An oil equipment operation and maintenance method and system based on an internet of things

Abstract

The application discloses an oil equipment operation and maintenance method and system based on the Internet of Things, which comprises collecting wellhead pressure data, extracting a deep time-frequency fault feature vector, calculating a trend coefficient K, a fluctuation index I, a mutation degree D, a stability S and a pressure value P of the pressure data, setting a threshold value, combining the calculated pressure data and the set threshold value to perform fault diagnosis, and the priority order of fault treatment is safety emergency fault, instrument fault, lifting equipment fault and wellbore fault in sequence; and performing maintenance operation according to the priority order of fault treatment; through the extraction of the deep fault feature vector of the wellhead pressure data, the trend coefficient K, the fluctuation index I, the mutation degree D, the stability S and the pressure value P are calculated, the fault diagnosis is performed in combination with the threshold value, the safety emergency fault, the wellbore fault, the lifting equipment fault and the instrument fault affecting the wellhead pressure change are obtained, the fault type is accurately known, and the fault is quickly and accurately positioned.

Description

Technical Field

[0001] This invention relates to the field of oil equipment operation and maintenance technology, and in particular to an Internet of Things-based oil equipment operation and maintenance method and system. Background Technology

[0002] In oil extraction, wellhead pressure data is a key parameter reflecting downhole conditions and equipment operating status. Current oil equipment maintenance methods mostly rely on setting maximum safe pressure limits and minimum pressure alarm thresholds for fault diagnosis and early warning.

[0003] However, in actual production environments, wellhead pressure data changes are affected by various factors such as pumping unit strokes, formation fluid supply capacity, tubing wax buildup, and gas influence. Relying solely on fixed thresholds can easily lead to false alarms or missed alarms. In particular, for safety emergencies caused by overpressure or a sharp rise in pressure, as well as wellbore faults caused by tubing leaks or insufficient formation fluid supply resulting in a continuous and slow pressure drop, current technology, which relies solely on pressure thresholds, may misjudge wellbore faults as safety emergencies, causing unnecessary shutdowns for maintenance. This necessitates time for comprehensive fault diagnosis, and even minor faults causing wellhead pressure changes require prolonged shutdowns for maintenance, leading to unnecessary halts in the entire oil extraction operation.

[0004] In view of this, this solution proposes an IoT-based operation and maintenance method and system for oil equipment to solve the problem of not being able to identify specific fault types caused by pressure changes. Summary of the Invention

[0005] The purpose of this invention is to provide an Internet of Things-based method and system for the operation and maintenance of petroleum equipment, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for the operation and maintenance of oil equipment based on the Internet of Things, comprising the following operational steps:

[0007] S1: Collect wellhead pressure data of oil equipment, process the collected pressure data using time-varying parameter multi-synchronous extrusion transformation to generate high-resolution pressure time-frequency spectrum, and extract depth time-frequency fault feature vector from the time-frequency spectrum;

[0008] S2, based on the pressure time-frequency spectrum and the deep time-frequency fault feature vector, calculates the trend coefficient K, fluctuation index I, abrupt change degree D, stability S and pressure value P of the pressure data;

[0009] S3, based on the design parameters of the oil equipment, sets the slow descent threshold K respectively. s Rapidly rising threshold K f Threshold K for near-zero change o Maximum fluctuation threshold Imax Minimum fluctuation threshold I min Threshold for sudden rate of change D t State stability threshold S u Data lag threshold S v Minimum pressure threshold P min Approaching zero pressure value P z Maximum safe pressure threshold P max Effective pressure lower limit P a and the upper limit of effective pressure P b ;

[0010] S4, combining the pressure data calculated by S2 and the threshold set by S3 to perform fault diagnosis;

[0011] When P > P max and / or K > K f When this occurs, it is determined to be a safety emergency.

[0012] When K < K s The downward trend lasts for more than 24 hours, and I < I max D < D t And P < P max At that time, it was determined to be a wellbore malfunction;

[0013] When I>I max When the main frequency obtained by spectrum analysis of the pressure data matches the pumping unit's stroke frequency, it is determined to be a lifting equipment failure.

[0014] When P < P min The state lasts for more than 2 hours, and <K o The problem was determined to be a malfunction of the lifting equipment.

[0015] When D > D t P < Pz, and S < S u When this occurs, it is determined to be an instrument malfunction;

[0016] When S < S v If the duration is greater than 30 minutes, it is determined to be an instrument malfunction;

[0017] Furthermore, the priority order for handling faults is as follows: safety emergency faults, instrument faults, lifting equipment faults, and wellbore faults.

[0018] S5, perform maintenance operations according to the priority order of fault handling.

[0019] Preferably, the extraction of the deep time-frequency fault feature vector specifically includes the following steps:

[0020] S101 continuously acquires the wellhead pressure data from the equipment to form a raw time-series pressure signal, then performs missing data filling and noise filtering, and finally performs sliding window segmentation;

[0021] S102, then perform time-frequency transformation on the pressure data of each sliding window to obtain the initial time-frequency representation and obtain the initial frequency positioning information of each time-frequency point. Then, extract the time-varying feature parameters of the pressure data at each time and determine the processing parameters at each time. Based on the processing parameters, perform iterative redistribution of multiple energy compression. In each iteration, the time-frequency energy is redistributed along the frequency direction to the corresponding instantaneous frequency position, so that the energy gradually converges. Then, for the converged time-frequency representation, within the local neighborhood limited by the frequency domain search range parameter at each time, the residual energy is further redistributed to the frequency position corresponding to the extreme point. Finally, a high-resolution pressure time-frequency map is output.

[0022] S103, based on the generated high-resolution pressure time-frequency spectrum, extract instantaneous frequency trajectory features, time-frequency energy distribution features, sideband structure features and transient impact features;

[0023] S104 integrates instantaneous frequency trajectory features, time-frequency energy distribution features, sideband structure features, and transient impact features, and uses a deep autoencoder network with attention mechanism for nonlinear dimensionality reduction and feature reconstruction to output a deep time-frequency fault feature vector.

[0024] Preferably, after obtaining the initial time-frequency representation in step 102, the partial derivative along the frequency direction is calculated and divided by the original transform value to obtain the preliminary estimate of the instantaneous frequency at each time-frequency point. Then, the time-varying characteristic parameters of the pressure data at the corresponding time moment are extracted within each window. Based on the absolute value of the first derivative of the instantaneous frequency, the optimal number of iterations is determined. Then, based on the local Rényi entropy minimization criterion, the optimal frequency interval is determined for each time moment. Combined with the optimal number of iterations, multiple iterative redistribution convergences are performed on the initial time-frequency representation. Finally, for the converged time-frequency representation obtained after iterative redistribution, the frequency position corresponding to the maximum modulus of the time-frequency coefficient is searched within the local frequency neighborhood defined by the optimal frequency interval at each time moment. The residual energy of each time-frequency point is further redistributed to the frequency position corresponding to the maximum modulus point in its local neighborhood.

