阅读说明：本技术 基于数字音频处理的储层预测方法 (Reservoir prediction method based on digital audio processing ) 是由 杨培杰 陈攀峰 屈冰 董立生 管晓燕 张娟 罗红梅 张志敬 王庆华 韦欣法 于 2019-07-22 设计创作，主要内容包括：本发明提供一种基于数字音频处理的储层预测方法,该基于数字音频处理的储层预测方法包括：步骤1,输入SEGY格式的地震数据,将地震信号转MIDI格式的音频信号；步骤2,进行MIDI特征提取；步骤3,进行MIDI特征学习与分类；步骤4,根据MIDI特征学习与分类,进行音乐储层预测。该基于数字音频处理的储层预测方法种通过数字音频处理技术对地震信号进行处理,得到地震信号的音乐属性,进而提高地震储层预测的适用范围和效果。(The invention provides a reservoir prediction method based on digital audio processing, which comprises the following steps: step 1, inputting seismic data in SEGY format, and converting seismic signals into audio signals in MIDI format; step 2, extracting MIDI characteristics; step 3, learning and classifying MIDI characteristics; and 4, predicting the music reservoir according to MIDI characteristic learning and classification. The reservoir prediction method based on digital audio processing processes the seismic signals through the digital audio processing technology to obtain the music attribute of the seismic signals, and further improves the application range and effect of seismic reservoir prediction.)

1. The reservoir prediction method based on the digital audio processing is characterized by comprising the following steps of:

Step 1, inputting seismic data in SEGY format, and converting seismic signals into audio signals in MIDI format;

step 2, extracting MIDI characteristics;

Step 3, learning and classifying MIDI characteristics;

And 4, predicting the music reservoir according to MIDI characteristic learning and classification.

2. The method of claim 1, wherein in step 2, a Stockwell Transform is applied to perform a precise time-frequency analysis on the seismic data, and the formula is as follows:

Where τ represents time delay, f represents instantaneous frequency, x represents seismic signal, and t represents time. The exponential function in the integration is in the frequency domain.

3. the method of claim 2, wherein in step 2, the physical quantities of the Stockwell spectrum including frequency, time and amplitude are converted into basic MIDI attributes including tone, sound intensity and note length, and the MIDI files are inputted into the sequencer and converted into audible audio by the audio device after proper frequency conversion.

4. A method of reservoir prediction based on digital audio processing according to claim 3, characterized in that in step 2, the following mathematical relationship is applied to establish the relationship between seismic frequency f and MIDI note n:

f(n)＝440·2^(n-58)/12 (3)

Where n represents the serial number of the MIDI note.

5. the method of claim 4, wherein in step 2, the MIDI features are extracted from the transformed data, and the features are single-valued attributes related to instantaneous frequency or amplitude, or multi-valued MIDI attributes including melody, harmony, rhythm and style of seismic signals.

6. The method of claim 1, wherein in step 3, the MIDI features are clustered and classified by a convolutional neural network method in deep learning.

7. The reservoir prediction method based on digital audio processing as claimed in claim 1, wherein in step 4, through MIDI feature learning and classification, mudstone cover and two different types of gas sandstone reservoirs are distinguished, and effective reservoir prediction is realized from different angles.

Technical Field

The invention relates to the technical field of reservoir prediction, in particular to a reservoir prediction method based on digital audio processing.

Background

Seismic attribute analysis is one of the key links of reservoir prediction and is a special measurement value of geometric, kinematic, dynamic or statistical characteristics derived from seismic data. There are many classification methods for seismic attributes, and there are hundreds of seismic attributes included in these classifications. These complex seismic attributes are sensitive to which reservoir characteristics, and should be known and selected prior to performing the attribute extraction calculations.

Seismic attributes extracted from post-stack processing parameters can be classified into three categories according to different picking methods (1) attributes based on sections: this attribute falls into the category of special processing such as velocity, wave impedance, and AVO amplitude extraction. (2) Attributes based on the in-phase axis: it is an attribute extracted from seismic data and associated with an interface that provides information about how the attribute changes above and below the interface or between interfaces. (3) Volume attributes based on three-dimensional seismic data volumes: mainly refers to the information of seismic signal similarity and continuity between seismic traces, and can represent seismic geological features from a three-dimensional perspective.

