CN104730122A - Underground oil gas detection method based on electronic nose - Google Patents

Abstract

The invention discloses an underground oil gas detection method based on an electronic nose. The underground oil gas detection method comprises the following steps: oil gas components selectively penetrate through a film to enter the other side of the film from sampling drilling liquid; after gas and liquid are separated, the drilling liquid and carrier gas exist at the two sides of the film; the carrier gas carries the separated oil gas components to the electronic nose to be detected; oil gas information is converted into one group of current signals by the electronic nose; the current signals are converted into voltage signals by a current-voltage conversion circuit; each voltage signal passes a primary amplification circuit and a filtering module to enable an input signal to be amplified and an interference signal is filtered by smoothing; then the voltage signals are converted into direct-current current signals by a voltage-current conversion circuit; finally, a wireless data signal is transmitted by a wireless emission module; a ground wireless receiving device receives the wireless data signal sent from a shaft bottom and records the wireless data signal; and a signal processing and oil gas judging circuit processes the received data signal and judges the content of oil gas in the drilling liquid. According to the underground oil gas detection method, the electronic nose is used for carrying out oil gas detection so that the cost is low, the size is small and the miniaturization of a detection device is facilitated. The continuous and real-time monitoring can be realized.

Description

A kind of down-hole gas-oil detecting method based on Electronic Nose

Technical field

The present invention relates to detection method, particularly a kind of down-hole gas-oil detecting method based on Electronic Nose under a kind of Oil/gas Well, belong to oil, natural gas exploration and development technical field.

Background technology

Rapidly, downhole drill detection is a Main Trends of The Development of Logging Industry in current MWD, LWD, SWD development.Compare traditional ground gas detection logging, downhole drill gas detect can find identification of hydrocarbon fast, eliminate the delayed and interference that oil gas returns, and can carry out toxic and harmful forecast in advance.It is just towards the future development detecting real time implementation, rapid, serialization and downhole drill gas-liquid separation, detection.

Current with boring the degas method mainly membrane separation technique adopted in gas separaion detection technique, the gas detection method of employing is photoelectric technology, miniature chromatographic technique.And chromatographic technique often needs complicated preprocessing process, this adds the complicacy of gas detecting system undoubtedly, is unfavorable for the miniaturization of gas-detecting device.

Electronic Nose utilizes the response pattern of gas sensor array to identify the electronic system of smell, and it can monitor the odor profile of ad-hoc location continuously, in real time within several hours, several days even time of several months.As the core devices-gas sensor of Electronic Nose, in recent years, it is widely used in oil, petrochemical industry, mining industry.Sulfuretted hydrogen, carbon monoxide, chlorine, methane and flammable hydrocarbon have become its main detected object.

Summary of the invention

The object of this invention is to provide a kind of down-hole gas-oil detecting method based on Electronic Nose.The present invention can go out hydrocarbon information in downhole detection and real-time being uploaded to by Detection Information is abovegroundly carried out analysis and judged, thus realizes real-time accurate measurements to hydrocarbon information.

Step of the present invention is as follows:

One, filtration treatment is carried out to the drilling fluid of sampling, eliminate down-hole drilling fluids to the pollution of film;

Two, oil-gas component is optionally entered into from sampling drilling fluid the opposite side of film through film, reach the object of gas-liquid separation;

Three, after gas-liquid separation, film both sides are respectively drilling fluid and carrier gas, and isolated oil-gas component is carried into Electronic Nose and detects by carrier gas, and hydrocarbon information is converted to one group of current signal through Electronic Nose;

Four, current signal is converted into voltage signal through electric current-voltage conversion circuit: input signal to amplify and through filtering filtering interference signals through elementary amplifying circuit and filtration module by voltage signal respectively; And then with voltage one current converter circuit, voltage signal is converted to DC current signal; Wireless data signal is sent finally by wireless transmitter module;

Five, terrestrial wireless receiving trap receives the wireless data signal record that send from shaft bottom; Signal transacting and oil gas judging circuit are by the data-signal process received and judge the hydrocarbon content in drilling fluid.

Described carrier gas is helium, and its air flow rate is 800 ~ 1200ml/min.

The sensor array of described Electronic Nose is made up of MP-4, MQ306A, MC101 tri-kinds of electrochemical sensors.