[0025] Preferably, the time-varying characteristic parameters include the instantaneous frequency value and its first derivative, wherein the larger the absolute value of the first derivative of the instantaneous frequency value, the more drastic the frequency change of the pressure data at the corresponding moment, and the smaller the absolute value, the more gradual the frequency change.

[0026] Preferably, the optimal number of iterations is based on the mapping relationship between the absolute value of the first derivative of the instantaneous frequency and the number of iterations. The larger the absolute value of the first derivative of the instantaneous frequency, the more iterations are allocated to focus on time-frequency energy. The smaller the absolute value of the first derivative of the instantaneous frequency, the fewer iterations are allocated to reduce computational resources. The number of iterations ranges from 2 to 10.

[0027] Preferably, when calculating the trend coefficient K, fluctuation index I, abrupt change degree D, stability S and pressure value P of the pressure data, an attention weight regressor is trained by using the depth time-frequency fault feature vector sequence in the historical wellhead pressure data and the corresponding original pressure data trend label. The weighted summation forms a weighted state value sequence that reflects the evolution of the pressure state, and then the trend coefficient K, fluctuation index I, stability S, abrupt change degree D and pressure value P are calculated in sequence.

[0028] Preferably, the slow descent threshold K s The rapid rise threshold K represents the maximum value of the slow downward trend of the pressure state. f The near-zero change threshold K represents the minimum value of the rapid increase in pressure. o The maximum fluctuation threshold I is set to the condition where the pressure state trend remains nearly unchanged. max The minimum fluctuation threshold I represents the maximum value of the pressure state fluctuation. min The threshold value of the sudden change rate D is the minimum value of the pressure state fluctuation. t The state stability threshold S is the minimum value of a sudden pressure drop. u The data stagnation threshold S is the maximum value of the actual pressure data fluctuation compared to the stable pressure data. v The minimum pressure threshold P is the minimum value of normal fluctuations in actual pressure data compared to the stable state pressure data. min The near-zero pressure value P is the minimum permissible pressure required to maintain effective lifting of the equipment. z The maximum safe pressure threshold P is used to determine an extremely low pressure setting value that is close to atmospheric pressure. max The effective lower pressure limit P represents the maximum safe pressure limit for equipment and pipelines. a and the upper limit of effective pressure P b These are the lower and upper limits of the optimal operating pressure range.

[0029] Preferably, the slow descent threshold K s Rapidly rising threshold K f and K o It calculates the distribution of the pressure change rate under normal wellhead operating conditions, where K s Take the 5th percentile on the left side of the distribution, K f Take the 95th percentile on the right side of the distribution. Ko Set it to a range close to 0;

[0030] The maximum fluctuation threshold I max and minimum fluctuation threshold I min It calculates the distribution of pressure fluctuations at the wellhead under normal operating conditions, where I max Take the 95th percentile, I min Take the 5th percentile;

[0031] The threshold of the sudden change rate D t It is the maximum rate of pressure decrease per unit time when the instrument fails;

[0032] The state stability threshold S u During stable production periods, the maximum deviation of pressure data points from their moving average is 1.2 to 1.5 times.

[0033] The data lag threshold S v During stable production periods, pressure data fluctuates within the minimum normal range of 20% to 50%.

[0034] Preferably, the minimum pressure threshold P min This is the minimum wellhead pressure for normal production;

[0035] The near-zero pressure value P z This is the atmospheric pressure value, or 1%–2% of the pressure transmitter's range;

[0036] The maximum safe pressure threshold P max It is 80%–90% of the equipment's rated working pressure or design pressure.

[0037] An Internet of Things (IoT)-based oil equipment operation and maintenance system is used to execute the aforementioned IoT-based oil equipment operation and maintenance method. The operation and maintenance system includes a data acquisition unit, a feature extraction unit, an index calculation unit, a threshold setting unit, a fault diagnosis unit, and a maintenance unit.

[0038] The data acquisition unit is configured at the wellhead of the oil equipment for acquiring wellhead pressure data.

[0039] The feature extraction unit is used to receive wellhead pressure data collected by the data acquisition unit and extract depth-time frequency fault feature vectors.

[0040] The indicator calculation unit calculates the trend coefficient K, fluctuation index I, abrupt change degree D, stability S and pressure value P of the pressure data by receiving the deep fault feature vector;

[0041] The threshold setting unit sets a threshold related to the calculated pressure data based on the design parameters of the oil equipment.

[0042] The fault diagnosis unit combines the data output by the index calculation unit with the threshold set by the threshold setting unit to perform fault type diagnosis.

[0043] The maintenance unit performs corresponding maintenance operations based on the fault type diagnosed by the fault diagnosis unit. The priority order of fault handling configured for the maintenance unit is as follows: safety emergency fault, instrument fault, lifting equipment fault, and well shaft fault.

[0044] The technical effects and advantages of this invention are as follows:

[0045] (1) This invention extracts the deep fault feature vector of wellhead pressure data, calculates the trend coefficient K, fluctuation index I, mutation degree D, stability S and pressure value P of the pressure data, and sets the corresponding thresholds related to the calculated pressure data in combination with the equipment design parameters. By comparing the calculated pressure data with the thresholds, fault diagnosis is performed to identify safety emergency faults, wellbore faults, lifting equipment faults and instrument faults that affect wellhead pressure changes. This allows for a more accurate identification of the fault type, enabling faster and more precise fault location during subsequent maintenance, and reducing unnecessary downtime required for continued fault diagnosis.

[0046] (2) This invention constructs differentiated two-dimensional judgment rules to jointly lock down slow-decreasing wellbore anomalies from five dimensions: trend rate, duration, fluctuation amplitude, mutation intensity and absolute pressure value. This setting avoids unplanned shutdowns caused by slow pressure drop in the wellbore being misjudged as a safety emergency failure. On the other hand, it prevents pressure fluctuations caused by the operation of the pumping unit from masking the real safety risks, thus realizing immediate response to emergency failures and classification and control of early warning of wellbore failures. Attached Figure Description

[0047] Figure 1 This is a flowchart illustrating the implementation steps of the operation method of the present invention.