Taner et al (1977 and 1979) and Bodine (1984) classically describe the theory and application of complex seismic trace attributes, commenting on conventional transient and response seismic attributes; the sensitivity of the seismic attributes to energy, bandwidth and phase is briefly described through a simple half-space model; the sensitivity of these seismic attributes to thickness is illustrated by a simple wedge model.

seismic reflection data may be subdivided into a number of components, such as energy, frequency, and phase. Seismic trace attributes, such as reflection intensity, instantaneous phase, instantaneous frequency, etc., are characterized by these seismic waveform components. Decomposing the attribute components of the seismic waveforms may allow interpreters to better identify and utilize these waveforms, as well as better distinguish and determine reservoir characteristics. The seismic attribute is mentioned again here, and has two broad categories, namely transient attribute and response attribute. The temporal attributes may represent the properties of each sample point and separate waveform components may be separated out in the seismic traces. Attributes of this type include quadrature amplitude, real amplitude, reflected intensity, instantaneous phase, cosine of instantaneous phase, and instantaneous frequency.

How to more effectively predict the reservoir depends on the familiarity of geology and earth scientists with seismic attributes and meanings of the seismic attributes, the seismic attributes are always hot spots for reservoir geophysical research, and in recent years, various seismic attributes emerge endlessly, and reservoir prediction is carried out by taking reference to other disciplines, so that one of reservoir geophysical development directions is at present. Therefore, a new reservoir prediction method based on digital audio processing is invented, and the technical problems are solved.

Disclosure of Invention

The invention aims to provide a reservoir prediction method based on digital audio processing, which is used for processing seismic signals through a digital audio processing technology to obtain the music attribute of the seismic signals so as to improve the application range and effect of seismic reservoir prediction.

The object of the invention can be achieved by the following technical measures: the reservoir prediction method based on the digital audio processing comprises the following steps: step 1, inputting seismic data in SEGY format, and converting seismic signals into audio signals in MIDI format; step 2, extracting MIDI characteristics; step 3, learning and classifying MIDI characteristics; and 4, predicting the music reservoir according to MIDI characteristic learning and classification.

the object of the invention can also be achieved by the following technical measures:

in step 2, using Stockwell Transform to perform accurate time-frequency analysis on seismic data, the formula is as follows:

where τ represents time delay, f represents instantaneous frequency, x represents seismic signal, and t represents that the exponential function in the time integral is in the frequency domain.

In step 2, the physical quantities of the Stockwell spectrum including frequency, time and amplitude are converted into basic MIDI attributes including tone, sound intensity and note length, and these MIDI files are input into the sequencer, and after proper frequency conversion, are converted into audible audio by the audio equipment.

In step 2, the following mathematical relationship is applied to establish the relationship between seismic frequency f and MIDI note n:

f(n)＝440·2^(n-58)/12 (3)

Where n represents the serial number of the MIDI note.

In step 2, the converted data is extracted with MIDI features, which are single-valued attributes related to instantaneous frequency or amplitude, or multi-valued MIDI attributes including melody, harmony, rhythm, and style of seismic signals.

In step 3, the MIDI features are clustered and classified by the convolutional neural network method in deep learning.

in step 4, through MIDI feature learning and classification, mudstone cover layers and two different gas-containing sandstone reservoirs are distinguished, and effective reservoir prediction is realized from different angles.

The reservoir prediction method based on digital audio processing converts seismic signals into audio signals, applies Stockwell transformation to carry out accurate time-frequency analysis on seismic data, and converts physical quantities (frequency, time and amplitude) of a Stockwell spectrum into basic MIDI attributes (such as tone, sound intensity and note length). MIDI is a standard protocol used in digital music, a powerful symbolic format, and this new "musical" attribute representation of seismic data offers many advantages. The audio attributes are clustered and classified through a Convolutional Neural Network (CNN) method in deep learning, so that a mudstone cover layer and two different gas-containing sandstone reservoirs (one is a reservoir with low gas saturation, and the other is high gas saturation) are distinguished, and reservoir prediction is further realized from different angles.

