Compressor temperature array abnormality prediction method based on twin iterative deduction

Abstract

The present invention discloses a compressor temperature array anomaly prediction method based on twin iterative deduction, including: acquiring compressor data; constructing a single temperature parameter twin prediction model; performing twin iterative deduction of all temperature parameters; realizing long-term trend prediction and accurate anomaly warning of compressor multi-dimensional temperature array data, overcoming the shortcomings of traditional methods in the collaborative prediction of multiple temperature parameters, and ultimately achieving significant technical effects of improving compressor operation reliability and ensuring safe and efficient operation of industrial production.

Description

Technical Field

[0001] The present invention relates to the technical field of compressor operating status monitoring, and in particular to a compressor temperature array anomaly prediction method based on twin iterative deduction. Background Art

[0002] Compressors are widely used in petrochemicals, refrigeration systems and various industrial processes. Their operating stability is directly related to the continuity and safety of the overall production process. Temperature parameters are one of the important monitoring indicators of the compressor's operating conditions. If the operating temperature is abnormal, it may not only lead to failure of the compressor's internal lubrication, thermal deformation of bearings and rotor materials, or degradation of sealing performance, but may also cause phase changes in the internal gas medium (such as condensation or freezing). In severe cases, it may cause reduced compressor operating efficiency, increased vibration, and even serious mechanical failure.

[0003] At present, compressor temperature monitoring mostly uses single-type sensors such as thermal resistors and thermocouples for data collection. However, these single components are easily affected by electromagnetic interference, working fluid contamination or mechanical vibration, and there are problems with insufficient data reliability and redundancy. At the same time, traditional monitoring methods usually rely on discrete sensors to collect data, lacking a comprehensive analysis of the evolution trend of the overall thermal state of the compressor. Especially under the operating conditions of multi-stage compressors, variable loads or variable speed systems, it is impossible to effectively achieve early and accurate warning of abnormal trends in temperature parameters, and it is difficult to meet the needs of modern industrial production for intelligent management of compressor operating status.

[0004] CN115717590A discloses a method for detecting abnormal conditions of compressors by establishing a reconstruction model using PLC data and feature engineering. Although this method does not require the addition of new sensors, its feature engineering relies on specific domain knowledge and lacks the prediction of multi-step evolution trends of temperature parameters, resulting in limited early warning accuracy and advance time.

[0005] CN116181635A discloses a method for detecting anomalies in the exhaust temperature data of subway air-conditioning compressors using an isolation forest algorithm. Although this method has the advantage of unsupervised detection, it is still limited to the static feature analysis of a single temperature parameter, ignoring the synergistic change characteristics between the overall temperature array of the compressor, and cannot effectively capture the associated abnormal trends of multi-dimensional temperature parameters under dynamic working conditions, affecting the accuracy and stability of fault warning.

[0006] Therefore, the existing compressor temperature anomaly prediction method has the problems of single monitoring method, insufficient data redundancy, and poor multi-dimensional temperature data correlation analysis capabilities. Our invention uses twin iterative deduction and recurrent neural network technology to solve the long-term accurate prediction and early warning of abnormal trends of multi-dimensional temperature array data of compressors under complex dynamic working conditions. Summary of the Invention

[0007] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract of the specification and the title of the invention of this application to avoid blurring the purpose of this section, the abstract of the specification and the title of the invention, and such simplifications or omissions cannot be used to limit the scope of the invention.

[0008] In view of the above existing problems, the present invention is proposed.

