Pipeline leakage diagnosis combined algorithm based on big data
Technical Field
The invention relates to a pipeline leakage algorithm, in particular to a pipeline leakage diagnosis combination algorithm based on big data, and belongs to the technical field of pipeline detection.
Background
The leakage of the petroleum pipeline is a potential threat to daily production and life, and brings great loss to people, so that the safe production of the pipeline can be effectively and timely ensured only by timely finding the leakage of the petroleum pipeline and the position of a leakage point.
The research of the pipeline leakage detection technology is relatively late in the beginning of China, but the development is fast, according to the difference of detection principles, various detection algorithms such as a negative pressure wave method, a sound wave method, a mass balance method and the like are mainly adopted, the positioning accuracy of the negative pressure wave method is poor compared with that of the sound wave method, the sound wave method is easily influenced by the external environment, the mass balance method is beneficial to the detection of tiny leakage, and the problems of contradiction between the detection sensitivity and false alarm and low positioning accuracy of the field are difficult to solve well by the methods.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a pipeline leakage diagnosis combination algorithm based on big data, which combines the results of various algorithms by adopting a means of combining various inspection methods, makes up the defects mutually, improves the detection effect, solves the contradiction between the detection sensitivity and the false alarm, and improves the positioning precision.
The invention provides a pipeline leakage combination algorithm based on big data, which specifically comprises the following steps:
the method comprises the following steps: acquiring pressure data, sound wave data and flow data of the pipeline in real time through a pressure transmitter, a sound wave sensor and a flowmeter which are respectively arranged at a head station and a tail station of the pipeline;
step two: connecting a PPS interface of a GPS system with a communication interface for synchronously storing pressure data, sound wave data and flow data;
step three: respectively obtaining measurement data by adopting a negative pressure wave method, a sound wave method and a mass balance method, fusing the pressure data, the sound wave data and the flow data in the step two in a plurality of data sources, and then performing leakage diagnosis by adopting a neural network algorithm to obtain a plurality of algorithm diagnosis results;
step four: synthesizing the diagnosis results of the algorithms in the third step, and obtaining a final diagnosis result by adopting a mode discrimination method of minimizing structural risk;
step five: and (4) integrating the positioning distances of all paths to obtain a combined algorithm positioning result by a statistical analysis method.
As a further improvement of the invention, the flow meter is installed at a distance of 200-300 meters from the pump.
As a further improvement of the invention, the pressure transmitter and the acoustic wave sensor are arranged at the same position and are connected through a tee joint.
As a further improvement of the invention, the PPS interface of the GPS synchronously stores real-time pressure data, sound wave data and flow data of the pipeline according to time sequence timing.
As a further improvement of the present invention, in step three, after pressure data, acoustic data and flow data are integrated, they are used as data input ends to establish a corresponding neural network input matrix.
As a further improvement of the invention, in step three, a neural network model based on the physical quantity of the pipeline state is established, and the current data is diagnosed.
As a further improvement of the invention, in the third step, the diagnosis results of the negative pressure wave, the sound wave method and the mass balance method are also input into the neural network, so that the neural network obtains the adaptive online learning capability.
As a further improvement of the method, in the fourth step, pressure data, sound wave data and flow data and the results of the negative pressure wave, the sound wave method and the mass balance method are gathered and input into a Support Vector Machine (SVM), an optimal segmentation hyperplane is constructed in the feature space based on the structure risk minimization theory, and the input multi-channel results are classified by using the SVM to obtain the diagnosis result of the combined algorithm.
As a further improvement of the invention, in the sixth step, the diagnosis positioning distances of different algorithms are summarized, and the positioning result of the combined algorithm is obtained by combining the historical positioning distances and the historical errors and applying a statistical analysis method.
Compared with the prior art, the pressure transmitter, the sound wave sensor and the flowmeter are respectively arranged at the head and tail stations of the pipeline, and the flowmeter is arranged in the pipeline, so that the pressure data, the sound wave data and the flow data of the pipeline are acquired in real time; the PPS interface of the GPS system is connected with the communication interface to synchronously store pressure data, sound wave data and flow data; respectively adopting a negative pressure wave method, a sound wave method and a mass balance method to obtain measurement data, fusing the pressure data, the sound wave data and the flow data under multiple scales, and then adopting a neural network algorithm to perform leakage diagnosis, and simultaneously obtaining multiple algorithm diagnosis results; synthesizing the diagnosis results of the algorithms, and obtaining a final diagnosis result by adopting a mode discrimination method of minimizing structural risk; and (4) integrating the positioning distances of all paths to obtain a combined algorithm positioning result by a statistical analysis method. The invention adopts the technical means of combining various inspection methods, based on the pipeline signal big data, applies the machine learning method, fuses the advantages of various algorithms, makes up for each other, and improves the sensitivity of leakage detection and the positioning precision of the leakage position.