Drawings

FIG. 1 is a flow chart of one embodiment of a method for reservoir prediction based on digital audio processing of the present invention;

FIG. 2 is a schematic illustration of seismic data and its Stockwell spectra in an embodiment of the invention;

FIG. 3 is a schematic representation of the conversion of seismic data to audio in one embodiment of the present invention.

Detailed Description

In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.

1. SEGY to MIDI format

The SEGY is that seismic data is generally organized in units of seismic traces and stored in an SEG-Y file format. The SEG-Y format is one of the standard tape data formats proposed by SEG (society of Exploration geography), which is one of the most common formats of seismic data in the oil Exploration industry. Midi (musical Instrument digital interface) musical Instrument digital interface was proposed in the beginning of the 20 th century, 80 s to solve the problem of communication between electric musical instruments. MIDI is the most widespread standard format for music in the composer world and may be referred to as "computer understandable musical scores". It records music with digital control signals of the notes.

Pure tones can be considered as the basic composition of a wide variety of tones in nature, and the function of a pure tone can be written as:

y＝Asin2πw (1)

Where A is amplitude and w controls frequency.

The general expression do, re, mi, fa, so, la, ti is actually a notation of pitch, and table 1 is a frequency table of commonly used tones.

TABLE 1 common Audio-frequency watch

Seismic data in SEGY format is input, converted into audio signals, subjected to accurate time-frequency analysis using Stockwell Transform, and converted into "fundamental" MIDI attributes (such as pitch, sound intensity, and note length) in the physical quantities (frequency, time, and amplitude) of the Stockwell spectrum.

2. MIDI feature extraction

And (3) applying Stockwell Transform to perform accurate time-frequency analysis on the seismic data, wherein the formula is as follows:

Where τ represents time delay and f represents instantaneous frequency, and the exponential function in the integration is in the frequency domain, this transformation is suitable for analyzing seismic signals where the instantaneous frequency information varies with time (non-stationary signals). FIG. 2 shows a trace of seismic data and its Stockwell spectrum.

The physical quantities (frequency, time and amplitude) of the Stockwell spectrum are converted into "fundamental" MIDI properties (e.g. pitch, sound intensity and note length). Fig. 3 is a schematic diagram of converting seismic data into audio, where the seismic data is input, time-frequency characteristics of the seismic data are extracted through Stockwell transformation, and then converted into MIDI attributes (such as tone, sound intensity, and note length), and these MIDI files are input into a sequencer, and after correct frequency transformation, converted into audible audio through audio equipment.

The relationship between seismic frequency f and MIDI note n is established in the present invention using the following mathematical relationship:

f(n)＝440·2^(n-58)/12 (3)

where n denotes the serial number of the MIDI note, for example, n 180 corresponds to note B8, i.e., f 7902 Hz. The seismic data are converted into a frequency range of audio frequencies (30Hz to 20000Hz) by equation 3, and then subjected to audio analysis.

From the converted data, MIDI features are extracted, which may be single-valued attributes related to instantaneous frequency or amplitude, or multi-valued MIDI attributes, such as: melody, harmony, rhythm, style and other characteristics of the seismic signals.

3. MIDI feature learning and classification

MIDI feature learning and classification typically uses a number of broader classification methods, combining several types of learning algorithms, such as k-nearest neighbor algorithms, artificial neural network algorithms, support vector machines, random forests, and the like. The patent can perform clustering and classification on the features through a Convolutional Neural Networks (CNN) method in deep learning.

4. Music reservoir prediction

The final purpose of reservoir prediction is fluid detection, and through MIDI characteristic learning and classification, a mudstone cover layer and two different gas-containing sandstone reservoirs (one is a reservoir with low gas saturation, and the other is high gas saturation) can be effectively distinguished, so that effective reservoir prediction is realized from different angles.