Intelligent gas self-closing valve with air tightness detection

Abstract

The present application belongs to the technical field of gas safety equipment detection, and provides an intelligent gas self-closing valve with air tightness detection, which comprises a detection triggering module, a pressure compensation module, a feature extraction module and an analysis and prediction module. The gas consumption in a historical period is collected and analyzed to mark a detection period and a working period, and it is judged whether to trigger air tightness detection. If triggered, the cavity temperature of the detection cavity is collected and it is judged whether to enter a thermal steady state. If entered, the key reference data is determined. The cavity pressure is collected in real time and the actual compensation pressure is calculated through compensation. The time sequence pressure sequence is generated. The time domain analysis technology based on sliding window is adopted to extract the leakage feature of the time sequence pressure sequence, to obtain the trend slope characteristic value, to generate the adaptive alarm threshold, to judge the air tightness state in the detection cavity in combination with the trend slope characteristic value, to judge whether to trigger the prediction analysis, and if triggered, to construct the trend prediction model to predict and analyze the air tightness state of the working period.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of gas safety equipment detection, in particular to an intelligent gas self-closing valve with gas tightness detection. BACKGROUND

[0002] In the field of gas safety monitoring, the gas tightness detection of the intelligent gas self-closing valve is a key link to prevent gas leakage.

[0003] The traditional detection method gradually exposes many defects in actual application. On the one hand, the detection period is fixed and does not match the user's habits. The existing technology usually detects the gas tightness at a preset fixed time, but the user's gas use scenario has randomness. The fixed period may be misaligned with the actual no-flow period, causing the pipe to still have gas flowing during detection, which interferes with the detection result. Even if the detection period covers the peak use period, it may miss the real leakage risk. Environmental temperature interference can easily cause pressure misjudgment. The gas pressure in the gas pipeline is significantly affected by temperature. The existing detection technology often ignores the interference of temperature fluctuations on pressure data and directly uses the measured pressure change to judge leakage, which may cause false positives or false negatives due to factors such as day-night temperature difference and indoor temperature change.

[0004] On the other hand, the interference of sensor noise and mechanical disturbance cannot be eliminated. During the detection process, short-term disturbances such as pipe vibration and sensor noise can cause fluctuations in pressure data. The existing feature extraction method lacks an anti-interference mechanism and cannot distinguish between real leakage trends and interference signals, resulting in low alarm reliability. The passive detection mode and trend prediction are missing. The traditional solution can only make real-time judgments within the detection period and cannot predict future gas tightness trends based on historical data. The fixed alarm threshold cannot adapt to small pressure drifts in different environments, which may miss slow leaks due to threshold rigidity or may cause frequent false positives due to environmental disturbances.

[0005] To solve the above problems, the application provides an intelligent gas self-closing valve with gas tightness detection. SUMMARY

[0006] To make up for the shortcomings of the prior art and solve at least one technical problem in the background art.

[0007] The technical solution adopted by the application to solve the technical problems is an intelligent gas self-closing valve with gas tightness detection, which includes:

[0008] A detection triggering module: set an analysis period and divide it into analysis sub-periods, collect and analyze the gas consumption in each analysis sub-period in the historical period, mark the detection period and the working period, and determine whether to trigger the gas tightness detection;

[0009] Pressure compensation module: if the air tightness detection is triggered, the cavity temperature of the detection cavity is collected and it is judged whether to enter the thermal stable state, if yes, the key reference data is determined, the cavity pressure is collected in real time and the actual compensation pressure is calculated according to the key reference data, and the time sequence pressure sequence is generated;

[0010] Feature extraction module: the time domain analysis technology based on sliding window is adopted to extract the leakage feature of the time sequence pressure sequence, and the trend slope characteristic value of the current time is obtained;

[0011] Analysis and prediction module: the adaptive alarm threshold is generated based on the trend slope characteristic value without leakage in the historical period, the air tightness state in the detection cavity is judged combined with the trend slope characteristic value, and it is judged whether to trigger the prediction analysis, if yes, the trend prediction model is constructed to predict and analyze the air tightness state in the working period;

[0012] The way of judging whether to trigger the air tightness detection is:

[0013] The period analysis cycle is set, and each period analysis cycle in the historical period is evenly divided into analysis sub-periods;