[0048] Figure 2 This is a block diagram of the system architecture of the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] Example 1: The present invention provides as follows Figure 1The method for operation and maintenance of oil equipment based on the Internet of Things (IoT) includes the following steps:

[0051] S1: Collect wellhead pressure data of oil equipment, process the collected pressure data using time-varying parameter multi-synchronous extrusion transformation to generate high-resolution pressure time-frequency spectrum, and extract depth time-frequency fault feature vector from the time-frequency spectrum;

[0052] Specifically, the extraction of deep time-frequency fault feature vectors includes the following steps:

[0053] S101 continuously acquires wellhead pressure data from the equipment, forming a raw time-series pressure signal. This signal is then filled with missing data and filtered for noise before being segmented into a sliding window. During continuous data acquisition, a high-frequency response pressure transmitter is first installed in the main pipeline or downstream straight section of the throttle valve in the oil production pipeline. Damping pads are used to eliminate mechanical vibration interference, ensuring the original fidelity of the physical signal. In the continuous acquisition phase, the hardware crystal oscillator clock of the data acquisition card triggers uninterrupted scanning, and FIFO buffering ensures that sampling points are buffered and not lost. In the data cleaning phase, invalid frames are identified to identify missing data. A short-time hold strategy is used to handle microsecond-level missing data. For timeout missing data, invalid segments are directly marked and outliers are removed. Low-pass filtering is used to physically filter out circuit noise. During sliding window segmentation, the window length is determined by the lowest frequency component of the wellhead pressure fluctuation, ensuring coverage of the complete stroke cycle and including multiple complete waveforms. After segmentation, the pressure data within each window, along with the start absolute time, well ID, and current operating condition label, are packaged into a structured data block, providing input for subsequent time-frequency analysis.

[0054] S102, Next, a time-frequency transformation is performed on the pressure data of each sliding window to obtain the initial time-frequency representation and the initial frequency positioning information of each time-frequency point. Specifically, a short-time Fourier transform combined with a Gaussian window function is used for the time-frequency transformation to balance the resolution requirements of time and frequency. After the transformation is completed, the partial derivative along the frequency direction is calculated and divided by the original transformed value to obtain the initial frequency positioning of each time-frequency point. The result is used as the preliminary estimate of the instantaneous frequency, which is the actual phase change rate of the signal at that frequency point at that moment. The calculation is expressed as:

[0055] (1);

[0056] in, This represents the frequency variable of the current analysis grid point, that is, the discrete value point on the frequency coordinate axis set when performing the short-time Fourier transform; The initial time-frequency representation result is represented by the complex matrix obtained after the short-time Fourier transform, which contains the amplitude and phase information of that time-frequency point; t represents the time variable, referring to the physical time corresponding to the sliding window currently being analyzed.

[0057] Then, the time-varying characteristic parameters of the pressure data at each moment are extracted, and the processing parameters for each moment are determined. The extracted time-varying characteristic parameters specifically refer to the first derivative of the instantaneous frequency. The magnitude of its absolute value reflects the intensity of frequency modulation of the pressure fluctuation at the current moment. The larger the absolute value, the more drastic the frequency change, corresponding to the transient impact of the fault. The smaller the absolute value, the smoother the change, corresponding to the steady-state periodic component. Based on the relative proportion of this absolute value in all moments, it is mapped to an integer range of 2 to 10 to determine the optimal number of iterations for each moment. At the same time, based on the local Rényi entropy minimization criterion, the candidate frequency interval values ​​are traversed, and the interval that makes the local time-frequency distribution sparsest and the disorder lowest is selected as the optimal frequency interval for the corresponding moment, thus completing the determination of all processing parameters.

[0058] Based on the processing parameters, an iterative redistribution of multiple energy compression is performed. In each iteration, the time-frequency energy is redistributed along the frequency direction to the corresponding instantaneous frequency position, so that the energy gradually converges. In specific execution, the iteration process is driven according to the optimal number of iterations determined at each time. In each iteration, the time-frequency energy scattered on each adjacent frequency point at the current time is moved according to the instantaneous frequency estimate and forcibly superimposed to the actual instantaneous frequency position at that time to eliminate the energy tailing effect caused by the window function. After each iteration, the instantaneous frequency estimate is updated in time and the above compression operation is performed again, thereby gradually shrinking the originally vague and broad frequency band into a sharp and concentrated ridge shape.

[0059] Then, for the converged time-frequency representation, within the local neighborhood defined by the frequency domain search range parameter at each time step, the residual energy is further redistributed to the frequency position corresponding to the extreme point. The frequency domain search range parameter is the optimal frequency interval described above. At each time step, the local neighborhood is defined with the converged instantaneous frequency position as the center and the optimal frequency interval as the radius. The frequency point with the largest modulus of the time-frequency coefficient in the search neighborhood is taken as the extreme point. Then, all the residual energy in the neighborhood at that time step is uniformly superimposed on the single-point frequency position corresponding to the extreme point, and the side lobes and diffusion residues around the ridge are cleared.

[0060] The final output is a high-resolution time-frequency spectrum of pressure. The final output is a high-resolution spectrum of the energy focused on the trajectory line of the real instantaneous frequency at each moment. The frequency ridges of the time-frequency spectrum are clear and sharp, the background noise is suppressed, and the fundamental frequency, harmonic frequency and weak sideband modulation components are distinguished, which can be directly used as the data base for subsequent feature extraction.

[0061] After obtaining the initial time-frequency representation in step 102, the partial derivative along the frequency direction is calculated and divided by the original transform value to obtain the preliminary estimate of the instantaneous frequency at each time-frequency point. Then, the time-varying feature parameters of the pressure data at the corresponding time are extracted in each window. The time-varying feature parameters include the instantaneous frequency value and its first derivative. The larger the absolute value of the first derivative of the instantaneous frequency value, the more drastic the frequency change of the pressure data at the corresponding time, and the smaller the absolute value, the smoother the frequency change.

[0062] The optimal number of iterations is determined based on the absolute value of the first derivative of the instantaneous frequency. Then, based on the local Rényi entropy minimization criterion, the optimal frequency interval is determined for each time step. Combined with the optimal number of iterations, multiple iterations are performed on the initial time-frequency representation to redistribute and converge. The optimal number of iterations is based on the mapping relationship between the absolute value of the first derivative of the instantaneous frequency and the number of iterations. The larger the absolute value of the first derivative of the instantaneous frequency, the more iterations are allocated to focus on time-frequency energy. The smaller the absolute value of the first derivative of the instantaneous frequency, the fewer iterations are allocated to reduce computational resources. The number of iterations ranges from 2 to 10.

[0063] Finally, for the converged time-frequency representation obtained after iterative redistribution, within the local frequency neighborhood defined by the optimal frequency interval at each time point, the frequency position corresponding to the maximum modulus of the time-frequency coefficient is searched, and the residual energy of each time-frequency point is further redistributed to the frequency position corresponding to the maximum modulus in its local neighborhood.