[0009] To solve the above technical problems, the present invention provides the following technical solutions: collecting the operating parameters of the compressor and the first temperature parameter array in real time to form a first state parameter vector at the current time t;

[0010] For each target temperature parameter in the first temperature parameter array, using the operating condition parameter at time t and the first temperature parameter array as input, a recurrent neural network is trained to obtain a single temperature parameter twin prediction model for predicting the target temperature at time t+1;

[0011] Repeat the previous step to establish corresponding single temperature parameter twin prediction models for all temperature parameters to form a temperature parameter model set;

[0012] Taking the first state parameter vector at the current time t as input, the temperature parameter model set is sequentially called to obtain a second temperature parameter array at the time t+1, and the second temperature parameter array and the current operating condition parameters are combined to form a second state parameter vector, which is iterated N times to obtain a multi-step deduction result;

[0013] Using a historical state matrix containing historical multi-time state parameter vectors, a conversion matrix is ​​established through principal component analysis, and the multi-step deduction results are subjected to dimensionality reduction and inverse transformation to obtain a deviation baseline of each temperature parameter;

[0014] The deviation ratio of each temperature parameter in the multi-step deduction result relative to its deviation baseline is calculated, and when the deviation ratio of any temperature parameter exceeds the warning threshold, a compressor temperature array abnormality warning signal is output.

[0015] As a preferred solution of the compressor temperature array abnormality prediction method based on twin iterative deduction described in the present invention, the operating condition parameters at least include current, load percentage and motor frequency;

[0016] The first temperature parameter array includes the intake temperature, exhaust temperature, main bearing temperature, male rotor inlet end bearing temperature, male rotor row end bearing temperature, male rotor thrust bearing temperature, female rotor inlet end bearing temperature, female rotor row end bearing temperature, female rotor thrust bearing temperature, motor A / B / C phase winding temperature and main motor drive end and non-drive end temperature at the current time t.

[0017] As a preferred solution of the compressor temperature array abnormality prediction method based on twin iterative deduction according to the present invention, the first state parameter vector formed at the current time t includes:

[0018] Normalizing the collected operating condition parameters and the first temperature parameter array to 0~1 respectively;

[0019] The normalized parameters are synchronized and aligned with a unified timestamp and combined into a vector in chronological order. ;

[0020] in, is the working condition parameter sub-vector, is the temperature parameter subvector, is the first state parameter vector at the current time t;

[0021] The first state parameter vector is stored in a state buffer.

[0022] As a preferred solution of the compressor temperature array anomaly prediction method based on twin iterative deduction described in the present invention, the recurrent neural network has six layers, each layer contains 64 neurons, the batch size is 32, the drop rate is 0.1, the learning rate is 0.001, and the number of iterations is not less than 1000 times.

[0023] As a preferred solution of the compressor temperature array anomaly prediction method based on twin iterative deduction described in the present invention, the single temperature parameter twin prediction model for predicting the target temperature at time t+1 obtained by using recurrent neural network training includes:

[0024] Build with As input, target temperature is the output supervised training sample set;

[0025] The forward calculation and reverse gradient update are performed using the following formula as the comprehensive loss function:

[0026]

[0027] Where M is the number of samples, is the normalized time variable, is the time weight attenuation coefficient, , is the regularization weight, , is the mapping matrix, b is the bias vector, and are the operating condition parameter sub-vector and temperature parameter sub-vector of the mth sample respectively, is the attention weight, is the hidden unit activation value, is an exponential linear unit, K is the hidden layer dimension, When it approaches 0, the prediction accuracy of the representation model is optimal. represents the normalized time variable Next The target time corresponding to the sample The actual single temperature parameter value, The predicted samples in the normalized time variable The single temperature parameter value under .

[0028] As a preferred solution of the compressor temperature array anomaly prediction method based on twin iterative deduction according to the present invention, obtaining the multi-step deduction results includes:

[0029] Using a sliding time window, the first state parameter vector is input into the temperature parameter model set to predict the second temperature parameter array at the next moment;

[0030] Fusing the second temperature parameter array with the real-time operating condition parameter to form a second state parameter vector as input for the next round;

[0031] Repeat the above two steps, iterate N times, until the sequence is generated , which is the multi-step deduction result.

[0032] As a preferred solution of the compressor temperature array anomaly prediction method based on twin iterative deduction described in the present invention, the number of loop iterations N is 50.