[0014] The gas consumption in each analysis sub-period is recorded by the gas meter and analyzed, and the preselected detection period is marked, if the absolute value difference between the start time in the same period analysis cycle meets the condition, it belongs to the same time group, the time group with the most start time is selected, and the data of all start times in the time group and the corresponding preselected detection period are processed respectively, and the detection period is marked;

[0015] If the current time is the start time of the detection period and the intelligent gas self-closing valve is closed, the air tightness detection is triggered, if not, it waits to be closed, and the start time of the detection period is changed to the closing time, and the air tightness detection is triggered;

[0016] The marking way of the preselected detection period is:

[0017] If the gas consumption in the analysis sub-period is 0, the analysis sub-period is marked as a no-flow period, the no-flow periods in the same period analysis cycle are analyzed, if the time sequence adjacent analysis sub-periods are all no-flow periods, the two analysis sub-periods are combined into one no-flow period, and the no-flow periods in each period analysis cycle are obtained, and the longest no-flow period in the same period analysis cycle is compared, and the longest no-flow period is marked as the preselected detection period;

[0018] The way of obtaining the key reference data is:

[0019] After the intelligent gas self-closing valve is closed, a closed detection cavity is formed in the downstream pipeline, a temperature sensor is used to monitor the cavity temperature in the detection cavity in real time, and the time interval between adjacent sampling points is a sampling period;

[0020] The cavity temperature change rate is calculated based on the cavity temperature difference between adjacent sampling points and the sampling period length, and if the temperature change rate meets the preset change rate threshold condition within a continuous preset time length, it is determined that the detection cavity reaches a thermal initial steady state, the instant when the thermal initial steady state is reached is recorded as a reference time, and key reference data including a reference temperature and an initial pressure are synchronously collected at the reference time;

[0021] The acquisition method of the time sequence pressure sequence is as follows:

[0022] The cavity temperature and the cavity pressure in the detection cavity are collected and matched according to the time stamp to obtain a synchronous data pair in the detection cavity at the sampling point, the cavity temperature in the synchronous data pair is converted into a Kelvin temperature, a standardized pressure compensation formula is derived based on the isochoric process of the ideal gas state equation, the cavity pressure in the synchronous data pair is compensated, and the actual compensation pressure is calculated to obtain the time sequence pressure sequence.

[0023] The acquisition method of the trend slope characteristic value is as follows:

[0024] The time sequence pressure sequence is obtained, a sliding window is set for the time sequence pressure sequence, the time sequence pressure sequence in each sliding window is intercepted to obtain a window sequence, and based on all the actual compensation pressures in the window sequence, the instantaneous slope of the corresponding sliding window is calculated and integrated to obtain a time sequence slope sequence.

[0025] A feature extraction sequence at the current time is intercepted in the time sequence slope sequence with the current time as the end point, a weighted moving average algorithm using an exponential decay weight distribution strategy is introduced, a weight based on an exponential function is distributed to each instantaneous slope in the feature extraction sequence, the instantaneous slopes are weighted and fused according to the distributed weights to obtain a trend slope characteristic value at the current time.

[0026] The calculation method of the instantaneous slope is as follows:

[0027] All the actual compensation pressures in each window sequence are obtained, a pressure change curve of all the actual compensation pressures in the window sequence is fitted by using a linear least square method, and the instantaneous slope in the corresponding sliding window is calculated.

[0028] The generation method of the adaptive alarm threshold is as follows:

[0029] Obtain the time trend slope characteristic value of confirming that there is no leakage in the detection cavity in the historical period, integrate according to the time sequence, obtain the no-leakage sample sequence, and cut the adaptive reference sequence in the no-leakage sample sequence with the current time as the end point, and for the adaptive reference sequence, generate the adaptive alarm threshold value based on the 3 sigma principle, if the length of the no-leakage sample sequence does not meet the requirement of the adaptive reference sequence, the preset initial threshold value is used as the adaptive alarm threshold value;

[0030] The judgment mode of whether to trigger the prediction analysis is:

[0031] The adaptive alarm threshold value and the trend slope characteristic value of the current time are used to carry out the air tightness analysis of the detection cavity, the leakage flag value is set and the initial value is 0, the trend slope characteristic value of the current time is obtained and compared with the current adaptive alarm threshold value, if the trend slope characteristic value of the current time exceeds the adaptive alarm threshold value limit, the leakage flag value is increased by one, otherwise the leakage flag value is reset to 0;

