Abnormality monitoring method and device for intelligent separate injection equipment of oil field

Abstract

The invention discloses an abnormity monitoring method and device for oil field intelligent separate injection equipment. The abnormity monitoring method comprises the steps that operation data of the oil field intelligent separate injection equipment are collected in real time through multiple sensors; the collected operation data are preprocessed; performing wavelet multi-scale decomposition on the preprocessed data, and constructing an abnormal mode database based on a decomposition result; an improved Transform network is adopted to carry out abnormal type identification on the features obtained after wavelet decomposition; and performing real-time monitoring on the identified abnormity based on a predictive maintenance algorithm, and generating alarm information. According to the method, real-time data are collected through various sensors, abnormal types are accurately identified in combination with wavelet multi-scale decomposition and an improved Transform algorithm, and a predictive maintenance strategy is supplemented, so that dynamic monitoring and efficient alarm of potential faults of equipment are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent diagnosis and maintenance of oilfield equipment, and particularly relates to an abnormal monitoring method and device for intelligent oilfield water injection allocation equipment. Background Art

[0002] With the continuous improvement of the digitalization and automation levels of the petroleum industry, intelligent oilfield water injection allocation equipment plays an increasingly crucial role in the processes of oil exploitation and water injection. Traditional water injection allocation systems mostly monitor and adjust indicators such as flow rate, pressure, and temperature based on mechanical or semi-automatic means. However, in the face of the increasing data scale and the more complex working condition requirements, it is difficult for such systems to determine in a timely and accurate manner whether the equipment is in an abnormal state.

[0003] In recent years, the booming development of data sensor technology, signal processing methods, and machine learning algorithms has provided new opportunities for the automatic monitoring and fault diagnosis of intelligent oilfield water injection allocation equipment. By deploying a variety of high-precision sensors to collect equipment operation data and cooperating with advanced time series analysis algorithms, potential fault signs can be identified in the early stage. However, how to extract strongly discriminative features in a long-term, high-noise, and strongly non-stationary environment remains a difficult problem. Existing technologies often use simple filtering or time-frequency analysis means for abnormal detection, but there are still deficiencies in multi-sensor fusion and rapid response. Especially for intelligent water injection allocation equipment, the coupling relationships among multiple indicators such as water injection volume, pressure, and temperature are becoming increasingly complex, and traditional methods are prone to limitations such as insufficient extraction of abnormal features, large modeling deviations, and inaccurate timing of maintenance.

[0004] Currently, due to the significant oilfield environmental noise and the variable working mechanisms of water injection allocation equipment, conventional fault diagnosis methods often only rely on simple signal analysis or single machine learning models, and it is difficult to obtain ideal results in scenarios involving multi-scale and multi-source data. In addition, limited by the system architecture and algorithm capabilities, traditional solutions often can only give rough alarm instructions after abnormal identification, and do not fully exploit the potential of predictive maintenance. It is difficult to quantitatively analyze the remaining life or potential faults of equipment in a timely and dynamic manner. Especially in the case of sudden abnormalities, if the signal details cannot be captured quickly and accurately, it may lead to the failure to detect faults in time and cause greater losses.

[0005] The existing technologies also have relatively limited capabilities in differentiating abnormal types and early warning, and are prone to false alarms or missed alarms. When the large-scale data collected by each sensor fluctuates significantly in both the time dimension and the frequency dimension, if there is a lack of in-depth modeling and comprehensive identification of multi-scale features, abnormal diagnosis often has the defects of lag and low accuracy.

[0006] It can be seen that for the intelligent water injection allocation equipment in oilfields, there is an urgent need to establish a more efficient abnormal monitoring method to achieve accurate detection and predictive maintenance in complex and changing environments, and effectively reduce maintenance costs and safety risks. Summary of the Invention

[0007] The purpose of this part is to outline some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Simplifications or omissions may be made in this part, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this part, the abstract, and the title. However, such simplifications or omissions shall not be used to limit the scope of the present invention.

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

[0009] Therefore, the technical problems solved by the present invention are: the existing fault monitoring technology for water injection allocation equipment has problems such as insufficient multi-scale heterogeneous data fusion, low accuracy in early warning and remaining life prediction, and lack of effective learning and recognition of new abnormal patterns.