[0033] As a preferred solution of the compressor temperature array anomaly prediction method based on twin iterative deduction according to the present invention, obtaining the deviation baseline of each temperature parameter includes:

[0034] The historical state matrix Zero mean normalization, where d is the dimension;

[0035] Calculate the covariance matrix using principal component analysis: , and solve its eigenvalue and the corresponding eigenvector is the principal component, where T represents the time length of the historical state matrix X, that is, the number of samples in X;

[0036] Select the smallest integer k that satisfies the cumulative contribution rate of not less than 95% so that , construct the transformation matrix for the first k principal components ;

[0037] Based on the transition matrix, that is, the twin parameter state vector at time t+1 Perform dimensionality reduction to obtain the principal component information matrix after dimensionality reduction ;

[0038] Based on the data after dimensionality reduction, the inverse matrix of the transformation matrix and the data after dimensionality reduction are used to restore the dimension. Due to the loss of information, only the maximum consistent information of all parameters is retained. Therefore, the deviation baseline of each temperature parameter can be obtained through inverse transformation. ,in is the inverse matrix of the transformation matrix;

[0039] The time mean curve of each temperature parameter in the sequence obtained from time t+1 to t+N is calculated in sequence and defined as the corresponding deviation baseline.

[0040] As a preferred solution of the compressor temperature array anomaly prediction method based on twin iterative deduction according to the present invention, the deviation ratio of each temperature parameter relative to its deviation baseline in the multi-step deduction result is calculated as follows:

[0041]

[0042] in, is the temperature parameter in the deduction step The predicted value in is the corresponding deviation baseline value, is the deviation ratio;

[0043] when When the temperature parameter is abnormal, the warning mark of the temperature parameter is triggered.

[0044] Beneficial effects of the present invention:

[0045] 1. By establishing an accurate and comprehensive data foundation for equipment operating status, the accuracy and completeness of the input data for subsequent prediction models are ensured, thereby providing high-quality data support for the prediction of compressor operating status and improving prediction accuracy and reliability;

[0046] 2. By independently and finely modeling each temperature parameter, the prediction accuracy of each individual temperature parameter is improved, the prediction deviation caused by mutual interference between different parameters is avoided, the dynamic change trend of a single temperature parameter is effectively captured, and the temperature prediction sensitivity and model refinement are enhanced;

[0047] 3. By comprehensively and systematically predicting various temperature parameters, the problem of insufficient data coverage of traditional monitoring methods using a single sensor or a small number of sensors is overcome, and an in-depth analysis of the overall temperature conditions of the compressor is achieved, allowing for accurate monitoring of the overall thermal conditions of the compressor.

[0048] 4. It overcomes the prediction hysteresis problem caused by existing technologies that only make simple inferences about the current or short-term status, detects potential anomalies in advance, and realizes active early warning and fault prevention;

[0049] 5. By reducing the dimensionality of the data to eliminate noise and redundant interference information, more representative features are extracted to accurately capture abnormal precursors. This overcomes the false positives or false negatives that may result from directly using the original prediction data in traditional methods, improves the reliability of anomaly detection, and reduces false alarms and misjudgments.

[0050] 6. Accurately quantify the degree of abnormality through deviation ratio, avoiding the inefficiency and uncertainty brought by subjective experience judgment or static threshold method of traditional monitoring system, improving the accuracy of abnormality judgment, predicting and preventing compressor failure risks in advance, and ensuring equipment safety and production continuity. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0052] Figure 1 This is a flow chart of the compressor temperature array abnormality prediction method based on twin iterative deduction shown in the present invention. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0054] Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without making any creative work should fall within the scope of protection of the present invention.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0056] According to an embodiment of the present invention, Figure 1 The flowchart shown is a method for predicting compressor temperature array abnormality based on twin iterative deduction, which specifically includes the following steps:

[0057] S1. Collect the operating parameters of the compressor and the first temperature parameter array in real time to form a first state parameter vector at the current time t;

[0058] S2. For each target temperature parameter in the first temperature parameter array, using the operating condition parameter at time t and the first temperature parameter array as input, a recurrent neural network is trained to obtain a single temperature parameter twin prediction model for predicting the target temperature at time t+1;