[0032] If the leakage flag value is greater than the preset flag value standard, it is judged that the air tightness of the detection cavity is reduced, otherwise it is judged that the air tightness of the detection cavity is stable, and if the air tightness of the detection cavity is stable and the current time is the end point of the detection period, the prediction analysis is triggered;

[0033] The prediction analysis mode of the air tightness state of the working period is:

[0034] All trend slope characteristic values and adaptive alarm threshold values in the current detection period are obtained, and are respectively arranged into characteristic sequences and threshold sequences and are respectively normalized, a trend prediction model containing a 2-layer LSTM neural network after quantization compression is used to process the normalized characteristic sequences and threshold sequences, and the characteristic sequences and threshold sequences are used as inputs to predict the trend slope characteristic values and adaptive alarm threshold values in the working period with the current time as the starting point;

[0035] Based on the prediction result, the leakage flag value is calculated, if there is a time leakage flag value greater than the preset flag value standard, it is judged that the air tightness of the detection cavity is reduced, otherwise it is judged that the air tightness of the detection cavity is stable.

[0036] The beneficial effects of the present application are as follows:

[0037] 1、The present application can accurately divide the period and analyze the gas consumption by setting the detection triggering module, intelligently judge whether to trigger the air tightness detection, avoid unnecessary detection operation, save energy and time cost, the pressure compensation module considers the cavity temperature factor after triggering the detection, ensures the accuracy of the key reference data, and then obtains the accurate actual compensation pressure and time sequence pressure sequence, provides reliable data support for subsequent accurate detection, and effectively improves the accuracy and reliability of the air tightness detection.

[0038] 2、The application extracts leakage characteristics from time series pressure sequence efficiently by using time domain analysis technology based on sliding window through feature extraction module, obtains accurate trend slope characteristic value, and analyzes and predicts module generates adaptive alarm threshold based on historical data, accurately judges airtightness state in combination with current characteristic value, can also perform prediction analysis when needed, and knows airtightness condition in working period in advance, which helps to discover potential problems in time, take measures in advance, guarantees gas use safety, and reduces accident risk. BRIEF DESCRIPTION OF DRAWINGS

[0039] The application will be further described below with reference to the drawings.

[0040] Figure 1 is a module architecture diagram of the intelligent gas self-closing valve with airtightness detection according to an embodiment of the application;

[0041] Figure 2 is a running step flow chart of the intelligent gas self-closing valve with airtightness detection according to an embodiment of the application. DETAILED DESCRIPTION

[0042] In order to make the technical means, creative features, purposes and effects achieved by the application easy to understand, the application will be further described below in combination with specific embodiments.

[0043] Embodiment 1: Please refer to Figure 1 and Figure 2 The intelligent gas self-closing valve with airtightness detection according to an embodiment of the application includes the following modules:

[0044] The detection triggering module sets an analysis period and divides the analysis period into analysis sub-periods, collects and analyzes gas consumption in each analysis sub-period in a historical period, marks a detection period and a working period, and judges whether to trigger airtightness detection;

[0045] In the airtightness detection of the intelligent gas self-closing valve, one day is set as an analysis period, the starting point of the analysis period is a preset gas common time, a historical period is set for the intelligent gas self-closing valve, each analysis period in the historical period is evenly divided into analysis sub-periods, and analysis sub-periods with the same sequential number in each analysis period have the same time range in a day;

[0046] The gas consumption in each analysis sub-period is recorded by the gas meter. If the gas consumption in the analysis sub-period is 0, the analysis sub-period is marked as a no-flow period. The no-flow periods in the same period analysis cycle are analyzed. If the two analysis sub-periods are adjacent in time sequence and are both no-flow periods, the two analysis sub-periods are combined into one no-flow period. The no-flow periods in each period analysis cycle are sorted. The longest no-flow period in the no-flow periods in the same period analysis cycle is marked as a pre-selected detection period;

[0047] The start time of each pre-selected detection period is recorded. For the start times in different period analysis cycles, if the absolute value difference between the two start times is less than a preset time interval threshold, the two start times are classified into the same time group. A plurality of time groups are sorted. The number of start times in each time group in the historical period is compared. The time group with the largest number of start times is selected. The mean value of all start times in the time group and the mean value of the lengths of the pre-selected detection periods corresponding to the start times are calculated. The detection period is marked according to the time mean value and the length mean value.