[0010] To solve the above technical problems, the present invention provides the following technical solutions: real-time collect the operation data of the intelligent water injection allocation equipment in oilfields by using a variety of sensors;

[0011] Preprocess the collected operation data;

[0012] Perform wavelet multi-scale decomposition on the preprocessed data, and construct an abnormal pattern database based on the decomposition results;

[0013] Use an improved Transformer network to identify the abnormal types of the features after wavelet decomposition;

[0014] Based on the predictive maintenance algorithm, monitor the identified abnormalities in real time and generate alarm information.

[0015] As a preferred solution of the abnormal monitoring method for the intelligent water injection allocation equipment in oilfields according to the present invention, the real-time collection of the operation data of the intelligent water injection allocation equipment in oilfields includes:

[0016] Install pressure, flow, temperature, and vibration sensors at key parts of the intelligent water injection allocation equipment in oilfields to capture signals that are prone to failure or abnormality during the operation of the intelligent water injection allocation equipment in oilfields;

[0017] Record the signals of each sensor through a multi-channel data acquisition system, and align the timestamps of different sensor data at the data acquisition end to obtain a synchronized multi-dimensional data sequence.

[0018] As a preferred solution of the abnormal monitoring method for the intelligent water injection equipment in oil fields of the present invention, the data preprocessing includes data cleaning, denoising processing, data standardization, data enhancement, and data labeling.

[0019] As a preferred solution of the abnormal monitoring method for the intelligent water injection equipment in oil fields of the present invention, the wavelet multi-scale decomposition of the preprocessed data and the construction of an abnormal pattern database based on the decomposition results include:

[0020] Performing continuous wavelet transform on the preprocessed sensor time series signal, selecting the Morlet wavelet as the mother wavelet function, and obtaining the wavelet coefficient distribution at each scale;

[0021] Extracting typical statistical features from the wavelet coefficients at each scale, screening each scale in combination with the band energy feature, and obtaining the key scale with the most significant fault diagnosis significance;

[0022] Inducing the historical fault cases corresponding to the key scale and its feature vector to form a database corresponding to the abnormal pattern - fault type;

[0023] Among them, the database is the abnormal pattern database.

[0024] As a preferred solution of the abnormal monitoring method for the intelligent water injection equipment in oil fields of the present invention, the typical statistical features at least include mean, variance, skewness, and kurtosis.

[0025] As a preferred solution of the abnormal monitoring method for the intelligent water injection equipment in oil fields of the present invention, the use of an improved Transformer network to identify the abnormal type of the features after wavelet decomposition includes:

[0026] Concatenating the multi-scale feature vectors in chronological order into an input tensor;

[0027] In the self-attention layer of the Transformer, performing weighted calculations on the key, query, and value matrices;

[0028] After the multi-head attention output, setting a residual connection and performing layer normalization;

[0029] Through the fully connected layer, mapping the final output vector to the predicted probability distribution of each fault type, and outputting the most likely fault type label according to the maximum probability term.

[0030] As a preferred solution of the abnormal monitoring method for the intelligent water injection equipment in oil fields of the present invention, the real-time monitoring of the identified abnormality based on the predictive maintenance algorithm and generating alarm information includes:

[0031] After the detected fault type is matched with the abnormal pattern database, the corresponding maintenance period or maintenance window is set by combining the real-time operating parameters and the equipment health assessment.

[0032] Use a Markov chain model combined with state transition probability to dynamically analyze the current health state of the equipment, and visualize the remaining life of the equipment as a probability distribution varying with time.

[0033] When the remaining life prediction result shows that the equipment is about to enter the high-risk area or the probability of critical component damage increases, the system automatically sends an alarm message to the operation and maintenance platform.

[0034] Among them, the alarm message at least includes the fault type, the recommended maintenance time point, and the key state parameters of the equipment.

[0035] As a preferred solution of the abnormal monitoring device for oilfield intelligent water injection equipment described in the present invention, it includes:

[0036] One or more processors;

[0037] A memory that stores operable instructions, and when the instructions are executed by the one or more processors, the one or more processors perform operations, and the operations include the processes of the abnormal monitoring method for oilfield intelligent water injection equipment as described above.

[0038] As a preferred solution of a computer-readable medium storing software described in the present invention, the software includes instructions that can be executed by one or more computers, and when the instructions are executed in this way, the one or more computers perform operations, and the operations include the processes of the abnormal monitoring method for oilfield intelligent water injection equipment as described above.