The algorithm that described signal transacting and oil gas judging circuit adopt is RBF neural algorithm, RBF neural algorithm construction has the RBF neural structure of three layers, this RBF neural structure of three layers has three input layers, three hidden nodes and an output layer node, and wherein input layer realization is to the acceptance of information and transmission merit; When data are delivered to hidden layer, the function of hidden neuron carries out spatial mappings conversion to information, and information is here processed and classifies, and the neuron weighting that information is output layer afterwards becomes linear junction fruit, then exports; Its specific algorithm is:

If the n that is input as of RBF neural ties up, learning sample is (X, Y), wherein, and X=(X ₁, X ₂..., X _n), X is input variable; Xi=(X _i1, X _i2..., X _iN) ^t, 1≤i≤Nj, Y=(y ₁, y ₂..., y _n), Y is for expecting output variable; N is number of training, when neural network is input as X _itime, the output of a hidden layer jth node is:

$G(X_i,C_j,{\mathrm{\σ}}_j)=\exp (-||X_i-C_j||/2{\mathrm{\σ}}_j^2)$

C in formula _j=(c _j1, c _j2..., c _jn) ^t, be the center of a jth hidden layer Gaussian function; σ _jfor the width of a jth hidden layer Gaussian function;

To entirety input learning sample, network desired output is:

$Y=\underset{j=1}{\overset{M}{\mathrm{\Σ}}}G(X,C_j,{\mathrm{\σ}}_j)w_j+e$

W in formula _jfor the network connection weight of jth between hidden node and output layer, for single neural network exported, w _jit is a scalar; M is node in hidden layer; E is error of fitting;

The function of the hidden neuron of described RBF neural algorithm adopts Gaussian function, for hidden layer Gaussian function center, adopt Orthogonal Least Square choose, and apply least square method to network export weights train, the target of its learning training makes total error reach minimum, namely

$\min E=\frac{1}{2}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{e}_i^2$

In formula wherein, y _ifor when input amendment be X _itime desired output;

Described RBF neural algorithm is carrying out, in down-hole drilling fluids before oil-gas recognition, to train it, and its training sample elects the drilling fluid being dissolved with different content oil gas as.

Beneficial effect of the present invention is:

1, with film as with boring the separate medium extracting oil gas gas, the tripping device volume of making is little, easily and drilling tool integrated.

2, carry out oil and gas detection by Electronic Nose, with low cost, volume is little, is beneficial to pick-up unit miniaturization.Continuous, Real-Time Monitoring can be realized.

3, the RBF neural adopted is a kind of feedforward neural network model of good performance, and it has the advantages that the overall situation is approached, and there is not Local Minimum problem.

Accompanying drawing explanation

Fig. 1 is oil gas UF membrane and detects schematic diagram.

Fig. 2 is RBF neural model structure schematic diagram.

Embodiment

Overall work principle of the present invention as shown in Figure 1, wherein signal final analysis, identify and adopt RBF neural algorithm, RBF neural structural model is as shown in Figure 2.

Step of the present invention is as follows:

One, filtration treatment is carried out to the drilling fluid of sampling, eliminate down-hole drilling fluids to the pollution of film;

Two, oil-gas component is optionally entered into from sampling drilling fluid 1 opposite side of film through film 2, reach the object of gas-liquid separation;

Three, after gas-liquid separation, film both sides are respectively drilling fluid 1 and carrier gas 3, and isolated oil gas 4 component is carried into Electronic Nose 5 and detects by carrier gas 3, and hydrocarbon information is converted to one group of current signal through Electronic Nose 5;

Four, current signal is converted into voltage signal through electric current-voltage conversion circuit: input signal to amplify and through filtering filtering interference signals through elementary amplifying circuit and filtration module by voltage signal respectively; And then with voltage one current converter circuit, voltage signal is converted to DC current signal; Wireless data signal is sent finally by wireless transmitter module;

Five, terrestrial wireless receiving trap receives the wireless data signal record that send from shaft bottom; Signal transacting and oil gas judging circuit are by the data-signal process received and judge the hydrocarbon content in drilling fluid.

Described carrier gas is helium, and its air flow rate is 800 ~ 1200ml/min.

The sensor array of described Electronic Nose is made up of MP-4, MQ306A, MC101 tri-kinds of electrochemical sensors.

The algorithm that described signal transacting and oil gas judging circuit adopt is RBF neural algorithm, RBF neural algorithm construction has the RBF neural structure of three layers, this RBF neural structure of three layers has three input layers, three hidden nodes and an output layer node, and wherein input layer realization is to the acceptance of information and transmission merit; When data are delivered to hidden layer, the function of hidden neuron carries out spatial mappings conversion to information, and information is here processed and classifies, and the neuron weighting that information is output layer afterwards becomes linear junction fruit, then exports; Its specific algorithm is:

If the n that is input as of RBF neural ties up, learning sample is (X, Y), wherein, and X=(X ₁, X ₂..., X _n), X is input variable; Xi=(X _i1, X _i2..., X _iN) ^t, 1≤i≤Nj, Y=(y ₁, y ₂..., y _n), Y is for expecting output variable; N is number of training, when neural network is input as X _itime, the output of a hidden layer jth node is:

$G(X_i,C_j,{\mathrm{\σ}}_j)=\exp (-||X_i-C_j||/2{\mathrm{\σ}}_j^2)$

C in formula _j=(c _j1, c _j2..., c _jn) ^t, be the center of a jth hidden layer Gaussian function; σ _jfor the width of a jth hidden layer Gaussian function;