[0048] It should be noted that the detection period is updated at the end of each period analysis cycle.

[0049] If the current time is the start time of the detection period, it is determined whether the intelligent gas self-closing valve is closed. If the intelligent gas self-closing valve is closed, the gas tightness detection is triggered. Otherwise, the intelligent gas self-closing valve is closed, and the start time of the detection period is changed to the time when the intelligent gas self-closing valve is closed. The gas tightness detection is triggered. The periods in each period analysis cycle that do not belong to the detection period are marked as working periods.

[0050] It should be noted that the purpose of this step is to dynamically determine the gas tightness detection period based on the gas usage rules. By dividing the analysis sub-period, counting the no-flow periods and combining adjacent no-flow periods, the longest no-flow period is selected as the pre-selected detection period. Then, the start times of the pre-selected detection periods in the historical period are clustered to determine the optimal detection period. At the same time, the working period is distinguished to avoid detection interference during gas use, improve detection accuracy, dynamically adapt to user gas usage habits, reduce manual intervention, improve detection efficiency, break the fixed period detection mode, and realize adaptive marking of the detection period through clustering analysis of historical flow data. The detection period matches the user's actual energy usage habits, enhancing the flexibility of the detection strategy.

[0051] Pressure compensation module: if the gas tightness detection is triggered, the cavity temperature of the detection cavity is collected and it is determined whether the hot stable state is entered. If the hot stable state is entered, the key reference data is determined. The cavity pressure is collected in real time and the actual compensation pressure is calculated based on the key reference data. The time sequence pressure sequence is generated.

[0052] If the airtightness detection is triggered, after the intelligent gas self-closing valve is closed, a closed detection cavity is formed in the downstream pipeline, a temperature sensor is used to monitor the cavity temperature in the detection cavity in real time, and an original temperature sequence containing the cavity temperature at each sampling point is generated, wherein the time interval between adjacent sampling points is one sampling period;

[0053] The cavity temperature difference of adjacent sampling points is calculated, and the cavity temperature change rate is calculated in combination with the sampling period length between adjacent sampling points. If the temperature change rate is less than a preset change rate threshold, it is judged that the temperature in the detection cavity between the two sampling points is stable. If the temperature in the detection cavity between all adjacent two sampling points is stable within a continuous preset time length, it is determined that the detection cavity reaches a thermal initial stable state;

[0054] It should be noted that this determination logic is based on the principle of thermodynamics. When the environmental temperature change rate is low enough, the pressure change of the gas under the condition of constant volume in a short time can be ignored, thereby forming a reliable reference;

[0055] At the moment when it is determined that the detection cavity reaches the thermal initial stable state, the current time is recorded as the reference time, and the key reference data including the reference temperature and the initial pressure are synchronously collected at the reference time. The reference temperature The initial pressure is obtained by collecting and converting the temperature sensor in the detection cavity to Kelvin temperature; It is collected by the pressure sensor arranged in the detection cavity;

[0056] After obtaining the key reference data, long-term monitoring is triggered, and the pressure sensor and the temperature sensor synchronously collect in real time with the same sampling period. The cavity temperature And the cavity pressure The hardware clock synchronization mechanism is used to ensure that the time stamps of the cavity temperature and the cavity pressure sampling points are strictly aligned. The collected cavity temperature and cavity pressure are matched to obtain a synchronous data pair in the detection cavity at the sampling point.

[0057] For the synchronous data pair, first, the temperature unit standardization processing is performed, the cavity temperature in the synchronous data pair is converted to Kelvin temperature, the isochoric process based on the ideal gas state equation is derived, the standardization pressure compensation formula is used to compensate the cavity pressure in the synchronous data pair, and the actual compensation pressure is calculated. The standardization pressure compensation formula is:

[0058] The actual compensation pressure at each time from the reference time to the current time is integrated according to the time sequence to obtain a time sequence pressure sequence;