[0059] S3. Repeat step S2 to establish corresponding single temperature parameter twin prediction models for all temperature parameters to form a temperature parameter model set;

[0060] S4. Using the first state parameter vector at the current time t as input, sequentially calling the temperature parameter model set to obtain a second temperature parameter array at time t+1, and then combining the second temperature parameter array with the current operating condition parameters to form a second state parameter vector, iterating N times to obtain a multi-step deduction result;

[0061] S5. Using the historical state matrix containing the historical multi-time state parameter vectors, a conversion matrix is ​​established through principal component analysis, and the multi-step deduction results are subjected to dimensionality reduction and inverse transformation to obtain the deviation baseline of each temperature parameter;

[0062] S6. Calculate the deviation ratio of each temperature parameter relative to its deviation baseline in the multi-step deduction results. When the deviation ratio of any temperature parameter exceeds the warning threshold, output a compressor temperature array abnormality warning signal.

[0063] It should be noted that the collected compressor state parameter data includes operating parameters and temperature parameters. The operating parameters and full temperature parameters at time t are used as input, and the single temperature parameter at time t+1 is used as output. A recurrent neural network is used to sequentially construct twin prediction models of single temperature parameters to form a set of temperature parameter twin prediction models.

[0064] Furthermore, based on the current operating status of the compressor, the twin prediction model of each temperature parameter is cyclically called to obtain the single-step deduction result of the single temperature parameter. The temperature parameter results of the single-step deduction are summarized to obtain the single-step deduction temperature array. Then, according to a fixed number of iteration cycles, the deduction is cyclically iterated to obtain the multi-step twin deduction temperature array.

[0065] During the iterative deduction process, the state deviation analysis of the deduced temperature array is performed. Specifically, based on a large amount of historical data, principal component analysis is used to obtain the transformation matrix of the twin distribution law, and the multi-step twin deduction temperature matrix of the current operating state of the compressor is transformed. The dimension is restored by the inverse matrix of the transformation matrix and the data after dimensionality reduction, and the main information is retained to extract the main characteristics of the twin distribution law as the deviation baseline; the difference between the deduced value of the temperature parameter in the twin state parameter and the deviation baseline is calculated. If the difference ratio deviation exceeds 5%, the deduced deviation of the temperature parameter is relatively large, and there is a potential temperature warning risk.

[0066] The following describes in more detail the implementation process and / or effects of certain embodiments of the present invention in conjunction with some preferred or optional examples of the present invention.

[0067] As an example, the operating parameters include at least current, load percentage, and motor frequency.

[0068] As an example, the first temperature parameter array includes the intake temperature, exhaust temperature, main bearing temperature, male rotor inlet end bearing temperature, male rotor row end bearing temperature, male rotor thrust bearing temperature, female rotor inlet end bearing temperature, female rotor row end bearing temperature, female rotor thrust bearing temperature, motor A / B / C phase winding temperature and main motor drive end and non-drive end temperature at the current time t.

[0069] In an optional embodiment, forming a first state parameter vector at the current time t includes:

[0070] Normalize the collected working condition parameters and the first temperature parameter array to 0~1 respectively;

[0071] The normalized parameters are synchronized and aligned with a unified timestamp and combined into a vector in chronological order. ;

[0072] in, is the working condition parameter sub-vector, is the temperature parameter subvector, is the first state parameter vector at the current time t;

[0073] The first state parameter vector is stored in the state buffer.

[0074] It should be noted that the recurrent neural network used in this embodiment has six layers, each layer contains 64 neurons, the batch size is 32, the drop rate is 0.1, the learning rate is 0.001, and the number of iterations is not less than 1000 times.

