[0157] Step 25: Determine a plurality of target blocking points where the plurality of sample performance curves intersect with a set blocking line as the blocking point labels.

[0158] The embodiment of the present invention uses a trained performance prediction model to predict performance curves at different speeds or inlet guide vane angles under different inlet conditions. For example, the embodiment of the present invention uses a trained performance prediction model to predict performance curves at different compressor speeds. However, the location of the surge point or blockage point of each curve has not yet been determined. Therefore, model prediction is performed using l groups of sample performance curve parameters (each group of sample performance curve parameters is a parameter at the same compressor speed), the speed is encrypted, and the intersection of the l performance curves and the blockage line is the supplementary blockage point, for example Figure 5 The intersection of the multiple predicted performance curves from Z1 to Z5 and the blocking line is the supplemented blocking point. The supplemented blocking points in the l performance curves are also normalized, and the data of the two are divided into three parts: training data, verification data, and test data.

[0159] For example, the embodiment of the present invention uses a trained performance prediction model to predict P in ∈[41,51,61,71,81](bar), inlet temperature T in ∈[0,10,20,30](℃) Performance curves at different speeds n, at the maximum speed n max With minimum speed n minThe performance curve clusters corresponding to each speed are encrypted, and the intersection of the encrypted performance curve and the congestion line is the supplemented congestion point. The supplemented congestion point data is also normalized, and the two data are divided into 70% training data, 15% verification data, and 15% test data.

[0160] Step 26: training the congestion point prediction model based on the sample performance curve parameters, the model prediction results and the congestion point labels corresponding to the sample performance curves on each of the sample performance curves.

[0161] The embodiment of the present invention constructs and trains the BP neural network for the blockage point data obtained in step 25. The neural network includes an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is 3. The input layer data is sample performance curve parameters and sample working conditions, for example, the input layer data is inlet pressure, inlet temperature, compressor speed, and volume flow rate. There are multiple hidden layers, multiple nodes in each layer, and 2 output layer nodes. The output layer variable is the blockage point. The remaining parameter settings and calculation process of the performance prediction model constructed based on the BP neural network are the same as step 22, and will not be repeated here.

[0162] The activation function of the hidden layer neurons is selected as the tansig function, so that the network has nonlinear fitting capabilities. The output layer neuron activation function uses an identity mapping to directly output the neuron value. The output value of the output layer is denormalized and compared with the true value in the verification data. The root mean square error of all sample output values ​​is used as the error function. The gradient descent training algorithm is used to backpropagate from the last layer to the previous layer, and the weight ω and threshold b of each layer of neurons are iteratively calculated. When the error function is minimized, the training is completed, and the corresponding BP neural network model, that is, the weight ω and threshold b of each neuron, is obtained after verification by the test data.

[0163] Therefore, the embodiment of the present invention can predict the performance curve including the clogging point under different inlet conditions and different rotation speeds or inlet guide vane angles based on the trained performance prediction model and the clogging point prediction model.

[0164] In summary, the embodiments of the present invention can predict the performance curves including surge points and blockage points at different speeds or inlet guide vane angles under different inlet conditions based on the trained performance prediction model, surge point prediction model and blockage point prediction model. The compressor performance prediction and adjustment method based on the BP neural network in the embodiments of the present invention can build a neural network based on performance curve data, and predict the performance curves including surge points and blockage points at different speeds or inlet guide vane angles under different inlet conditions, which has an important guiding role in the adjustment of the compressor. Compared with the existing methods, it is simple and easy to implement, and improves accuracy and efficiency.

