A cloud computing-based exploration electronic measurement data analysis system and method

Abstract

The application relates to the technical field of exploration, and discloses an exploration electronic measurement data analysis system and method based on cloud computing. A preprocessing module is used for collecting multi-source exploration electronic measurement data, performing unified format conversion and data index establishment on different types of data through a cloud computing environment, removing noise and separating effective frequency band subbands, and obtaining preprocessed electronic measurement data; a classification module is used for inputting the preprocessed electronic measurement data into a CNN-LSTM hybrid model, dynamically adjusting a classification threshold according to signal distribution by adopting a self-adaptive particle swarm optimization algorithm, and outputting a classification result; a dynamic allocation module is used for constructing a PPO-DRL model and dynamically allocating computing resources; and a result generation module is used for analyzing classified data based on the classification result output by the CNN-LSTM hybrid model, generating an exploration result, and presenting the exploration result in a visual mode; and the overall analysis efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of exploration, in particular to an exploration electronic measurement data analysis system and method based on cloud computing. BACKGROUND

[0002] In the process of resource exploration such as oil and gas and mineral resources, the analysis of exploration electronic measurement data is a key link to obtain underground geological structure information; the existing exploration electronic measurement data analysis system mainly relies on the cloud platform to centralizedly process the electronic measurement data generated on the exploration site; when collecting the magnetotelluric signals and the seismic reflection signals simultaneously in oil and gas exploration, the existing system usually uses a unified preprocessing and filtering framework, ignoring the essential differences in the propagation mechanism and frequency response characteristics of the two types of signals; for example, if the same high-pass filter is used to remove low-frequency noise, the effective electromagnetic low-frequency components may be excessively attenuated, while the low-frequency components in the seismic signals are crucial to the resolution of tomographic imaging, resulting in systematic deviation of the inversion results.

[0003] The existing parallel computing architecture usually allocates computing resources averagely to various types of signals; for example, when processing high-frequency acoustic wave signals and low-frequency gravity signals, the system still uses the same slice size and iteration number, resulting in insufficient analysis of key signals and excessive calculation of secondary signals, causing low overall energy efficiency. SUMMARY

[0004] The purpose of the present application is to solve the above problems, and a kind of exploration electronic measurement data analysis system and method based on cloud computing are designed.

[0005] The present application provides an exploration electronic measurement data analysis system based on cloud computing, which comprises:

[0006] A preprocessing module is used to collect multi-source exploration electronic measurement data including electromagnetic wave signals, seismic wave signals and resistivity signals by an exploration device, to perform unified format conversion and data index establishment on different types of data by cloud computing environment, to denoise and separate effective frequency band subbands of different signals by WPT decomposition strategy, and to obtain preprocessed electronic measurement data.

[0007] A classification module is used to input the preprocessed electronic measurement data into a CNN-LSTM hybrid model, to dynamically adjust the classification threshold according to the signal distribution by using a self-adaptive particle swarm optimization algorithm, to perform feature extraction and classification on the preprocessed electronic measurement data, and to output the classification results.

[0008] A dynamic allocation module is used to construct a PPO-DRL model, to take signal type, data volume and node load as state, and to take resource allocation strategy as action, to dynamically allocate computing resources.

[0009] A result generation module is configured to analyze the classified data based on the classification result output by the CNN-LSTM hybrid model, generate exploration results, and present the exploration results in a visual manner.

[0010] Optionally, in the first implementation manner of the present application, the preprocessing module comprises:

[0011] A transmission sub-module is configured to collect multi-source exploration electronic measurement data including electromagnetic wave signals, seismic wave signals and resistivity signals by an exploration device, store the data in an original format and transmit the data to distributed storage nodes of a cloud computing environment;

[0012] A format conversion sub-module is configured to start a format conversion engine in the cloud computing environment, formulate mapping rules for different original format data, and perform unified format conversion on different types of data;

[0013] An index construction sub-module is configured to extract metadata based on the converted unified format data, construct a multi-level index by using a B+ tree index algorithm, and store index information in an index server of the cloud computing environment, wherein the metadata includes collection time, measuring point latitude and longitude, signal type and data volume;