[0064] S103, based on the generated high-resolution pressure time-frequency spectrum, extracts instantaneous frequency trajectory features, time-frequency energy distribution features, sideband structure features, and transient impact features. It should be noted that the instantaneous frequency trajectory features reflect the real-time fluctuations in the rotational speed of rotating equipment or reciprocating pumps; the undulating pattern of the trajectory line corresponds to shaft torsional vibration or the opening and closing rhythm of pumps and valves. The time-frequency energy distribution features reflect the intensity concentration of the pressure signal in different time and frequency intervals; changes in the distribution center of gravity correspond to the transfer of operating load or changes in fluid properties. The sideband structure features reflect the modulation sideband spacing and amplitude on both sides of the fundamental frequency and its harmonics; density and strength correspond to fault types. The transient impact features reflect sudden broadband energy bands appearing in the time-frequency spectrum, corresponding to abnormal events such as sudden pressure rises or falls or particulate impacts.

[0065] In actual extraction operations, the instantaneous frequency trajectory is obtained by searching for the maximum modulus point along the frequency direction at each moment in the time-frequency spectrum and connecting the points. If necessary, a dynamic programming path tracking algorithm is used to avoid frequency jumps. The time-frequency energy distribution is obtained by integrating the square of the amplitude of the time-frequency coefficients within a specified frequency band to obtain the energy change curve over time. The sideband structure is obtained by demodulating and analyzing the local spectrum near the fundamental frequency to extract the sideband spacing and sideband energy ratio. The transient impact characteristics are obtained by using a time-frequency domain joint threshold discrimination method to mark and quantize the time-frequency regions that exceed the background energy mean plus three standard deviations.

[0066] The frequency conversion interval is obtained by subtracting the positions of the spectral peaks symmetrically distributed on both sides of the fundamental frequency and taking the average. The calculation is expressed as follows:

[0067] (2);

[0068] in, It represents the carrier frequency, which corresponds to the frequency value of the central spectral peak with the strongest energy in the time-frequency spectrum, serving as the main modulation element; This indicates the frequency of the right-side spectral peak, which is higher than the carrier frequency. The frequency value corresponding to the first significant modulation sideband spectral peak; This indicates the frequency of the left-side spectral peak, which is below the carrier frequency. The frequency value corresponding to the first significant modulation sideband spectral peak.

[0069] S104 fuses instantaneous frequency trajectory features, time-frequency energy distribution features, sideband structure features, and transient impact features. During the fusion operation, the four types of features are first normalized to eliminate dimensional differences, and then they are aligned point by point on the time axis to form a multi-dimensional feature matrix. Each time point corresponds to a joint feature vector containing the four types of feature values, thereby preserving the synchronous correlation between the features on the time series. This fusion matrix also contains amplitude modulation information and frequency modulation information. The amplitude modulation information comes from energy distribution and transient impact, and the frequency modulation information comes from frequency trajectory and sideband structure.

[0070] A deep autoencoder network with an attention mechanism is used for nonlinear dimensionality reduction and feature reconstruction. The purpose of introducing the attention mechanism is to allow the network to automatically determine the contribution weights of different time points and different feature dimensions to the final fault diagnosis. When the frequency trajectory fluctuates abnormally sharply but the energy change is stable in certain time periods, the attention weights will automatically focus on the frequency trajectory dimension and suppress the energy dimension, and vice versa, thus avoiding the blindness of manually setting weights based on experience. In actual operation, a self-attention module is embedded in the encoding stage of the autoencoder to calculate the correlation weight matrix between each dimension of the fused feature matrix, which is expressed as follows:

[0071] (3);

[0072] Where Q represents the query matrix, which is obtained by linear transformation of the input fused feature matrix, and has a dimension of n×d. k K represents the key matrix, which is obtained from the input fused feature matrix through another linear transformation, and has a dimension of n×d. k , as the object of the matching search for matrix Q; T represents the matrix transpose symbol, K T This indicates that the rows and columns of the key matrix K are interchanged; dk The dimension of the key vector is represented by the feature length of each row of the query matrix Q and the key matrix K. The encoder part of the autoencoder network gradually compresses the high-dimensional fusion features into a low-dimensional bottleneck space through multiple fully connected layers. The decoder part reconstructs the original fusion features from the bottleneck vector. During training, the goal is to minimize the reconstruction error, forcing the bottleneck layer to retain the core feature components that are most representative and physically distinguishable, while discarding redundant noise information that is unrelated to the fault.

[0073] The output deep time-frequency fault feature vector is the low-dimensional encoding result of the bottleneck layer. Its dimension is much lower than that of the original fused features, but it condenses the time-frequency physical information in the pressure signal that is closely related to the health status of the equipment. This vector can be directly input into the classifier for fault mode recognition, or it can be used as a state baseline to compare similarity with historical data to evaluate the current operational health and abnormal evolution trend of the wellhead equipment.

[0074] S2, based on the pressure time-frequency spectrum and deep time-frequency fault feature vector, calculates the trend coefficient K, fluctuation index I, abrupt change degree D, stability S and pressure value P of the pressure data. The trend coefficient K is used to describe the long-term evolution direction and speed, the fluctuation index I is used to describe the degree of short-term random fluctuations, the abrupt change degree D is used to describe the intensity of sudden transient changes, the stability S is used to describe the comprehensive ability of the system to maintain stable operation, and the pressure value P is used to describe the absolute dimension of the current overall level.

[0075] When calculating the trend coefficient K, fluctuation index I, abrupt change D, stability S and pressure value P of pressure data, an attention weight regressor is trained by using the depth time-frequency fault feature vector sequence in historical wellhead pressure data and the corresponding original pressure data trend label. The weighted summation forms a weighted state value sequence that reflects the evolution of pressure state. Then, the trend coefficient K, fluctuation index I, stability S, abrupt change D and pressure value P are calculated in sequence.

[0076] Specifically, when calculating the trend coefficient K, fluctuation index I, abrupt change degree D, stability S, and pressure value P of pressure data, the deep fault feature vector sequence and the corresponding original pressure data trend labels from historical wellhead pressure data are first retrieved to form a dataset. The attention weight regressor is trained using the dataset to extract the contribution weight of each feature dimension to trend judgment and form a feature weight vector. During the training process, the attention weight regressor uses minimizing the trend label prediction error as the loss function and iteratively updates the attention weight of each feature dimension through backpropagation. The larger the value of each element in the final weight vector, the more discriminative the corresponding feature dimension is in describing changes in pressure state. Conversely, a smaller value indicates that the corresponding dimension is less correlated with trend evolution or contains more redundant information.