[0059] It should be noted that the role of this step is to determine the initial stable state of the gas heat after the intelligent gas self-closing valve is closed, to synchronously collect the reference temperature and the initial pressure, to derive the pressure compensation formula combined with the ideal gas state equation, to generate the time sequence pressure sequence, to exclude the interference of temperature fluctuation on pressure detection, to establish reliable reference data, to ensure the accuracy of subsequent pressure analysis, to provide a scientific basis for gas tightness judgment, to introduce the principle of thermodynamics, to use the rate of temperature change as the heat stability judgment standard, to realize the alignment of temperature and pressure data through the hardware clock synchronization mechanism, to eliminate the temperature influence through the standardized pressure compensation formula, and to improve the anti-environmental interference ability of data acquisition;

[0060] The feature extraction module: a time domain analysis technology based on a sliding window is adopted to extract leakage features from the time sequence pressure sequence to obtain the trend slope characteristic value at the current time;

[0061] The time sequence pressure sequence is obtained, a time domain analysis technology based on a sliding window is adopted to extract leakage features from the time sequence pressure sequence to obtain the trend slope characteristic value as the leakage feature;

[0062] It should be noted that although the time sequence pressure sequence has excluded the environmental temperature interference, it may still contain sensor noise, mechanical vibration and other short-term disturbances, and the anti-interference leakage feature extraction function is to extract a quantitative index that can clearly represent the trend of gas tightness state change;

[0063] Specifically, a sliding window is set for the time sequence pressure sequence, the sliding window length is constant, and the sliding window is slid within the time period corresponding to the time sequence pressure sequence, a 50% overlap rate is set between adjacent sliding windows, the time sequence pressure sequence within each sliding window is intercepted to obtain a window sequence;

[0064] It should be noted that the role of the sliding window is to realize continuous monitoring within the time sequence pressure sequence, to ensure that no time sequence pressure sequence is missed, and to enable the feature extraction process to proceed smoothly and capture the start and change of the trend in a timely manner;

[0065] For all actual compensation pressures in each window sequence, a linear least squares method is used to fit the pressure change curve of all actual compensation pressures in the window sequence, to calculate the instantaneous slope representing the short-term trend within the corresponding sliding window, and to represent the downward trend of the actual compensation pressure when the instantaneous slope is negative;

[0066] The instantaneous slope within the corresponding sliding window is calculated for all window sequences, all instantaneous slopes are detected for abnormalities according to the time sequence, and a linear interpolation method is used to process the abnormal values, all processed instantaneous slopes are integrated according to the time sequence to obtain a time sequence slope sequence;

[0067] For the time series slope sequence, a feature analysis period ending at the current time is set, the feature analysis period slides with the change of the current time and has a constant length, the time series slope sequence in the feature analysis period is intercepted and marked as the feature extraction sequence of the current time, if the time length of the time series slope sequence corresponding to the period is less than the feature analysis period, the actual compensation pressure acquisition and the instantaneous slope calculation of the sliding window are continuously performed until the time series slope sequence meets the time length of the feature analysis period, and the feature extraction sequence of the current time is obtained;

[0068] For the feature extraction sequence, a weighted moving average algorithm using an exponential decay weight distribution strategy is introduced to perform leaky feature extraction on the feature extraction sequence;

[0069] Specifically, a weight based on an exponential function is assigned to each instantaneous slope in the feature extraction sequence, an attenuation factor for controlling the attenuation speed of the weight is set as the base of the exponential function, and the weight assigned to each instantaneous slope is calculated by taking the inverse order number of the instantaneous slope in the feature extraction sequence as the independent variable;

[0070] The trend slope feature value of the current time is obtained by performing weighted average calculation on the instantaneous slopes in the feature extraction sequence according to the weight assigned to each instantaneous slope;

[0071] It should be noted that the trend slope feature value is a continuous updating and highly robust quantitative index, which reflects the evolution direction and rate of the gas tightness in the detection cavity and provides a unique and reliable basis for the final leakage decision;

[0072] It should be noted that the role of this step is to use the sliding window technology on the time series pressure sequence, fit the pressure change curve by the linear least square method, calculate the instantaneous slope, generate the time series slope sequence after abnormal value processing, and extract the trend slope feature value by using the weighted moving average algorithm with exponential decay weight, effectively filter the short-term disturbance such as sensor noise and mechanical vibration, extract the quantitative index representing the change trend of the gas tightness, improve the anti-interference and continuity of the leakage feature, combine the sliding window overlap mechanism and the exponential decay weight distribution strategy, continuously monitor the pressure sequence, give higher weight to the recent data, enhance the capture ability of the leakage trend starting and change, and realize dynamic feature extraction;