To entirety input learning sample, network desired output is:

$Y=\underset{j=1}{\overset{M}{\mathrm{\Σ}}}G(X,C_j,{\mathrm{\σ}}_j)w_j+e$

W in formula _jfor the network connection weight of jth between hidden node and output layer, for single neural network exported, w _jit is a scalar; M is node in hidden layer; E is error of fitting;

The function of the hidden neuron of described RBF neural algorithm adopts Gaussian function, for hidden layer Gaussian function center, adopt Orthogonal Least Square choose, and apply least square method to network export weights train, the target of its learning training makes total error reach minimum, namely

$\min E=\frac{1}{2}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{e}_i^2$

In formula wherein, y _ifor when input amendment be X _itime desired output;

Described RBF neural algorithm is carrying out, in down-hole drilling fluids before oil-gas recognition, to train it, and its training sample elects the drilling fluid being dissolved with different content oil gas as.

Claims (4)

1., based on a down-hole gas-oil detecting method for Electronic Nose, the step of the method is as follows:

One, filtration treatment is carried out to the drilling fluid of sampling, eliminate down-hole drilling fluids to the pollution of film;

Two, oil-gas component is optionally entered into from sampling drilling fluid the opposite side of film through film, reach the object of gas-liquid separation;

Three, after gas-liquid separation, film both sides are respectively drilling fluid and carrier gas, and isolated oil-gas component is carried into Electronic Nose and detects by carrier gas, and hydrocarbon information is converted to one group of current signal through Electronic Nose;

Four, current signal is converted into voltage signal through electric current-voltage conversion circuit: input signal to amplify and through filtering filtering interference signals through elementary amplifying circuit and filtration module by voltage signal respectively; And then with voltage one current converter circuit, voltage signal is converted to DC current signal; Wireless data signal is sent finally by wireless transmitter module;

Five, terrestrial wireless receiving trap receives the wireless data signal record that send from shaft bottom; Signal transacting and oil gas judging circuit are by the data-signal process received and judge the hydrocarbon content in drilling fluid.

2. a kind of down-hole gas-oil detecting method based on Electronic Nose according to claim 1, it is characterized in that: described carrier gas is helium, its air flow rate is 800 ~ 1200ml/min.

3. a kind of down-hole gas-oil detecting method based on Electronic Nose according to claim 1, is characterized in that: the sensor array of described Electronic Nose is made up of MP-4, MQ306A, MC101 tri-kinds of electrochemical sensors.

4. a kind of down-hole gas-oil detecting method based on Electronic Nose according to claim 1, it is characterized in that: the algorithm that described signal transacting and oil gas judging circuit adopt is RBF neural algorithm, RBF neural algorithm construction has the RBF neural structure of three layers, this RBF neural structure of three layers has three input layers, three hidden nodes and an output layer node, and wherein input layer realization is to the acceptance of information and transmission merit; When data are delivered to hidden layer, the function of hidden neuron carries out spatial mappings conversion to information, and information is here processed and classifies, and the neuron weighting that information is output layer afterwards becomes linear junction fruit, then exports; Its specific algorithm is:

If the n that is input as of RBF neural ties up, learning sample is (X, Y), wherein, and X=(X ₁, X ₂..., X _n), X is input variable; Xi=(X _i1, X _i2..., X _iN) ^t, 1≤i≤Nj, Y=(y ₁, y ₂..., y _n), Y is for expecting output variable; N is number of training, when neural network is input as X _itime, the output of a hidden layer jth node is:

$G(X_i,C_j,{\mathrm{\σ}}_j)=\exp (-||X_i-C_j||/2{\mathrm{\σ}}_j^2)$

C in formula _j=(c _j1, c _j2..., c _jn) ^t, be the center of a jth hidden layer Gaussian function; σ _jfor the width of a jth hidden layer Gaussian function;

To entirety input learning sample, network desired output is:

$Y=\underset{j=1}{\overset{M}{\mathrm{\Σ}}}G(X,C_j,{\mathrm{\σ}}_j)w_j+e$

W in formula _jfor the network connection weight of jth between hidden node and output layer, for single neural network exported, w _jit is a scalar; M is node in hidden layer; E is error of fitting;

The function of the hidden neuron of described RBF neural algorithm adopts Gaussian function, for hidden layer Gaussian function center, adopt Orthogonal Least Square choose, and apply least square method to network export weights train, the target of its learning training makes total error reach minimum, namely

$\min E=\frac{1}{2}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{e}_i^2$

In formula wherein, y _ifor when input amendment be X _itime desired output;

Described RBF neural algorithm is carrying out, in down-hole drilling fluids before oil-gas recognition, to train it, and its training sample elects the drilling fluid being dissolved with different content oil gas as.