[0014] A wavelet packet decomposition sub-module is configured to configure wavelet packet decomposition parameters for different signal types in the unified format, wherein the electromagnetic wave signals are decomposed by using a db6 wavelet base with 5 layers, the seismic wave signals are decomposed by using a sym8 wavelet base with 3 layers, and the resistivity signals are decomposed by using a db4 wavelet base with 2 layers;

[0015] A screening sub-module is configured to perform threshold denoising on each frequency band sub-band obtained by decomposition, screen effective sub-bands according to signal characteristics, and output preprocessed electronic measurement data, wherein the effective sub-bands are low frequency sub-bands of 0.001–62.5 Hz for the electromagnetic wave signals, main frequency sub-bands of 12.5–100 Hz for the seismic wave signals, and sub-bands without 50 or 60 Hz power frequency interference for the resistivity signals.

[0016] Optionally, in the second implementation manner of the present application, the format conversion sub-module comprises:

[0017] The seismic wave signals in the SEG-Y format are converted into the JSON format containing trace header information and sampling point values;

[0018] The electromagnetic wave signals in the EDIS format are converted into the JSON format containing frequency, amplitude and phase;

[0019] The resistivity signals in the CSV format are converted into the JSON format containing measuring point coordinates and resistivity values.

[0020] Optionally, in the third implementation manner of the present application, the index construction sub-module comprises:

[0021] The first index is divided according to a collection time interval, the second index is divided according to a measurement point geographical region, and the third index is divided according to a signal type.

[0022] Optionally, in the fourth implementation manner of the present application, the classification module comprises:

[0023] The first convolutional submodule is configured to input the preprocessed electronic measurement data into a CNN-LSTM hybrid model, and a first convolutional layer is configured to perform sliding convolution by using 32 3*3 convolutional kernels and extract local frequency domain features by using a ReLU activation function;

[0024] The second convolutional submodule is configured to perform deep feature extraction by using 64 3*3 convolutional kernels in a second convolutional layer, and enhance the capture of signal detail patterns.

[0025] The capture submodule is configured to perform 2*2 maximum pooling on the feature map output by the convolutional layer, input the pooled features into an LSTM layer with 128 hidden units, learn the dynamic change rule of the signal in the time dimension, and capture the time sequence dependency.

[0026] The weight distribution submodule is configured to introduce a Bahdanau attention mechanism, and distribute weights to the time sequence features output by the LSTM layer, so as to highlight the feature contribution of the time node.

[0027] The classification submodule is configured to output the classification results and corresponding confidence degrees of the electromagnetic signal, the seismic signal and the resistivity signal by using a fully connected layer and a softmax classifier.

[0028] Optionally, in the fifth implementation manner of the present application, the classification module further comprises:

[0029] The initialization submodule is configured to set 50 particles as threshold candidate solutions, each particle contains 3 dimensions, and the position and velocity of each particle are randomly initialized.

[0030] The first calculation submodule is configured to take the classification error rate as an adaptive function, apply the threshold value of the current particle to the classification result, and calculate the adaptive value of each particle.

[0031] The update submodule is configured to compare the current adaptive value of each particle with the historical optimal value, update the individual optimal position if the current value is better, compare the adaptive values of all particles, and update the global optimal position.

[0032] The iteration submodule is configured to update the velocity and position according to the individual optimal position and the global optimal position, guide the particles to move to a better solution, stop after 50 iterations, and take the threshold value corresponding to the global optimal position as the final classification threshold, wherein the final classification threshold is used for confidence judgment of real-time signal classification.

[0033] Optionally, in the sixth implementation mode of the present application, the dynamic allocation module comprises:

[0034] a consolidator module, configured to consolidate the signal type, the data volume, the real-time load rate of the computing node, and the signal importance weight into a state vector as an input of the PPO-DRL model;

[0035] a division module, configured to divide the action into three dimensions of computing node selection, CPU core number allocation, and memory allocation, and to calculate a reward value of the reward function by weighting a load deviation rate, a delay rate, and a classification error rate;

[0036] a second calculation sub-module, configured to calculate an action advantage value by using generalized advantage estimation, and to update the policy network and the value network by the reward function;

[0037] an output sub-module, configured to input the state vector of the to-be-processed signal into the PPO-DRL model, and to output an optimal action, so that the cloud computing environment schedules resources according to the scheme to realize dynamic allocation of computing resources.