[0165] Device Embodiment

[0166] Please refer to Figure 6 On the other hand, an embodiment of the present invention further provides a compressor performance prediction device, comprising:

[0167] An acquisition module 601 is used to acquire the performance curve parameters and working conditions to be tested of the compressor;

[0168] A first prediction module 602, used for inputting the performance curve parameters to be measured and the working conditions into a performance prediction model to obtain a performance prediction result output by the performance prediction model;

[0169] A second prediction module 603, used for inputting the performance curve parameter to be measured and the working condition into a surge point prediction model to obtain a surge point prediction result output by the surge point prediction model;

[0170] A third prediction module 604 is used to input the performance curve parameter to be measured and the working condition into a congestion point prediction model to obtain a congestion point prediction result output by the congestion point prediction model;

[0171] Among them, the surge point prediction model is trained based on multiple sample performance curves and the surge point labels corresponding to each of the sample performance curves; the congestion point prediction model is trained based on multiple sample performance curves and the congestion point labels corresponding to each of the sample performance curves; each of the sample performance curves is constructed based on multiple sample performance curve parameters, sample working conditions and multiple model prediction results input by the trained performance prediction model; the surge point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set surge line; the congestion point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set congestion line.

[0172] Optionally, the surge point prediction model is trained by the following steps:

[0173] Obtaining multiple sample performance curve parameters and multiple sample working conditions;

[0174] Training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters;

[0175] Predicting the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results;

[0176] Constructing a single sample performance curve based on all sample performance curve parameters under each of the sample working conditions and the model prediction results corresponding to each sample performance curve parameter to obtain multiple sample performance curves;

[0177] determining a plurality of target surge points where the plurality of sample performance curves intersect a set surge line as the surge point labels;

[0178] The surge point prediction model is trained based on the sample performance curve parameters, the model prediction results and the surge point labels corresponding to the sample performance curves on each of the sample performance curves.

[0179] Optionally, the congestion point prediction model is trained by the following steps:

[0180] Obtaining multiple sample performance curve parameters and multiple sample working conditions;

[0181] Training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters;

[0182] Predicting the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results;

[0183] Constructing a single sample performance curve based on all sample performance curve parameters under each of the sample working conditions and the model prediction results corresponding to each sample performance curve parameter to obtain multiple sample performance curves;

[0184] determining a plurality of target blocking points where the plurality of sample performance curves intersect a set blocking line as the blocking point labels;

[0185] The congestion point prediction model is trained based on the sample performance curve parameters on each of the sample performance curves, the model prediction results and the congestion point labels corresponding to the sample performance curves.

[0186] Optionally, the training of the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions and the performance label corresponding to each of the sample performance curve parameters includes:

[0187] Repeat the following steps until the set number of iterations is reached:

[0188] Inputting a target sample performance curve parameter and a target sample working condition into a performance prediction model to obtain a model output result of the performance prediction model; the target sample performance curve parameter is any one of the multiple sample performance curve parameters, and the target sample working condition is any one of the multiple sample working conditions;

[0189] Calculate the loss function based on the target sample performance curve parameters, the real performance labels corresponding to the target sample working conditions, and the model output results;

[0190] Model parameters of the loss function are adjusted based on the loss function.

[0191] Optionally, the performance prediction model is constructed using any one of a BP neural network, a generalized regression neural network and a probabilistic neural network.

[0192] Optionally, the surge point prediction model is constructed by using any one of a BP neural network, a generalized regression neural network and a probabilistic neural network.

[0193] Optionally, the congestion point prediction model is constructed by using any one of BP neural network, generalized regression neural network and probabilistic neural network.

[0194] Optionally, the performance curve parameter to be measured includes at least one of pressure, polynomial efficiency, flow rate, power and compression ratio of the compressor.

[0195] Optionally, the operating conditions include at least one of a rotation speed of the compressor, an inlet pressure of the compressor, an inlet temperature of the compressor, and an inlet guide vane angle of the compressor.

[0196] The compressor performance prediction device includes a processor and a memory. The acquisition module 601, the first prediction module 602, the second prediction module 603 and the third prediction module 604 are all stored in the memory as program units, and the processor executes the program units stored in the memory to implement corresponding functions.

[0197] The processor includes a kernel, which calls the corresponding program unit from the memory. There can be one or more kernels.