[0073] Analysis and prediction module: based on the trend slope feature values in the historical period without leakage, an adaptive alarm threshold is generated, the gas tightness state in the detection cavity is judged combined with the trend slope feature value, and it is judged whether the prediction analysis is triggered, if triggered, a trend prediction model is constructed to predict and analyze the gas tightness state in the working period;

[0074] Obtaining a trend slope characteristic value calculated at each detection period in a historical period and integrating the trend slope characteristic value according to time sequence to obtain a no-leakage sample sequence, and cutting a threshold adaptive period from the current time as an end point, the threshold adaptive period sliding with the change of the current time and the length being unchanged, cutting the no-leakage sample sequence in the threshold adaptive period and marking the no-leakage sample sequence as an adaptive reference sequence;

[0075] For the adaptive reference sequence, an arithmetic mean μ and a standard deviation σ of the trend slope characteristic values in the adaptive reference sequence are calculated, the arithmetic mean μ represents an inherent and small pressure drift reference of the detection cavity in a no-leakage state, and the standard deviation σ quantifies a disturbance level, and an adaptive alarm threshold is generated based on a 3σ principle;

[0076] If the length of the period corresponding to the no-leakage sample sequence is less than the threshold adaptive period, a preset initial threshold is used as the adaptive alarm threshold;

[0077] Based on the trend slope characteristic value of the current time and the adaptive alarm threshold, the air tightness of the detection cavity is analyzed;

[0078] Specifically, a leakage flag value is set and an initial value of the leakage flag value is 0, a trend slope characteristic value of the current time is obtained and compared with the current adaptive alarm threshold, if the trend slope characteristic value of the current time exceeds the adaptive alarm threshold, the leakage flag value is increased by one, otherwise the leakage flag value is reset to 0;

[0079] If the leakage flag value is greater than a preset flag value standard, it is judged that the air tightness of the detection cavity is reduced and the detection cavity leaks, and an alarm is triggered and sent, otherwise it is judged that the air tightness of the detection cavity is stable, and if the air tightness of the detection cavity is stable and the current time is the end point of the detection period, a prediction analysis is triggered;

[0080] It should be noted that the leakage flag value is used to effectively filter out false signals of pressure fluctuations caused by temporary pipeline vibration or external impact, and to improve the reliability of the alarm;

[0081] If the prediction analysis is triggered, all trend slope characteristic values and adaptive alarm thresholds in the current detection period are obtained, and are respectively arranged into a feature sequence and a threshold sequence, the feature sequence and the threshold sequence are respectively normalized, a trend prediction model including a quantized and compressed 2-layer LSTM neural network is used to process the normalized feature sequence and the threshold sequence, and the feature sequence and the threshold sequence are input to the trend prediction model to predict the trend slope characteristic values and the adaptive alarm thresholds in a working period starting from the current time;

[0082] It should be noted that, since only the detection period and the working period exist in each period analysis cycle, the end of the detection period is the start of the working period, and the current time is the end of the detection period, so the current time is the start of the working period.

[0083] According to the predicted trend slope characteristic value in the working period and the adaptive alarm threshold, the gas tightness of the detection cavity in the working period is predicted and analyzed, and a leakage flag value is calculated.

[0084] If it is judged that the gas tightness of the detection cavity is stable, the next detection period is waited for to continue detection, otherwise an alarm is triggered to send.

[0085] If the alarm is triggered to send, an alarm is generated and sent to the user terminal in combination with the current or predicted trend slope characteristic value, and if the alarm is triggered to send, it is reset by manual means, or it is judged that the gas tightness of the detection cavity is stable in the detection period after confirming that the leakage has been repaired, and the alarm is removed.