[0075] In an optional embodiment, a single temperature parameter twin prediction model for predicting the target temperature at time t+1 is obtained by training a recurrent neural network, including:

[0076] Build with As input, target temperature is the output supervised training sample set;

[0077] The forward calculation and reverse gradient update are performed using the following formula as the comprehensive loss function:

[0078]

[0079] Where M is the number of samples, is the normalized time variable, is the time weight attenuation coefficient, , is the regularization weight, , is the mapping matrix, b is the bias vector, and are the operating condition parameter sub-vector and temperature parameter sub-vector of the mth sample respectively, is the attention weight, is the hidden unit activation value, is an exponential linear unit, K is the hidden layer dimension, When it approaches 0, the prediction accuracy of the representation model is optimal. represents the normalized time variable Next The target time corresponding to the sample The actual single temperature parameter value, The predicted samples in the normalized time variable Single temperature parameter value under ;

[0080] And the original output of the network Introduce interval constraint mapping:

[0081]

[0082] Among them, σ is the Sigmoid function, and are the minimum and maximum values ​​of the jth temperature parameter in the historical data;

[0083] An interval constraint mapping is introduced into the comprehensive loss function update to make the model converge within the physically feasible temperature range.

[0084] It should be noted that the comprehensive loss function in this embodiment is a single temperature parameter twin prediction model, and the physically feasible temperature range is the industry standard range.

[0085] Furthermore, multi-step deduction results are obtained, including:

[0086] Using a sliding time window, the first state parameter vector is input into the temperature parameter model set to predict the second temperature parameter array at the next moment;

[0087] The second temperature parameter array is integrated with the real-time operating condition parameters to form a second state parameter vector as the next round of input;

[0088] Repeat the above two steps, iterate N times, until the sequence is generated , which is the result of multi-step deduction.

[0089] As an example, the number of times the loop iterates N times is 50.

[0090] It should be noted that for the deduction results obtained by the single temperature parameter twin prediction model, the state parameters including each temperature should obey the distribution law of the state parameters. Therefore, it is necessary to perform deviation analysis on the deduced and predicted state vector to identify whether the compressor temperature has a deterioration trend;

[0091] For example, the principal component analysis method is used to construct a deviation identification model based on the distribution law of state parameters. By extracting the main change direction in the data, high-dimensional data is projected into a low-dimensional space to complete the trend feature extraction of the twin state parameters. After back-projecting each parameter, the fluctuation deviation is eliminated.

[0092] Based on the data after dimensionality reduction, the inverse matrix of the transformation matrix and the data after dimensionality reduction are used to restore the dimension. Due to the loss of information, only the maximum consistent information of all parameters is retained;

[0093] For each time step prediction result, the difference between the predicted value and the deviation baseline can be analyzed; in this embodiment, Taking the twin deduction result at the moment as an example, the deviation baseline of each twin temperature parameter can be obtained through inverse transformation.

[0094] Exemplarily, obtaining the deviation baseline of each temperature parameter includes:

[0095] The historical state matrix Zero mean normalization is used to eliminate the dimension effect, where d is the dimension;

[0096] Calculate the covariance matrix using principal component analysis: , and solve its eigenvalue and the corresponding eigenvector is the principal component, where T represents the time length of the historical state matrix X, that is, the number of samples in X;

[0097] Select the smallest integer k that satisfies the cumulative contribution rate of not less than 95% so that , construct the transformation matrix for the first k principal components ;

[0098] Based on the transition matrix, that is, the twin parameter state vector at time t+1 Perform dimensionality reduction to obtain the principal component information matrix after dimensionality reduction ;

[0099] Based on the data after dimensionality reduction, the inverse matrix of the transformation matrix and the data after dimensionality reduction are used to restore the dimension. Due to the loss of information, only the maximum consistent information of all parameters is retained. Therefore, the deviation baseline of each temperature parameter can be obtained through inverse transformation. ,in is the inverse matrix of the transformation matrix;

[0100] The time mean curve of each temperature parameter in the sequence obtained from time t+1 to t+N is calculated in sequence and defined as the corresponding deviation baseline.

[0101] It is not difficult to understand that principal component analysis is an unsupervised dimensionality reduction technique used to simplify high-dimensional data sets while retaining the key information (variance) in the original data as much as possible. Its core goal is to convert the original features into a set of new, independent (uncorrelated) features (principal components), which are sorted according to the defined data variance.