[0077] Then, the feature weight vector is used to perform a weighted summation on each feature vector in the sequence, and the summation is converted into a weighted state value sequence reflecting the pressure state;

[0078] The slope of change is obtained by fitting the weighted state value sequence with a univariate linear regression. The slope of change is then divided by the range of the weighted state value sequence. If the range is zero, the trend coefficient K is directly defined. The sign of the calculated trend coefficient K reflects the overall upward or downward trend of the pressure state, and the absolute value reflects the relative magnitude of the trend change.

[0079] Calculate the average absolute deviation of the weighted state value sequence, and then divide it by the range of the weighted state value sequence. If the range is zero, then I is directly defined as zero. The fluctuation index I is used to measure the average dispersion of the state values ​​at each time point in the sequence relative to the mean. The larger the value, the more violent the pressure fluctuation, reflecting the instability of the fluid or the aggravation of mechanical vibration. A value of zero indicates that the state at each time point is completely consistent and is in an ideal steady state.

[0080] Calculate the instantaneous rate of change sequence of the weighted state value sequence and find the maximum value. Subtract the maximum value from the median of the entire instantaneous rate of change sequence. Divide the difference by the range of the instantaneous rate of change sequence. If the range is zero, then define D as zero. The abruptness degree D is used to capture the largest abnormal jump in the instantaneous rate of change sequence that deviates from the normal range, and quantifies the intensity of sudden pressure spikes or drops. The larger the value, the more significant the abrupt event.

[0081] The volatility index I and the mutation degree D are weighted and fused to obtain the comprehensive instability. The comprehensive instability is then subtracted from 1 to obtain the statistical and trend stability S. The stability S ranges from 0 to 1. The closer it is to 1, the more stable the operation is. The closer it is to 0, the more prominent the unstable factors are. It serves as a convenient reference for evaluating the overall health.

[0082] The pressure value P is obtained by summing all weighted state values ​​in the weighted state value sequence and dividing by the sequence length. Unlike the physical average pressure value directly collected by the original sensor, the pressure value P calculated by this scheme has undergone deep time-frequency feature extraction and attention weighted mapping, representing the effective pressure component related to the health of equipment operation. It can be directly compared with the P value under the same historical working conditions to determine whether the current overall state deviates from the normal benchmark.

[0083] S3, based on the design parameters of the oil equipment, sets the slow descent threshold K respectively. s Rapidly rising threshold K f Threshold K for near-zero change o Maximum fluctuation threshold I max Minimum fluctuation threshold I min Threshold for sudden rate of change D t State stability threshold Su Data lag threshold S v Minimum pressure threshold P min Approaching zero pressure value P z Maximum safe pressure threshold P max Effective pressure lower limit P a and the upper limit of effective pressure P b ;

[0084] Slow descent threshold K s The maximum value of the slow downward trend of the pressure state, and the rapid rise threshold K. f The minimum value of the rapid increase in pressure, the near-zero change threshold K. o As the pressure state trend remains nearly unchanged, K s K f and K o The maximum fluctuation threshold I was determined by analyzing the statistical distribution of the pressure trend coefficient K under a large number of historical normal operating conditions. max The maximum value of the pressure state fluctuation, and the minimum fluctuation threshold I. min The minimum value of pressure state fluctuation, the threshold value of sudden change rate D t The minimum value of the sudden pressure drop, the state stability threshold S u The data stagnation threshold S represents the maximum fluctuation of the actual pressure data relative to the stable pressure data. v The minimum pressure threshold P represents the minimum normal fluctuation of actual pressure data relative to a stable pressure state. min To maintain the minimum permissible pressure value required for effective lifting of the equipment, close to the zero pressure value P z The maximum safe pressure threshold P is used to determine the extremely low pressure setting value where the pressure is close to atmospheric pressure. max The upper limit of the safe pressure that the equipment and pipelines can withstand, and the lower limit of the effective pressure P. a and the upper limit of effective pressure P b The lower and upper limits of the optimal operating pressure range;

[0085] Slow descent threshold K s Rapidly rising threshold K f and K o It calculates the distribution of the pressure change rate under normal wellhead operating conditions, where K s Take the 5th percentile on the left side of the distribution, K f Take the 95th percentile on the right side of the distribution. K o Set it to a range close to 0;

[0086] Maximum fluctuation threshold I max and minimum fluctuation threshold I minIt calculates the distribution of pressure fluctuations at the wellhead under normal operating conditions, where I max Take the 95th percentile, I min Take the 5th percentile;

[0087] Threshold for sudden rate of change D t D is the maximum rate of pressure decrease per unit time when the instrument malfunctions. t The pressure drop response rate is determined based on the typical pressure drop response rate when the instrument pressure tapping line is damaged or the transmitter diaphragm fails. It is obtained through simulation experiments or statistical analysis of historical instrument failure cases and is used to distinguish between the actual pressure drop in the wellbore and the abnormal sudden drop caused by instrument sensing failure.

[0088] State stability threshold S u It is 1.2-1.5 times the maximum deviation between the pressure data point and its moving average during a stable production period, used to determine whether the current pressure status is within the normal fluctuation range allowed for stable operation;

[0089] Data lag threshold S v During stable production periods, pressure data fluctuations are typically within the minimum range of 20% to 50%. This is used to identify data stagnation. When the absolute value of the rate of change between adjacent pressure data points remains below S... v If the data change is too small, it may be due to insufficient resolution caused by an oversized instrument range, signal transmission interruption causing numerical freezing, or A / D conversion failure causing numerical locking. Even if the pressure value P is within the normal range at this time, it should still be judged as abnormal instrument data.

[0090] Minimum pressure threshold P min This is the minimum wellhead pressure for normal production;

[0091] Approaching zero pressure value P z The pressure value is atmospheric pressure, or 1%–2% of the pressure transmitter's range, used to determine whether the pressure has dropped to near the ambient atmospheric pressure level.

[0092] Maximum safe pressure threshold P max It is 80%-90% of the equipment's rated working pressure or design pressure, used to cope with short-term pressure fluctuation overshoot without triggering false alarms.

[0093] S4, combining the pressure data calculated by S2 and the threshold set by S3 to perform fault diagnosis;

[0094] When P > P max and / or K > K fWhen a pressure value exceeds the limit, it is determined to be a safety emergency fault. This determination has the highest priority. The pressure value exceeds the limit, which means that the equipment is facing a direct threat of physical damage or even blowout and loss of control. The rapid rise in pressure means that some kind of blockage or pressure accumulation is rapidly deteriorating. The safety emergency fault is triggered as soon as any condition is met, without waiting for other indicators to be verified, ensuring timely response.