[0039] The beneficial effects of the present invention:

[0040] 1. Realize the real-time monitoring of the equipment operation state, ensure the accuracy and integrity of the data foundation, lay a solid foundation for abnormal feature extraction and fault diagnosis, and improve the adaptability of the system to dynamic changes and complex environments;

[0041] 2. Realize the effective regularization and optimization of complex and diverse original data, provide more representative data input for abnormal feature extraction, and improve the stability and accuracy of model training;

[0042] 3. Realize the multi-scale capture and systematic induction of abnormal features, not only improve the resolution ability for multi-type faults, but also provide strong support for the dynamic update and efficient query of fault patterns;

[0043] 4. Achieved precise classification and dynamic adaptation of complex abnormal features, greatly enhancing the recognition ability for multiple types of abnormalities, significantly reducing the possibilities of false alarms and missed detections, and providing guarantee for the reliability of equipment operation;

[0044] 5. Achieved closed-loop management from fault recognition to maintenance decision-making, significantly improving the safety and operation and maintenance efficiency of equipment, extending the service life of equipment, and reducing the operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Among them:

[0046] Figure 1 It is a schematic flowchart of the abnormal monitoring method for the intelligent water injection allocation equipment in oilfields shown in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In order to make the above-mentioned objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention with reference to the drawings of the specification. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments.

[0048] Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0049] In the following description, many specific details are set forth to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0050] According to the embodiments of the present invention, in combination with Figure 1 the flowchart shown, an abnormal monitoring method for intelligent water injection allocation equipment in oilfields specifically includes the following steps:

[0051] S1. Use multiple sensors to collect the operation data of the intelligent water injection allocation equipment in oilfields in real time. Among them, it should be noted in this step that:

[0052] Deploy pressure, flow, temperature, and vibration sensors at key parts (such as injection valves, pipeline interfaces, and pump bodies) of the intelligent water injection allocation equipment in oilfields to capture signals (i.e., operation data) that are prone to failures or abnormalities during the operation of the equipment;

[0053] Select a multi-channel data acquisition system with a high sampling rate (such as 1 kHz to 10 kHz) to capture fast-changing dynamic signals;

[0054] Connect the signals collected by each sensor to the data acquisition end, and perform unified clock synchronization settings or timestamp recording on the acquisition end to ensure that the multi-channel signals are strictly aligned on the time axis and avoid timing misalignment of the data of each sensor;

[0055] In an alternative embodiment, to ensure the timing consistency of the data, high-precision timestamps are added to the data of each channel at the data acquisition end. For example, the NTP (Network Time Protocol) or GPS time synchronization method is used to generate a synchronized multi-dimensional data sequence;

[0056] Based on the characteristics of each sensor, the measurement range and tolerance range are preset (for example, the measurement range of the pressure sensor is 100 m 3 / h, tolerance ±0.5 m 3 / h). If the signal of a certain sensor exceeds the preset threshold (that is, the set measurement range and tolerance range), the high-frequency sampling mode is enabled, and this high-frequency sampling segment is uploaded to the data management platform.

[0057] It should be noted that the operation data collected in this step is multi-channel, multi-sensor and raw operation data aligned in timestamps.

[0058] S2. Preprocess the collected operation data. Among them, it should be noted in this step that:

[0059] If the sensor fails or fails to record data successfully during a certain period, a linear interpolation algorithm is selected to complete the data filling. When the missing data is irregular, the corresponding segment of data is removed;

[0060] If the record appears repeatedly or exceeds the upper / lower limit of the physical measurement range of the sensor, it is determined as dirty data or error data and deleted;

[0061] For low-frequency or high-frequency noise caused by vibration or electromagnetic interference in the oilfield operation environment, a band-pass filter is used for large-scale noise reduction to filter out extreme low-frequency drift and high-frequency clutter. The filtered signal is further decomposed into several intrinsic mode functions (IMFs), and the remaining high-frequency noise or invalid components are removed according to the energy characteristics of each IMF component, so as to obtain a relatively stable and less noisy signal;

[0062] Perform the operations of subtracting the mean and dividing by the standard deviation on the data of each sensor channel to make it conform to the distribution of μ = 0 and σ = 1. Exemplarily, its data expression formula is:

[0063]

[0064] Among them, is the mean value of this sensor channel, and σ x is the standard deviation of this sensor channel;

[0065] Then, perform min-max normalization on this data:

[0066]

[0067] Among them, x min and x max are respectively the minimum value and the maximum value of this sensor in the operating data, x is the collected operating data, and x norm is the data value after normalization;

[0068] Select a window size w and a sliding step s, and divide the time series into several consecutive overlapping small segments, each with a length of w, to form a time series signal x(t);

[0069] Add a small amount of Gaussian noise to the slices (i.e., small segments) to increase the diversity of training data and reduce the risk of model overfitting;

[0070] Mark the known fault events in the historical records and attach fault labels to the corresponding time periods and sensor data slices;

[0071] According to the time sequence and fault distribution, divide the labeled data into a sample training set, a validation set, and a test set (the division ratio is 7:2:1) for verifying the abnormal type identification in the subsequent steps.