[0102] In this embodiment, the essence of the matrix transformation using principal component analysis is: extracting core features by projecting into the principal component space; generating a low-noise baseline by restoring the dimensions using inverse transformation; and establishing a dynamic and robust deviation baseline in a data-driven manner, which solves the bottlenecks of traditional methods in multi-parameter collaborative analysis and dynamic working condition adaptability, thereby improving the accuracy and reliability of abnormal warnings.

[0103] It should be further explained that principal component analysis filters high-frequency noise and minor fluctuations by retaining the principal component with the largest variance (95% cumulative contribution rate), so that the baseline focuses on the stable patterns of historical data and reduces false alarms; the baseline after inverse transformation is essentially the "average pattern" of the historical normal state, which can reflect the inherent coordinated change patterns of temperature parameters (such as the correlation between bearing temperature and load), rather than an isolated threshold; under variable load / speed scenarios, traditional static thresholds are prone to failure, while the baseline through principal component analysis dynamically captures the joint distribution of multiple parameters and is more adaptable to complex working conditions.

[0104] It should be noted that the calculation method for calculating the deviation ratio of each temperature parameter relative to its deviation baseline in the multi-step deduction results is:

[0105]

[0106] in, is the temperature parameter in the deduction step The predicted value in is the corresponding deviation baseline value, is the deviation ratio;

[0107] when When the temperature parameter is abnormal, the warning mark of the temperature parameter is triggered.

[0108] Further, Set a dynamic warning threshold for the i-th temperature parameter:

[0109]

[0110] in, is the global minimum safety threshold, is the deviation amplification factor, The number of the most recent S deduction steps (S ≥ N) The rolling standard deviation of

[0111] For each deduction step Generate a Boolean decision flag:

[0112]

[0113] in, For deduction step The deviation ratio of the i-th temperature parameter in, is the warning threshold dynamically calculated for the i-th temperature parameter;

[0114] 1 means that the temperature parameter triggers an early warning in this deduction step;

[0115] 0 means no warning is triggered;

[0116] For example, when an early warning is triggered, it is immediately written into the alarm queue and the temperature array abnormality early warning signal is issued to the compressor control system through the MQTT protocol, and the triggering timestamp, the exceeded parameter number and its value.

[0117] The aforementioned principal component analysis and matrix conversion method of the historical state matrix can be performed using methods and means in the existing technology and will not be described in detail in this example.

[0118] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A compressor temperature array abnormality prediction method based on twin iterative deduction, characterized in that: include: Collect the operating parameters of the compressor and the first temperature parameter array in real time to form a first state parameter vector at the current time t; For each target temperature parameter in the first temperature parameter array, using the operating condition parameter at time t and the first temperature parameter array as input, a recurrent neural network is trained to obtain a single temperature parameter twin prediction model for predicting the target temperature at time t+1; include: Build with As input, target temperature is the output supervised training sample set; The forward calculation and reverse gradient update are performed using the following formula as the comprehensive loss function: Where M is the number of samples, is the normalized time variable, is the time weight attenuation coefficient, , is the regularization weight, , is the mapping matrix, b is the bias vector, and are the operating condition parameter sub-vector and temperature parameter sub-vector of the mth sample respectively, is the attention weight, is the hidden unit activation value, is an exponential linear unit, K is the hidden layer dimension, When it approaches 0, the prediction accuracy of the representation model is optimal. represents the normalized time variable Next The target time corresponding to the sample The actual single temperature parameter value, The predicted samples in the normalized time variable Single temperature parameter value under ; Repeat the previous step to establish corresponding single temperature parameter twin prediction models for all temperature parameters to form a temperature parameter model set; Taking the first state parameter vector at the current time t as input, the temperature parameter model set is sequentially called to obtain a second temperature parameter array at the time t+1, and the second temperature parameter array and the current operating condition parameters are combined to form a second state parameter vector, which is iterated N times to obtain a multi-step deduction result; Using a historical state matrix containing historical multi-time state parameter vectors, a conversion matrix is ​​established through principal component analysis, and the multi-step deduction results are subjected to dimensionality reduction and inverse transformation to obtain a deviation baseline of each temperature parameter; The deviation ratio of each temperature parameter in the multi-step deduction result relative to its deviation baseline is calculated, and when the deviation ratio of any temperature parameter exceeds the warning threshold, a compressor temperature array abnormality warning signal is output.