[0095] When K < K s The downward trend lasts for more than 24 hours, and I < I max D < D t And P < P max When the condition is determined to be a wellbore failure, it indicates a slow-changing process of deteriorating flow conditions inside the wellbore. When the trend coefficient K remains below the slow-decline threshold, it indicates a stable negative evolution. A duration exceeding 24 hours excludes interference from short-term disturbances and measurement noise. The fluctuation index I and abrupt change degree D are both within the non-abnormal range, indicating that the decline is not caused by equipment failure or operational abrupt changes. The pressure value is below P. max After ensuring that the descent has not yet reached the top, and considering all factors, the problem is determined to be an increase in wellbore flow resistance due to wax buildup, scaling, or near-wellbore blockage, requiring downhole maintenance operations to be arranged.

[0096] When I>I max When the main frequency obtained by spectral analysis of the pressure data matches the pumping unit stroke frequency, it is determined to be a lifting equipment failure. This indicates that the operating parameters of the lifting equipment do not match the wellbore fluid supply capacity, resulting in periodic pressure pulsation. This is caused by improper stroke and stroke settings, imbalance, or decreased pump efficiency, and needs to be resolved by adjusting the lifting parameters.

[0097] When P < P min The state lasts for more than 2 hours, and <K o If the lifting equipment is identified as malfunctioning, it means that the energy currently provided by the lifting equipment is insufficient to overcome the gravity pressure difference and flow resistance of the well fluid, but it has not yet completely failed. It is determined that the pump efficiency is insufficient or the fluid supply capacity has decreased, and the lifting capacity needs to be restored by checking the pump or adjusting the parameters.

[0098] When D > D t P < Pz, and S < S u If the problem is identified as an instrument malfunction, priority should be given to checking the pressure transmitter, pressure lines, signal transmission, and A / D conversion module.

[0099] When S < S vIf the duration is greater than 30 minutes, it is determined to be an instrument failure, indicating that the pressure data has almost no change or extremely low change in the time series, which violates the physical law that there must be small disturbances in the real fluid system. At the same time, the duration of more than 30 minutes excludes short-term static conditions, and it is determined to be a measurement failure caused by instrument signal freezing or insufficient resolution. It is necessary to check whether the instrument is stuck or whether the signal link is interrupted.

[0100] The priority order for handling faults is as follows: safety emergency faults, instrument faults, lifting equipment faults, and shaft faults. Safety emergency faults involve equipment integrity and personal safety, so they are set as the highest priority. Instrument faults take precedence over lifting and shaft faults because instrument data is the basis for all subsequent diagnostic judgments; if the instruments fail, other diagnostic results are unreliable. Lifting equipment faults take precedence over shaft faults because lifting equipment directly affects current production, and its problems are often sudden. Shaft faults are mostly slow-deterioration types, and maintenance can be arranged in an orderly manner after the first two types of problems are resolved.

[0101] S5, perform maintenance operations according to the priority order of fault handling;

[0102] Specifically, maintenance and repair procedures for safety emergencies include the following:

[0103] When P > P max When the pressure at the wellhead equipment exceeds the limit, the system will automatically perform a remote emergency shutdown and / or safety pressure relief, and notify safety personnel to confirm and handle the situation on-site. The remote emergency shutdown will stop the pressure from rising further by cutting off the drive source. The safety pressure relief will release the pressure to the flare system or drain tank by opening the vent valve or bypass valve. Notifying safety personnel to confirm on-site is to ensure that the automatic action is effectively executed and to investigate the root cause of the overpressure.

[0104] When K > K f When the pressure at the wellhead equipment is rapidly increasing and approaching the danger threshold, remote throttling control is automatically executed to suppress the pressure rise. At this time, the operation and maintenance personnel are notified to check the status of the process valves and the function of the downhole safety valves. The remote throttling control increases the flow resistance by finely adjusting the opening of the flow valve to suppress the rate of pressure rise. At the same time, the operation and maintenance personnel are notified to check whether the surface process valves are abnormally closed and whether the downhole safety valves are malfunctioning, to prevent the situation from developing into an overpressure accident.

[0105] The maintenance and troubleshooting procedures for instrument malfunctions include the following:

[0106] When D > D t P < P z And S < S uWhen this occurs, it indicates a sudden drop in pressure data accompanied by abnormal fluctuations, suggesting possible instrument failure or a serious pipeline leak. In this case, immediately notify instrument personnel to check the pressure transmitter, pressure tapping lines, and the sealing of connections, and perform calibration or replacement. After triggering, instrument inspection personnel are required to arrive at the site within 30 minutes, prioritizing checks for blockages or leaks in the pressure tapping lines, damage to the transmitter diaphragm, and whether the circuit output is normal. If the transmitter is functioning normally and the pressure tapping lines are unobstructed, further investigation is needed to determine if there is an actual leak in the wellhead pipeline.

[0107] When S < S v If the duration of the delay exceeds 30 minutes, it indicates that the pressure data has not changed for an extended period, which is considered a data stagnation. In this case, instrumentation personnel should be notified to check whether the instrument is stuck, whether the signal transmission is interrupted, and to calibrate or replace the instrument. The data stagnation response requires instrumentation personnel to check whether the transmitter output is frozen at a certain value, whether there is an open circuit or short circuit in the signal transmission cable, and whether the data acquisition card channel is damaged. For intelligent transmitters, it is also necessary to check whether the internal software is in a watchdog reset state or whether the configuration parameters are lost. After calibration or replacement, the data authenticity must be verified by comparing it with the local pressure gauge.

[0108] Maintenance and troubleshooting procedures for lifting equipment malfunctions include the following:

[0109] When I>I max When the main frequency obtained by spectrum analysis of the pressure data matches the frequency of the pumping unit stroke, it indicates that the lifting equipment is not operating properly and is causing periodic pulsations. At this time, a parameter adjustment work order is generated to guide the on-site adjustment of the pumping unit stroke and frequency or to perform balance correction.

[0110] When P < P min The state lasts for more than 2 hours, and |K| < K o When the pressure remains low and shows no upward trend, it is determined that the pump efficiency is insufficient. At this time, a pump inspection work order is generated to arrange pump condition diagnosis, current test, and plan pump inspection or maintenance work.

[0111] Maintenance and repair procedures for wellbore malfunctions include the following:

[0112] When K < K s The downward trend lasts for more than 24 hours, and I < I min D < D t And P < P max When the pressure continues to drop slowly, it indicates that there is a risk of wax buildup, scaling, or blockage in the near-wellbore area. At this time, a downhole maintenance work order is generated to arrange well washing, chemical unblocking, or wireline operation to inspect downhole tools.