[0072] S3. Perform wavelet multi-scale decomposition on the preprocessed data and construct an abnormal pattern database based on the decomposition results. Among them, it should be noted in this step that:

[0073] Select a mother wavelet function ψ (such as Morlet or Mexican Hat wavelet) to perform multi-scale decomposition on the preprocessed time series signal x(t);

[0074] As an example, taking the Morlet wavelet as an example, its calculation formula is:

[0075]

[0076] Among them, x(t) is the input sensor signal, a is the scale factor, b is the time translation factor, and ψ * (·) represents the complex conjugate, and the wavelet coefficient W ψ (a,b) is used to represent the characteristic distribution of the signal at different scales a and times b;

[0077] In the wavelet coefficient matrix, for different scales (such as a 1, a 2 , …, a n ) to extract statistical features such as mean, variance, skewness, and kurtosis, and calculate the band energy to measure the importance of a specific frequency band for fault features;

[0078] Exemplarily, its mathematical calculation formula is:

[0079]

[0080] where E is the band energy, a i is the i-th scale;

[0081] Select the key scale a that can best reflect the fault features according to the preset threshold k , and form a multi-scale feature vector:

[0082] f t = [f t,1 , f t,2 , …, f t,D

[0083] where D represents the finally retained feature dimension;

[0084] Associate the feature vector extracted from the key scale with the labeled historical fault types to form a mapping relationship between the feature vector and the fault type;

[0085] After sorting, store it in the abnormal pattern database. When a feature vector with a high similarity to the pattern in this database is detected in the subsequent steps, it can be determined as the corresponding fault type or suspected fault;

[0086] It should be noted that the abnormal pattern database can be iteratively updated as new data continuously enters to improve the coverage of multiple fault patterns.

[0087] S4. Use an improved Transformer network to identify the abnormal type of the features after wavelet decomposition. Among them, it should be noted in this step:

[0088] Concatenate the multi-scale feature vector f t output in step S3 into an input tensor X ∈ R T×D ;

[0089] where T is the time step (or the number of sliding windows), and D is the feature dimension corresponding to each time step;

[0090] In an alternative embodiment, if it is necessary to process multi-sensor parallel features, multi-sensor channel concatenation or nesting can be performed on dimension D;

[0091] In the self-attention layer of the Transformer encoder, weighted calculations are performed on the query vector Q, the key vector K, and the value vector V:

[0092]

[0093] where d k is the dimension of the key vector k;

[0094] A learnable scale weighting vector α ∈ R D is introduced, which represents the weight factor for different scale features during attention calculation. Specifically, before calculating Q, K, and V, it is achieved by element-wise multiplication (f t ⊙α) for the corresponding channels or scales, enabling the above weighted calculation model to adaptively identify the scales or sensor channels with the most fault diagnosis value;

[0095] The output of the multi-head attention is added to its input residually, allowing the gradient to be directly backpropagated to the deep structure and alleviating the vanishing gradient;

[0096] Layer normalization is used to normalize the output distribution, reducing the impact of excessive numerical differences between outputs at different time steps;

[0097] The final output vector of the Transformer encoder is input into a fully connected layer or a classification module based on attention-weighted aggregation to output a fault type prediction probability vector:

[0098] p = softmax(W out ·h enc + b out )S(t)

[0099] where W out and b out are trainable parameters, and h enc represents the output vector of the Transformer encoder;

[0100] The specific fault type is determined according to the dimension where the maximum value in p is located (e.g., 1 for vibration anomaly, 2 for temperature anomaly, 3 for pressure anomaly);

[0101] If the highest probability value is less than a set threshold (e.g., the threshold is set to 0.6), then this data sample is determined to be undecidable or suspected of being a new fault and is updated through manual discrimination;

[0102] Via the above process operations, the identified fault type is output.