[0198] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0199] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7As shown, the electronic device may include: a processor (processor) 710, a communication interface (Communications Interface) 720, a memory (memory) 730 and a communication bus 740, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other through the communication bus 740. The processor 710 can call the logic instructions in the memory 730 to execute the compressor performance prediction method, which includes: obtaining the performance curve parameters and working conditions to be tested of the compressor; inputting the performance curve parameters and the working conditions to be tested into a performance prediction model to obtain the performance prediction results output by the performance prediction model; inputting the performance curve parameters and the working conditions to be tested into a surge point prediction model to obtain the surge point prediction results output by the surge point prediction model; inputting the performance curve parameters and the working conditions to be tested into a congestion point prediction model to obtain the congestion point prediction results output by the congestion point prediction model; wherein the surge The surge point prediction model is obtained by training based on multiple sample performance curves and surge point labels corresponding to each of the sample performance curves; the congestion point prediction model is obtained by training based on multiple sample performance curves and congestion point labels corresponding to each of the sample performance curves; each of the sample performance curves is constructed based on multiple sample performance curve parameters, sample working conditions and multiple model prediction results input by the trained performance prediction model; the surge point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set surge line; the congestion point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set congestion line.

[0200] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0201] On the other hand, the present invention further provides a computer program product, which includes a computer program. The computer program can be stored on a machine-readable storage medium. When the computer program is executed by a processor, the computer can execute a compressor performance prediction method, which includes: obtaining the performance curve parameters and working conditions to be tested of the compressor; inputting the performance curve parameters and the working conditions to be tested into a performance prediction model to obtain a performance prediction result output by the performance prediction model; inputting the performance curve parameters and the working conditions to be tested into a surge point prediction model to obtain a surge point prediction result output by the surge point prediction model; inputting the performance curve parameters and the working conditions to be tested into a blockage point prediction model to obtain The blocking point prediction result output by the blocking point prediction model; wherein, the surge point prediction model is obtained by training based on multiple sample performance curves and surge point labels corresponding to each of the sample performance curves; the blocking point prediction model is obtained by training based on multiple sample performance curves and blocking point labels corresponding to each of the sample performance curves; each of the sample performance curves is constructed based on multiple sample performance curve parameters, sample operating conditions and multiple model prediction results input by the trained performance prediction model; the surge point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set surge line; the blocking point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set blocking line.

[0202] On the other hand, the present invention also provides a machine-readable storage medium having a computer program stored thereon, which is implemented when the processor executes the compressor performance prediction method, the method comprising: obtaining the performance curve parameters and working conditions to be tested of the compressor; inputting the performance curve parameters and the working conditions to be tested into a performance prediction model to obtain the performance prediction results output by the performance prediction model; inputting the performance curve parameters and the working conditions to be tested into a surge point prediction model to obtain the surge point prediction results output by the surge point prediction model; inputting the performance curve parameters and the working conditions to a congestion point prediction model to obtain the congestion point prediction results output by the congestion point prediction model Point prediction results; wherein, the surge point prediction model is obtained by training based on multiple sample performance curves and surge point labels corresponding to each of the sample performance curves; the congestion point prediction model is obtained by training based on multiple sample performance curves and congestion point labels corresponding to each of the sample performance curves; each of the sample performance curves is constructed based on multiple sample performance curve parameters, sample operating conditions and multiple model prediction results input by the trained performance prediction model; the surge point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set surge line; the congestion point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set congestion line.

[0203] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0204] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0205] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A compressor performance prediction method, characterized in that: include: Obtaining the performance curve parameters and working conditions to be tested of the compressor; Inputting the performance curve parameters to be measured and the working conditions into a performance prediction model to obtain a performance prediction result output by the performance prediction model; Inputting the performance curve parameter to be measured and the operating conditions into a surge point prediction model to obtain a surge point prediction result output by the surge point prediction model; Inputting the performance curve parameters to be measured and the working conditions into a congestion point prediction model to obtain a congestion point prediction result output by the congestion point prediction model; Wherein, the surge point prediction model is obtained by training based on multiple sample performance curves and the surge point labels corresponding to each of the sample performance curves; the congestion point prediction model is obtained by training based on multiple sample performance curves and the congestion point labels corresponding to each of the sample performance curves; each of the sample performance curves is constructed based on multiple sample performance curve parameters, sample working conditions and multiple model prediction results inputted from the trained performance prediction model; the surge point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set surge line; The congestion point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set congestion line.