[0113] Example 2: The present invention provides as follows Figure 2The illustrated petroleum equipment operation and maintenance system is used to execute the petroleum equipment operation and maintenance method based on the Internet of Things in Embodiment 1. The operation and maintenance system includes a data acquisition unit, a feature extraction unit, an index calculation unit, a threshold setting unit, a fault diagnosis unit, and a maintenance unit.

[0114] The data acquisition unit is configured at the wellhead of the oil equipment for acquiring wellhead pressure data. It performs uninterrupted scanning by triggering a hardware clock, marking each sampling point with an absolute timestamp, and enabling breakpoint continuation through a circular buffer. The raw pressure sequence formed by the acquisition is filled with missing data and filtered for noise in the edge terminal, and then divided into standardized data blocks with time labels, well IDs, and operating condition labels by a sliding window, and then passed up to the feature extraction unit.

[0115] The feature extraction unit receives wellhead pressure data acquired by the data acquisition unit and extracts deep time-frequency fault feature vectors. First, it performs time-frequency transformation on the data in each window to obtain an initial time-frequency representation and acquire initial frequency location information. After extracting time-varying feature parameters to determine the processing parameters at each moment, it performs multiple energy squeezing iterations to gradually converge the time-frequency energy along the frequency direction. Then, it redistributes the residual energy in the local neighborhood at each moment, finally outputting a high-resolution time-frequency map. Based on the time-frequency map, it extracts four types of physical features: instantaneous frequency trajectory, time-frequency energy distribution, sideband structure, and transient impact. After fusion, these features are nonlinearly reduced in dimensionality and reconstructed using a deep autoencoder network with an attention mechanism, outputting the deep time-frequency fault feature vector to the index calculation unit.

[0116] The indicator calculation unit calculates the trend coefficient K, fluctuation index I, mutation degree D, stability S, and pressure value P of the pressure data by receiving the deep fault feature vector. Based on historical data, it trains an attention weight regressor, weights and sums the feature vectors at each time point to form a weighted state value sequence, and then calculates the five indicators in sequence to characterize the long-term evolution direction of the pressure state, the degree of short-term fluctuation, the intensity of sudden transients, the ability of the system to maintain stability, and the current overall level. After completion, the five indicators are packaged and output to the fault diagnosis unit.

[0117] The threshold setting unit sets thresholds related to the calculated pressure data based on the design parameters of the oil equipment; these include trend thresholds, fluctuation thresholds, stability thresholds, and pressure thresholds. All thresholds are adjusted based on the equipment's rated parameters, historical normal operating condition statistical distribution, and field experience. They are injected before operation and can be dynamically corrected during operation.

[0118] The fault diagnosis unit combines the data output by the indicator calculation unit with the thresholds set by the threshold setting unit to perform fault type diagnosis. Faults are determined hierarchically according to priority: the highest priority is safety emergency faults, the second priority is instrument faults, the third priority is lifting equipment faults, and the fourth priority is well shaft faults. When multiple faults are triggered simultaneously, they are categorized and output according to priority.

[0119] The maintenance unit executes corresponding maintenance operations based on the fault type diagnosed by the fault diagnosis unit. The priority order for fault handling configured in the maintenance unit is as follows: safety emergency faults, instrument faults, lifting equipment faults, and wellbore faults. Safety emergency faults trigger remote emergency shutdown or throttling control; instrument faults trigger transmitter and pressure line inspection and calibration; lifting equipment faults generate parameter adjustment or pump inspection work orders; and wellbore faults generate well cleaning, unblocking, or wireline operation work orders. All operations are scheduled and executed according to priority. Upon completion, feedback is sent to the diagnostic unit to trigger a re-diagnosis. Work orders are pushed to personnel handheld terminals via the Internet of Things and simultaneously recorded in the maintenance log.

[0120] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for the operation and maintenance of oil equipment based on the Internet of Things, characterized in that, The following steps are included: S1: Collect wellhead pressure data of oil equipment, process the collected pressure data using time-varying parameter multi-synchronous extrusion transformation to generate high-resolution pressure time-frequency spectrum, and extract depth time-frequency fault feature vector from the time-frequency spectrum; S2, based on the pressure time-frequency spectrum and the deep time-frequency fault feature vector, calculates the trend coefficient K, fluctuation index I, abrupt change degree D, stability S and pressure value P of the pressure data; S3, based on the design parameters of the oil equipment, sets the slow descent threshold K respectively. s Rapidly rising threshold K f Threshold K for near-zero change o Maximum fluctuation threshold I max Minimum fluctuation threshold I min Threshold for sudden rate of change D t State stability threshold S u Data lag threshold S v Minimum pressure threshold P min Approaching zero pressure value P z Maximum safe pressure threshold P max Effective pressure lower limit P a and the upper limit of effective pressure P b ; S4, combining the pressure data calculated by S2 and the threshold set by S3 to perform fault diagnosis; When P > P max and / or K > K f When this occurs, it is determined to be a safety emergency. When K < K s The downward trend lasts for more than 24 hours, and I < I max D < D t And P < P max At that time, it was determined to be a wellbore malfunction; When I>I max When the main frequency obtained by spectrum analysis of the pressure data matches the pumping unit's stroke frequency, it is determined to be a lifting equipment failure. When P < P min The state lasts for more than 2 hours, and <K o The problem was determined to be a malfunction of the lifting equipment. When D > D t P < Pz, and S < S u When this occurs, it is determined to be an instrument malfunction; When S < S v If the duration is greater than 30 minutes, it is determined to be an instrument malfunction; Furthermore, the priority order for handling faults is as follows: safety emergency faults, instrument faults, lifting equipment faults, and wellbore faults. S5, perform maintenance operations according to the priority order of fault handling.

2. The method for operation and maintenance of petroleum equipment based on the Internet of Things according to claim 1, characterized in that, The extraction of the deep time-frequency fault feature vector specifically includes the following steps: S101 continuously acquires the wellhead pressure data from the equipment to form a raw time-series pressure signal, then performs missing data filling and noise filtering, and finally performs sliding window segmentation; S102, then perform time-frequency transformation on the pressure data of each sliding window to obtain the initial time-frequency representation and obtain the initial frequency positioning information of each time-frequency point. Then, extract the time-varying feature parameters of the pressure data at each time and determine the processing parameters at each time. Based on the processing parameters, perform iterative redistribution of multiple energy compression. In each iteration, the time-frequency energy is redistributed along the frequency direction to the corresponding instantaneous frequency position, so that the energy gradually converges. Then, for the converged time-frequency representation, within the local neighborhood limited by the frequency domain search range parameter at each time, the residual energy is further redistributed to the frequency position corresponding to the extreme point. Finally, a high-resolution pressure time-frequency map is output. S103, based on the generated high-resolution pressure time-frequency spectrum, extract instantaneous frequency trajectory features, time-frequency energy distribution features, sideband structure features and transient impact features; S104 integrates instantaneous frequency trajectory features, time-frequency energy distribution features, sideband structure features, and transient impact features, and uses a deep autoencoder network with attention mechanism for nonlinear dimensionality reduction and feature reconstruction to output a deep time-frequency fault feature vector.