[0103] S5. Based on the predictive maintenance algorithm, real-time monitoring of the identified anomalies is performed, and alarm information is generated. It should be noted in this step that:

[0104] Match the currently recognized fault type with the abnormal pattern database constructed in step S3;

[0105] Combine real-time operating parameters (such as current load, temperature, flow rate) and equipment health assessment (obtained by integrating past maintenance records and cumulative working hours) to set the corresponding maintenance window;

[0106] If the fault or abnormal pattern has a greater impact on the safe operation of the equipment, shorten the maintenance window and arrange for priority maintenance;

[0107] Use a Markov chain model combined with state transition probabilities to dynamically analyze the current health state of the equipment, and visualize the remaining life of the equipment as a probability distribution that changes over time;

[0108] As an example, let the health state of the equipment be represented as S(t) at different times t, and its discrete state is modeled by a Markov chain. Suppose there is a state set {s 0 , s 1 , …, s n};

[0109] Among them, s 0 is the normal state, and s n is the complete fault state;

[0110] Set the transition probability matrix:

[0111] P ∈ R (n+1)×(n+1)

[0112] P ij = P(S(t + 1) = s j | S(t) = s i )

[0113] Among them, s i , s j are the i-th and j-th states respectively, P is the transition probability, P ij is the transition probability from state i to state j, R is the set to which the transition probability belongs, and n is the quantity;

[0114] Modify the transition probability in real time by combining the currently detected fault type. For example, when a severe vibration fault occurs, the transition probability from s i to s n will increase;

[0115] The remaining useful life RUL is inferred by the probability of the complete fault state s n occurring at future times. If P(s n , T) exceeds the critical value (such as 0.9) around time T, it can be determined that the remaining useful life is approaching;

[0116] When the prediction result indicates that the device is about to enter the high-risk area, the system automatically sends an alarm message to the operation and maintenance platform. The alarm message includes the fault type, the recommended maintenance time point, and the key status parameters of the device.

[0117] Furthermore, after the maintenance personnel complete the repair, record the actual fault location, damage degree, repair or replacement cycle information.

[0118] Compare the actual repair results (such as replaced parts, repair cycle, etc.) with the prediction information, and integrate the real situation and new fault modes into the abnormal mode database, and iteratively update the transition probability of the Markov chain and the Transformer network parameters.

[0119] Preferably, through the implementation of the above steps, it is possible to achieve a closed-loop of abnormal monitoring, fault identification, and predictive maintenance during the actual operation of the intelligent water injection equipment in the oilfield. Compared with the traditional regular maintenance or passive reactive maintenance, the method of the present invention can detect abnormal signs at an early stage and quantitatively predict the fault process, greatly improving production efficiency and equipment safety.

[0120] The generation of the foregoing multi-dimensional data sequence, the time synchronization method, the typical statistical feature extraction method, and the equipment health assessment method can be carried out by means and methods in the prior art, and will not be elaborated in this example.

[0121] It should be noted that some other aspects disclosed in the embodiments of the present invention also propose an abnormal monitoring device for intelligent water injection equipment in the oilfield, including: one or more processors and a memory.

[0122] The memory is used to store executable instructions, which, when executed by the one or more processors, cause the one or more processors to perform operations, including the processes of an abnormal monitoring method for intelligent water injection equipment in the oilfield in the foregoing embodiments, especially Figure 1 the processes of the method shown.

[0123] Some other aspects disclosed in the embodiments of the present invention also propose a computer-readable medium storing software, and these software include instructions executable by one or more computers, and these instructions, when executed in this way, cause the one or more computers to perform operations, including the processes of an abnormal monitoring method for intelligent water injection equipment in the oilfield in the foregoing embodiments, especially Figure 1 the processes of the method shown.

[0124] It should be recognized that the embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory.

[0125] The described method can be implemented using standard programming techniques, including in a non-transitory computer-readable storage medium configured with a computer program in a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner.

[0126] Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program can be implemented in assembly or machine language.

[0127] In any case, the language can be a compiled or interpreted language.

[0128] In addition, for this purpose the program is capable of running on a programmed application-specific integrated circuit.

[0129] The processes described herein (or variations and / or combinations thereof) can be executed under the control of one or more computer systems configured with executable instructions and can be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) executed jointly on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions executable by one or more processors.

[0130] Further, the method can be implemented in any type of computing platform operably connected, including but not limited to personal computers, minicomputers, mainframes, workstations, network or distributed computing environments, separate or integrated computer platforms, or communicating with charged particle tools or other imaging devices.