2. The compressor performance prediction method according to claim 1, characterized in that: The surge point prediction model is trained by the following steps: Obtaining multiple sample performance curve parameters and multiple sample working conditions; Training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters; Predicting the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results; Constructing a single sample performance curve based on all sample performance curve parameters under each of the sample working conditions and the model prediction results corresponding to each sample performance curve parameter to obtain multiple sample performance curves; determining a plurality of target surge points where the plurality of sample performance curves intersect a set surge line as the surge point labels; The surge point prediction model is trained based on the sample performance curve parameters, the model prediction results and the surge point labels corresponding to the sample performance curves on each of the sample performance curves.

3. The compressor performance prediction method according to claim 1, characterized in that: The congestion point prediction model is trained by the following steps: Obtaining multiple sample performance curve parameters and multiple sample working conditions; Training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters; Predicting the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results; Constructing a single sample performance curve based on all sample performance curve parameters under each of the sample working conditions and the model prediction results corresponding to each sample performance curve parameter to obtain multiple sample performance curves; determining a plurality of target blocking points where the plurality of sample performance curves intersect a set blocking line as the blocking point labels; The congestion point prediction model is trained based on the sample performance curve parameters on each of the sample performance curves, the model prediction results and the congestion point labels corresponding to the sample performance curves.

4. The compressor performance prediction method according to claim 2 or 3, characterized in that: The training of the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions and the performance label corresponding to each of the sample performance curve parameters includes: Repeat the following steps until a set number of iterations is reached: input the target sample performance curve parameter and the target sample working condition into the performance prediction model to obtain a model output result of the performance prediction model; the target sample performance curve parameter is any one of the multiple sample performance curve parameters, and the target sample working condition is any one of the multiple sample working conditions; Calculate the loss function based on the target sample performance curve parameters, the real performance labels corresponding to the target sample working conditions, and the model output results; Model parameters of the loss function are adjusted based on the loss function.

5. The compressor performance prediction method according to claim 1, characterized in that: The performance prediction model is constructed by using any one of BP neural network, generalized regression neural network and probabilistic neural network.

6. The compressor performance prediction method according to claim 1, characterized in that: The surge point prediction model is constructed by using any one of BP neural network, generalized regression neural network and probabilistic neural network.

7. The compressor performance prediction method according to claim 1, characterized in that: The congestion point prediction model is constructed by using any one of BP neural network, generalized regression neural network and probabilistic neural network.

8. The compressor performance prediction method according to claim 1, characterized in that: The performance curve parameter to be measured includes at least one of the pressure, polytropic efficiency, flow rate, power and compression ratio of the compressor.

9. The compressor performance prediction method according to claim 1, characterized in that: The working condition includes at least one of a rotation speed of the compressor, an inlet pressure of the compressor, an inlet temperature of the compressor, and an inlet guide vane angle of the compressor.

10. A compressor performance prediction device, characterized in that: include: An acquisition module, used to acquire the performance curve parameters and working conditions to be tested of the compressor; A first prediction module, used for inputting the performance curve parameters to be measured and the working conditions into a performance prediction model to obtain a performance prediction result output by the performance prediction model; A second prediction module, used for inputting the performance curve parameter to be measured and the working condition into a surge point prediction model to obtain a surge point prediction result output by the surge point prediction model; A third prediction module, used for inputting the performance curve parameter to be measured and the working condition into a congestion point prediction model to obtain a congestion point prediction result output by the congestion point prediction model; Wherein, the surge point prediction model is obtained by training based on multiple sample performance curves and the surge point labels corresponding to each of the sample performance curves; the congestion point prediction model is obtained by training based on multiple sample performance curves and the congestion point labels corresponding to each of the sample performance curves; each of the sample performance curves is constructed based on multiple sample performance curve parameters, sample working conditions and multiple model prediction results inputted from the trained performance prediction model; the surge point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set surge line; The congestion point prediction result is the intersection of the performance prediction curve where the performance prediction result is located and the set congestion line.