3. The method for operation and maintenance of petroleum equipment based on the Internet of Things according to claim 2, characterized in that, After obtaining the initial time-frequency representation in step 102, the partial derivative along the frequency direction is calculated and divided by the original transform value to obtain the preliminary estimate of the instantaneous frequency at each time-frequency point. Then, the time-varying characteristic parameters of the pressure data at the corresponding time moment are extracted within each window. Based on the absolute value of the first derivative of the instantaneous frequency, the optimal number of iterations is determined. Then, based on the local Rényi entropy minimization criterion, the optimal frequency interval is determined for each time moment. Combined with the optimal number of iterations, multiple iterative redistribution convergences are performed on the initial time-frequency representation. Finally, for the converged time-frequency representation obtained after iterative redistribution, the frequency position corresponding to the maximum modulus of the time-frequency coefficient is searched within the local frequency neighborhood defined by the optimal frequency interval at each time moment. The residual energy of each time-frequency point is further redistributed to the frequency position corresponding to the maximum modulus point in its local neighborhood.

4. The method for operation and maintenance of petroleum equipment based on the Internet of Things according to claim 3, characterized in that, The time-varying characteristic parameters include the instantaneous frequency value and its first derivative. The larger the absolute value of the first derivative of the instantaneous frequency value, the more drastic the frequency change of the pressure data at the corresponding moment, and the smaller the absolute value, the smoother the frequency change.

5. The method for operation and maintenance of petroleum equipment based on the Internet of Things according to claim 3, characterized in that, The optimal number of iterations is based on the mapping relationship between the absolute value of the first derivative of the instantaneous frequency and the number of iterations. The larger the absolute value of the first derivative of the instantaneous frequency, the more iterations are allocated to focus on time-frequency energy. The smaller the absolute value of the first derivative of the instantaneous frequency, the fewer iterations are allocated to reduce computational resources. The number of iterations ranges from 2 to 10.

6. The method for operation and maintenance of petroleum equipment based on the Internet of Things according to claim 1, characterized in that, When calculating the trend coefficient K, fluctuation index I, abrupt change D, stability S, and pressure value P of pressure data, an attention weight regressor is trained by using the depth time-frequency fault feature vector sequence in historical wellhead pressure data and the corresponding original pressure data trend labels. The weighted summation forms a weighted state value sequence that reflects the evolution of pressure state. Then, the trend coefficient K, fluctuation index I, stability S, abrupt change D, and pressure value P are calculated sequentially.

7. The method for operation and maintenance of petroleum equipment based on the Internet of Things according to claim 1, characterized in that, The slow-decreasing threshold K s The rapid rise threshold K represents the maximum value of the slow downward trend of the pressure state. f The near-zero change threshold K represents the minimum value of the rapid increase in pressure. o The maximum fluctuation threshold I is set to the condition where the pressure state trend remains nearly unchanged. max The minimum fluctuation threshold I represents the maximum value of the pressure state fluctuation. min The threshold value of the sudden change rate D is the minimum value of the pressure state fluctuation. t The state stability threshold S is the minimum value of a sudden pressure drop. u The data stagnation threshold S is the maximum value of the actual pressure data fluctuation compared to the stable pressure data. v The minimum pressure threshold P is the minimum value of normal fluctuations in actual pressure data compared to the stable state pressure data. min The near-zero pressure value P is the minimum permissible pressure required to maintain effective lifting of the equipment. z The maximum safe pressure threshold P is used to determine an extremely low pressure setting value that is close to atmospheric pressure. max The effective lower pressure limit P represents the maximum safe pressure limit for equipment and pipelines. a and the upper limit of effective pressure P b These are the lower and upper limits of the optimal operating pressure range.

8. The method for operation and maintenance of oil equipment based on the Internet of Things according to claim 7, characterized in that, The slow-decreasing threshold K s Rapidly rising threshold K f and K o It calculates the distribution of the pressure change rate under normal wellhead operating conditions, where K s Take the 5th percentile on the left side of the distribution, K f Take the 95th percentile on the right side of the distribution, K o Set it to a range close to 0; The maximum fluctuation threshold I max and minimum fluctuation threshold I min It calculates the distribution of pressure fluctuations at the wellhead under normal operating conditions, where I max Take the 95th percentile, I min Take the 5th percentile; The threshold of the sudden change rate D t It is the maximum rate of pressure decrease per unit time when the instrument fails; The state stability threshold S u During stable production periods, the maximum deviation of pressure data points from their moving average is 1.2 to 1.5 times. The data lag threshold S v During stable production periods, pressure data fluctuates within the minimum normal range of 20% to 50%.

9. A method for the operation and maintenance of petroleum equipment based on the Internet of Things according to claim 8, characterized in that, The minimum pressure threshold P min This is the minimum wellhead pressure for normal production; The near-zero pressure value P z The pressure value is atmospheric pressure, or 1%–2% of the pressure transmitter's range. The maximum safe pressure threshold P max It is 80%–90% of the equipment's rated working pressure or design pressure.

10. An Internet of Things (IoT)-based oil equipment operation and maintenance system, used to execute the IoT-based oil equipment operation and maintenance method according to any one of claims 1-9, characterized in that, The operation and maintenance system includes a data acquisition unit, a feature extraction unit, an indicator calculation unit, a threshold setting unit, a fault diagnosis unit, and a maintenance unit. The data acquisition unit is configured at the wellhead of the oil equipment for acquiring wellhead pressure data. The feature extraction unit is used to receive wellhead pressure data collected by the data acquisition unit and extract depth-time frequency fault feature vectors. The indicator calculation unit calculates the trend coefficient K, fluctuation index I, abrupt change degree D, stability S and pressure value P of the pressure data by receiving the deep fault feature vector; The threshold setting unit sets a threshold related to the calculated pressure data based on the design parameters of the oil equipment. The fault diagnosis unit combines the data output by the index calculation unit with the threshold set by the threshold setting unit to perform fault type diagnosis. The maintenance unit performs corresponding maintenance operations based on the fault type diagnosed by the fault diagnosis unit. The priority order of fault handling configured for the maintenance unit is as follows: safety emergency fault, instrument fault, lifting equipment fault, and well shaft fault.

Images