[0131] Aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it is readable by a programmable computer and can be used to configure and operate the computer to perform the processes described herein when the storage medium or device is read by the computer.

[0132] In addition, the machine-readable code, or portions thereof, can be transmitted via a wired or wireless network.

[0133] When such media includes instructions or programs that implement the above-described steps in conjunction with a microprocessor or other data processor, the inventions described herein include these and other different types of non-transitory computer-readable storage media.

[0134] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for abnormal monitoring of intelligent injection equipment in oil fields, characterized in that: include: Use a variety of sensors to collect real-time operating data of oilfield intelligent injection equipment; Preprocess the collected operation data; Perform wavelet multi-scale decomposition on the preprocessed data and build an abnormal pattern database based on the decomposition results; An improved Transformer network is used to identify abnormal types of the features after wavelet decomposition; Based on the predictive maintenance algorithm, the identified anomalies are monitored in real time and alarm information is generated.

2. The abnormality monitoring method for oilfield intelligent injection equipment according to claim 1 is characterized in that: The real-time collection of operating data of the oilfield intelligent injection equipment includes: Pressure, flow, temperature and vibration sensors are arranged at key parts of the oilfield intelligent dispensing equipment to capture signals that are prone to failure or abnormality during the operation of the oilfield intelligent dispensing equipment; The signals of each sensor are recorded separately through a multi-channel data acquisition system, and the timestamps of different sensor data are aligned at the data acquisition end to obtain a synchronized multi-dimensional data sequence.

3. The abnormality monitoring method for oilfield intelligent injection equipment according to claim 1 is characterized in that: The data preprocessing includes data cleaning, denoising, data standardization, data enhancement and data labeling.

4. The abnormality monitoring method for oilfield intelligent injection equipment according to claim 1 is characterized in that: The method of performing wavelet multi-scale decomposition on the pre-processed data and constructing an abnormal pattern database based on the decomposition results includes: The preprocessed sensor time series signal is transformed by continuous wavelet transform, and Morlet wavelet is selected as the mother wavelet function to obtain the wavelet coefficient distribution at each scale. Typical statistical features are extracted from the wavelet coefficients of each scale, and each scale is screened in combination with the frequency band energy features to obtain the key scale with the most significance for fault diagnosis; Summarizing the historical fault cases corresponding to the key scales and their characteristic vectors to form a database corresponding to abnormal modes and fault types; Wherein, the database is an abnormal pattern database.

5. The abnormality monitoring method for oilfield intelligent injection equipment according to claim 4 is characterized in that: The typical statistical characteristics at least include mean, variance, skewness and kurtosis.

6. The abnormality monitoring method for oilfield intelligent injection equipment according to claim 1 is characterized in that: The improved Transformer network is used to identify abnormal types of features after wavelet decomposition, including: Concatenate the multi-scale feature vectors into the input tensor in time order; In the Transformer’s self-attention layer, weighted calculations are performed on the key, query, and value matrices. After the multi-head attention output, set the residual connection and perform layer normalization; Through the fully connected layer, the final output vector is mapped to the predicted probability distribution of each fault type, and the most likely fault type label is output according to the maximum probability item.

7. The abnormality monitoring method for oilfield intelligent injection equipment according to claim 1 is characterized in that: The predictive maintenance algorithm is used to monitor the identified anomalies in real time and generate alarm information, including: When the detected fault type is matched with the abnormal pattern database, the corresponding maintenance cycle or maintenance window is set in combination with the real-time operating parameters and equipment health assessment; The Markov chain model combined with the state transition probability is used to dynamically analyze the current health status of the equipment and visualize the remaining life of the equipment as a probability distribution that changes over time. When the remaining life prediction results show that the equipment is about to enter a high-risk area or the probability of damage to key components increases, the system automatically sends an alarm message to the operation and maintenance platform; The alarm information at least includes the fault type, the recommended maintenance time point and key equipment status parameters.

8. An abnormality monitoring device for oilfield intelligent injection equipment, characterized in that: include: one or more processors; A memory storing operable instructions, wherein when the instructions are executed by the one or more processors, the one or more processors perform operations, wherein the operations include the process of the abnormality monitoring method for oilfield intelligent injection equipment as described in any one of claims 1 to 7.

9. A computer-readable medium storing software, characterized in that: The software includes instructions that can be executed by one or more computers, and the instructions enable the one or more computers to perform operations through such execution, and the operations include the process of the abnormality monitoring method for oilfield intelligent injection equipment as described in any one of claims 1 to 7.

Images