[0086] It should be noted that the purpose of this step is to calculate the adaptive alarm threshold based on historical non-leakage data, to determine the leakage flag value by comparing the current trend slope characteristic value with the threshold, to predict the gas tightness trend in the working period by combining the LSTM neural network, to realize leakage judgment and early warning, to dynamically generate the threshold by the 3σ principle, to reduce false alarms caused by environmental disturbances, to introduce the LSTM prediction model to identify the gas tightness trend in advance, to improve the timeliness and reliability of early warning, to combine the adaptive threshold algorithm with the deep learning prediction model, to filter false signals through the leakage flag value, to realize a dual determination mechanism of real-time detection combined with trend prediction, and to promote the gas tightness detection from passive response to active early warning upgrade.

[0087] The technical scheme of the embodiment of the application is: setting a period analysis cycle and dividing an analysis sub-period, collecting and analyzing the gas consumption of each analysis sub-period in the historical period, marking the detection period and the working period, and judging whether to trigger the gas tightness detection, if the gas tightness detection is triggered, collecting the cavity temperature of the detection cavity and judging whether to enter the thermal steady state, if it is entered, determining the key reference data, collecting the cavity pressure in real time and compensating the actual compensation pressure according to the key reference data, generating a time series pressure sequence, using a time domain analysis technology based on a sliding window to extract the leakage characteristics of the time series pressure sequence, obtaining the trend slope characteristic value of the current time, generating an adaptive alarm threshold based on the trend slope characteristic value of the historical period without leakage, judging the gas tightness state in the detection cavity based on the trend slope characteristic value, and judging whether to trigger the prediction analysis, if the prediction analysis is triggered, a trend prediction model is constructed to predict and analyze the gas tightness state in the working period.

[0088] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. An intelligent gas self-closing valve with air tightness detection, characterized in that: The method comprises the following steps: The detection trigger module sets a period analysis cycle and divides the analysis sub-period, collects and analyzes the gas consumption of each analysis sub-period in the historical period, marks the detection period and the working period, and judges whether to trigger the gas tightness detection; The way to judge whether to trigger the gas tightness detection is: Set a period analysis cycle, and evenly divide each period analysis cycle into analysis sub-periods in the historical period; Use the gas meter to record and analyze the gas consumption in each analysis sub-period, mark the pre-selected detection period, if the absolute value difference between the start time of each period analysis cycle meets the condition, they belong to the same time group, select the time group with the most start times, and perform data processing on all start times and corresponding pre-selected detection period lengths in the time group, and mark the detection period; If the current time is the start time of the detection period and the intelligent gas self-closing valve is closed, trigger the gas tightness detection, if it is not closed, wait for it to be closed, and change the start time of the detection period to the closing time, trigger the gas tightness detection; The marking method of the pre-selected detection period is: If the gas consumption in the analysis sub-period is 0, mark the analysis sub-period as a no-flow period, analyze the no-flow periods in the same period analysis cycle, if the time-adjacent analysis sub-periods are both no-flow periods, combine the two analysis sub-periods into one no-flow period, and obtain the no-flow period in each period analysis cycle, compare the no-flow periods in the same period analysis cycle, and mark the no-flow period with the longest length as the pre-selected detection period; The pressure compensation module: if the gas tightness detection is triggered, collect the cavity temperature of the detection cavity and judge whether to enter the thermal steady state, if yes, determine the key reference data, real-time collect the cavity pressure and compensate the actual compensation pressure according to the key reference data, and generate a time sequence pressure sequence; The feature extraction module: using the time domain analysis technology based on sliding window, the leakage feature of the time sequence pressure sequence is extracted, and the trend slope characteristic value of the current time is obtained; The analysis and prediction module: based on the trend slope characteristic value without leakage in the historical period, an adaptive alarm threshold is generated, the gas tightness state in the detection cavity is judged, and whether to trigger the prediction analysis is judged, if yes, a trend prediction model is constructed to predict and analyze the gas tightness state in the working period.

2. The intelligent gas self-closing valve with air tightness detection according to claim 1, characterized in that: The way to obtain the key reference data is: After the intelligent gas self-closing valve is closed, the downstream pipeline forms a closed detection cavity, a temperature sensor is used to monitor the cavity temperature in the detection cavity in real time, and the time interval between adjacent sampling points is a sampling period; Based on the cavity temperature difference between adjacent sampling points and the sampling period length, the cavity temperature change rate is calculated, if the temperature change rate meets the preset change rate threshold condition in a continuous preset time length, it is determined that the detection cavity reaches the thermal initial steady state, the instant when the thermal initial steady state is reached is recorded as the reference time, and the key reference data is synchronously collected at the reference time, the key reference data includes the reference temperature and the initial pressure.