11. The compressor performance prediction device according to claim 10, characterized in that: The surge point prediction model is trained by the following steps: Obtaining multiple sample performance curve parameters and multiple sample working conditions; Training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters; Predicting the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results; Constructing a single sample performance curve based on all sample performance curve parameters under each of the sample working conditions and the model prediction results corresponding to each sample performance curve parameter to obtain multiple sample performance curves; determining a plurality of target surge points where the plurality of sample performance curves intersect a set surge line as the surge point labels; The surge point prediction model is trained based on the sample performance curve parameters, the model prediction results and the surge point labels corresponding to the sample performance curves on each of the sample performance curves.

12. The compressor performance prediction device according to claim 10, characterized in that: The congestion point prediction model is trained by the following steps: Obtaining multiple sample performance curve parameters and multiple sample working conditions; Training the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions, and the performance label corresponding to each of the sample performance curve parameters; Predicting the performance of each of the sample performance curve parameters under each of the sample working conditions based on the trained performance prediction model to obtain multiple model prediction results; Constructing a single sample performance curve based on all sample performance curve parameters under each of the sample working conditions and the model prediction results corresponding to each sample performance curve parameter to obtain multiple sample performance curves; determining a plurality of target blocking points where the plurality of sample performance curves intersect a set blocking line as the blocking point labels; The congestion point prediction model is trained based on the sample performance curve parameters on each of the sample performance curves, the model prediction results and the congestion point labels corresponding to the sample performance curves.

13. The compressor performance prediction device according to claim 11 or 12, characterized in that: The training of the performance prediction model based on the multiple sample performance curve parameters, the multiple sample working conditions and the performance label corresponding to each of the sample performance curve parameters includes: Repeat the following steps until the set number of iterations is reached: Inputting a target sample performance curve parameter and a target sample working condition into a performance prediction model to obtain a model output result of the performance prediction model; the target sample performance curve parameter is any one of the multiple sample performance curve parameters, and the target sample working condition is any one of the multiple sample working conditions; Calculate the loss function based on the target sample performance curve parameters, the real performance labels corresponding to the target sample working conditions, and the model output results; Model parameters of the loss function are adjusted based on the loss function.

14. The compressor performance prediction device according to claim 10, characterized in that: The performance prediction model is constructed by using any one of BP neural network, generalized regression neural network and probabilistic neural network.

15. The compressor performance prediction device according to claim 10, characterized in that: The surge point prediction model is constructed by using any one of BP neural network, generalized regression neural network and probabilistic neural network.

16. The compressor performance prediction device according to claim 10, characterized in that: The congestion point prediction model is constructed by using any one of BP neural network, generalized regression neural network and probabilistic neural network.

17. The compressor performance prediction device according to claim 10, characterized in that: The performance curve parameter to be measured includes at least one of the pressure, polytropic efficiency, flow rate, power and compression ratio of the compressor.

18. The compressor performance prediction device according to claim 10, characterized in that: The working condition includes at least one of a rotation speed of the compressor, an inlet pressure of the compressor, an inlet temperature of the compressor, and an inlet guide vane angle of the compressor.

19. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the compressor performance prediction method according to any one of claims 1 to 9 is implemented.

20. A machine-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the compressor performance prediction method according to any one of claims 1 to 9 is implemented.

21. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the compressor performance prediction method according to any one of claims 1 to 9 is implemented.

Images