Detailed Description
The present invention is further described below.
A pipeline leakage combination algorithm based on big data specifically comprises the following steps:
the method comprises the following steps: acquiring pressure data, sound wave data and flow data of the pipeline in real time through a pressure transmitter, a sound wave sensor and a flowmeter which are respectively arranged at a head station and a tail station of the pipeline;
step two: connecting a PPS interface of a GPS system with a communication interface for synchronously storing pressure data, sound wave data and flow data;
step three: respectively adopting a negative pressure wave method, a sound wave method and a mass balance method to obtain measurement data, fusing the pressure data, the sound wave data and the flow data in the step two under multiple scales, and then adopting a neural network algorithm to perform leakage diagnosis to obtain multiple algorithm diagnosis results;
step four: synthesizing the diagnosis results of the algorithms in the third step, and obtaining a final diagnosis result by adopting a mode discrimination method of minimizing structural risk;
step five: and (4) integrating the positioning distances of all paths to obtain a combined algorithm positioning result by a statistical analysis method.
In order to reduce the noise generated by the pump and make the data measured by the flow meter more accurate, the flow meter is installed at a distance of 200-300 meters from the pump.
And in order to accurately position the precision, the pressure transmitter and the acoustic wave sensor are arranged at the same position and are connected through a tee joint.
As a further improvement of the invention, the PPS interface of the GPS synchronously stores real-time pressure data, sound wave data and flow data of the pipeline according to time sequence timing.
As a further improvement of the present invention, in step three, after pressure data, acoustic data and flow data are integrated, they are used as data input ends to establish a corresponding neural network input matrix.
As a further improvement of the invention, in step three, a neural network model based on the physical quantity of the pipeline state is established, and the current data is diagnosed.
As a further improvement of the invention, in the third step, the diagnosis results of the negative pressure wave, the sound wave method and the mass balance method are also input into the neural network, so that the neural network obtains the adaptive online learning capability.
As a further improvement of the method, in the fourth step, pressure data, sound wave data and flow data and the results of the negative pressure wave, the sound wave method and the mass balance method are gathered and input into a Support Vector Machine (SVM), an optimal segmentation hyperplane is constructed in the feature space based on the structure risk minimization theory, and the input multi-channel results are classified by using the SVM to obtain the diagnosis result of the combined algorithm.
As a further improvement of the invention, in the sixth step, the diagnosis positioning distances of different algorithms are summarized, and the positioning result of the combined algorithm is obtained by combining the historical positioning distances and the historical errors and applying a statistical analysis method.
Examples
The summary logic of the system is that the leakage position mainly adopts the result of the negative pressure wave method, whether the leakage mainly adopts the result of the infrasonic wave method or not, and the result of the mass balance method is mainly used for checking and correcting.
For the negative pressure wave method, the instantaneous alarm output can generate a B-level alarm in the system, if the alarm result is verified by the infrasonic wave method or the mass balance method, the B-level alarm can be upgraded to an A-level alarm, and meanwhile, the leakage position adopts the position interval determined by the negative pressure wave method.
For the infrasonic wave method, instantaneous alarm output can generate A-level alarm in the system, if negative pressure wave also outputs alarm, the leakage position is corrected by the result of the negative pressure wave. If the negative pressure wave does not output the alarm, the mass balance method outputs the alarm, and the leakage position is corrected according to the result of the mass balance method.
For the mass balance method, the instantaneous alarm output can generate B-level alarm in the system, and if the instantaneous alarm output is verified by one or all of the other two methods, the instantaneous alarm output is upgraded to A-level alarm. If the negative pressure wave outputs an alarm, the leakage position is corrected according to the detection result of the negative pressure wave. In addition, if the time accumulation result of the mass balance method is output, the alarm is also upgraded to an A-level alarm.
And if the negative pressure wave output alarm is given, adopting the leakage position of the negative pressure wave.
In the case of the present example without integrating the mass balance method, the system output logic is as follows:
negative pressure wave method result
|
Result of infrasonic wave method
|
System alarm level
|
System alarm positioning
|
0
|
0
|
0
|
/
|
0
|
1
|
A
|
According to the infrasonic wave method
|
1
|
0
|
B
|
According to the negative pressure wave method
|
1
|
1
|
A
|
According to the negative pressure wave method |
The invention can be applied to various occasions, and can be used for a leakage detection algorithm for detecting leakage of different pipeline pressure amplitudes, different noise levels and different noise distributions in a certain range.