2. The compressor temperature array abnormality prediction method based on twin iterative deduction according to claim 1 is characterized in that: The operating parameters include at least current, load percentage and motor frequency; The first temperature parameter array includes the intake temperature, exhaust temperature, main bearing temperature, male rotor inlet end bearing temperature, male rotor row end bearing temperature, male rotor thrust bearing temperature, female rotor inlet end bearing temperature, female rotor row end bearing temperature, female rotor thrust bearing temperature, motor A / B / C phase winding temperature and main motor drive end and non-drive end temperature at the current time t.

3. The compressor temperature array abnormality prediction method based on twin iterative deduction according to claim 1 or 2, characterized in that: The forming of the first state parameter vector at the current time t includes: Normalizing the collected operating condition parameters and the first temperature parameter array to 0~1 respectively; The normalized parameters are synchronized and aligned with a unified timestamp and combined into a vector in chronological order. ; in, is the working condition parameter sub-vector, is the temperature parameter subvector, is the first state parameter vector at the current time t; The first state parameter vector is stored in a state buffer.

4. The compressor temperature array abnormality prediction method based on twin iterative deduction according to claim 1 is characterized in that: The recurrent neural network has six layers, each layer contains 64 neurons, the batch size is 32, the drop rate is 0.1, the learning rate is 0.001, and the number of iterations is not less than 1000 times.

5. The compressor temperature array abnormality prediction method based on twin iterative deduction according to claim 1 is characterized in that: The obtaining of multi-step deduction results includes: Using a sliding time window, the first state parameter vector is input into the temperature parameter model set to predict the second temperature parameter array at the next moment; Fusing the second temperature parameter array with the real-time operating condition parameter to form a second state parameter vector as input for the next round; Repeat the above two steps, iterate N times, until the sequence is generated , which is the multi-step deduction result.

6. The compressor temperature array abnormality prediction method based on twin iterative deduction according to claim 5 is characterized in that: The number of times the loop iterates N times is 50.

7. The compressor temperature array abnormality prediction method based on twin iterative deduction according to claim 1 is characterized in that: Obtain the deviation baseline of each temperature parameter, including: The historical state matrix Zero mean normalization, where d is the dimension; Calculate the covariance matrix using principal component analysis: , and solve its eigenvalue and the corresponding eigenvector is the principal component, where T represents the time length of the historical state matrix X, that is, the number of samples in X; Select the smallest integer k that satisfies the cumulative contribution rate of not less than 95% so that , construct the transformation matrix for the first k principal components ; Based on the transition matrix, that is, the twin parameter state vector at time t+1 Perform dimensionality reduction to obtain the principal component information matrix after dimensionality reduction ; Based on the data after dimensionality reduction, the inverse matrix of the transformation matrix and the data after dimensionality reduction are used to restore the dimension. Due to the loss of information, only the maximum consistent information of all parameters is retained. Therefore, the deviation baseline of each temperature parameter can be obtained through inverse transformation. ,in is the inverse matrix of the transformation matrix; The time mean curve of each temperature parameter in the sequence obtained from time t+1 to t+N is calculated in sequence and defined as the corresponding deviation baseline.

8. The compressor temperature array abnormality prediction method based on twin iterative deduction according to claim 7 is characterized in that: Calculate the deviation ratio of each temperature parameter in the multi-step deduction result relative to its deviation baseline, and the calculation method is: in, is the temperature parameter in the deduction step The predicted value in is the corresponding deviation baseline value, is the deviation ratio; when When the temperature parameter is abnormal, the warning mark of the temperature parameter is triggered.

Images