3. The intelligent gas self-closing valve with air tightness detection according to claim 2, characterized in that: The way to obtain the time sequence pressure sequence is: The cavity temperature and the cavity pressure in the detection cavity are collected, matched according to the time stamp, and the synchronous data pairs in the detection cavity at the sampling points are obtained, the cavity temperature in the synchronous data pairs is converted into Kelvin temperature, the standardized pressure compensation formula is derived based on the isochoric process of the ideal gas state equation, the cavity pressure in the synchronous data pairs is compensated, the actual compensation pressure is calculated, and the actual compensation pressure from the reference time to the current time is integrated according to the time sequence to obtain the time sequence pressure sequence.

4. The intelligent gas self-closing valve with air tightness detection according to claim 1, characterized in that: The trend slope characteristic value is obtained in the following manner: The time sequence pressure sequence is obtained, a sliding window is set for the time sequence pressure sequence, the time sequence pressure sequence in each sliding window is intercepted to obtain a window sequence, the instantaneous slope of the corresponding sliding window is calculated based on all the actual compensation pressures in the window sequence and integrated to obtain a time sequence slope sequence; A feature extraction sequence at the current time is intercepted from the time sequence slope sequence with the current time as the end point, a weighted moving average algorithm using an exponential decay weight distribution strategy is introduced, a weight based on an exponential function is distributed to each instantaneous slope in the feature extraction sequence, the instantaneous slopes are fused according to the distributed weights to obtain the trend slope characteristic value at the current time.

5. The intelligent gas self-closing valve with air tightness detection according to claim 4, characterized in that: The instantaneous slope is calculated in the following manner: All the actual compensation pressures in each window sequence are obtained, a pressure change curve of all the actual compensation pressures in the window sequence is fitted by using a linear least square method, and the instantaneous slope in the corresponding sliding window is calculated.

6. The intelligent gas self-closing valve with air tightness detection according to claim 1, characterized in that: The adaptive alarm threshold is generated in the following manner: The trend slope characteristic values at the time when no leakage is confirmed in the detection cavity in the historical period are obtained and integrated according to the time sequence to obtain a no-leakage sample sequence, an adaptive reference sequence is intercepted from the no-leakage sample sequence with the current time as the end point, and the adaptive alarm threshold is generated based on the 3σ principle for the adaptive reference sequence, if the length of the no-leakage sample sequence does not meet the requirement of the adaptive reference sequence, a preset initial threshold is used as the adaptive alarm threshold.

7. The intelligent gas self-closing valve with air tightness detection according to claim 6, characterized in that: The judgment of whether to trigger the prediction analysis is in the following manner: The adaptive alarm threshold and the trend slope characteristic value at the current time are used to perform airtightness analysis on the detection cavity, a leakage flag value is set and initialized to 0, the trend slope characteristic value at the current time is obtained and compared with the current adaptive alarm threshold, if the trend slope characteristic value at the current time exceeds the adaptive alarm threshold limit, the leakage flag value is incremented by one, otherwise the leakage flag value is reset to 0; If the leakage flag value is greater than a preset flag value standard, it is judged that the airtightness of the detection cavity is reduced, otherwise it is judged that the airtightness of the detection cavity is stable, and if the airtightness of the detection cavity is stable and the current time is the end point of the detection period, the prediction analysis is triggered.

8. The intelligent gas self-closing valve with air tightness detection according to claim 7, characterized in that: The prediction analysis on the airtightness state of the working period is in the following manner: All trend slope characteristic values in the current detection period and adaptive alarm thresholds are obtained, and are respectively arranged into a characteristic sequence and a threshold sequence and are respectively normalized. A trend prediction model including a quantized and compressed 2-layer LSTM neural network is used to process the normalized characteristic sequence and threshold sequence. The characteristic sequence and the threshold sequence are taken as inputs, and the trend slope characteristic values and the adaptive alarm thresholds in a work period starting from the current time are predicted. Based on the prediction results, a leakage flag value is calculated. If there is a time leakage flag value greater than a preset flag value standard, it is determined that the detection cavity gas tightness is reduced, otherwise it is determined that the detection cavity gas tightness is stable.

Images