[0038] Optionally, in the seventh implementation mode of the present application, the generation result module comprises:

[0039] a distribution sub-module, configured to distribute data to corresponding computing nodes according to categories by a task scheduler of the cloud computing environment based on a classification result output by the CNN-LSTM hybrid model;

[0040] a generation sub-module, configured to calculate resistivity values at different depths of the underground by the electromagnetic signal node through an inversion algorithm, to construct a seismic wave velocity model of the underground medium by the seismic signal node, and to identify stratum lithology according to resistivity differences by the resistivity signal node to generate preliminary analysis results;

[0041] a summary sub-module, configured to summarize the preliminary results of the nodes, to perform spatial coordinate matching based on GPS information of the exploration points, and to generate a multi-dimensional exploration report;

[0042] a conversion sub-module, configured to convert the aggregated multi-dimensional exploration report into a WebGL compatible format, to store the report as grid data in layers according to depths, and to convert attribute information into a texture map;

[0043] a display sub-module, configured to load the converted data through a 3D platform constructed by WebGL, and to display signal classification logs and resource allocation records to realize visualization of the exploration results.

[0044] Optionally, in the eighth implementation mode of the present application, the electromagnetic signal is distributed to the resistivity inversion node, the seismic signal is distributed to the velocity modeling node, and the resistivity signal is distributed to the lithology identification node.

[0045] Optionally, in the ninth implementation mode of the present application, a method for realizing a cloud computing-based exploration electronic measurement data analysis system is implemented, and the method comprises the following steps:

[0046] Multi-source exploration electronic measurement data including electromagnetic wave signals, seismic wave signals and resistivity signals are collected by an exploration device, unified format conversion and data index establishment are performed on different types of data in a cloud computing environment, WPT decomposition strategy is adopted to denoise and separate effective frequency band subbands of different signals, and preprocessed electronic measurement data are obtained;

[0047] The preprocessed electronic measurement data are input into a CNN-LSTM hybrid model, a self-adaptive particle swarm optimization algorithm is adopted to dynamically adjust classification threshold values according to signal distribution, feature extraction and classification are performed on the preprocessed electronic measurement data, and a classification result is output;

[0048] A PPO-DRL model is constructed, signal types, data volumes and node loads are taken as states, resource allocation strategies are taken as actions, and computing resources are dynamically allocated;

[0049] Based on the classification result output by the CNN-LSTM hybrid model, the classified data are analyzed, an exploration result is generated, and the exploration result is presented in a visual manner.

[0050] In the technical scheme provided by the present application, multi-source exploration electronic measurement data including electromagnetic wave signals, seismic wave signals and resistivity signals are collected by an exploration device, unified format conversion and data index establishment are performed on different types of data in a cloud computing environment, WPT decomposition strategy is adopted to denoise and separate effective frequency band subbands of different signals, and preprocessed electronic measurement data are obtained; the preprocessed electronic measurement data are input into a CNN-LSTM hybrid model, a self-adaptive particle swarm optimization algorithm is adopted to dynamically adjust classification threshold values according to signal distribution, feature extraction and classification are performed on the preprocessed electronic measurement data, and a classification result is output; a PPO-DRL model is constructed, signal types, data volumes and node loads are taken as states, resource allocation strategies are taken as actions, and computing resources are dynamically allocated; based on the classification result output by the CNN-LSTM hybrid model, the classified data are analyzed, an exploration result is generated, and the exploration result is presented in a visual manner; the CNN-LSTM hybrid model can effectively distinguish different types of exploration signals, avoids the signal aliasing problem in traditional methods, improves classification accuracy, the self-adaptive computing resource scheduling algorithm ensures that key signals obtain sufficient computing resources, overall analysis efficiency is improved, system energy consumption is reduced, data processing time is shortened through cloud computing, exploration analysis is more timely and efficient, and the accuracy and reliability of exploration results are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0051] Various other advantages and benefits will become apparent to those of ordinary skill in the art, upon reading the following detailed description of the preferred embodiments with reference made to the accompanying drawings. The drawings are for purposes of illustration only and are not intended to be limiting on the application.

[0052] Figure 1 A structural schematic diagram of a cloud computing-based exploration electronic measurement data analysis system provided by an embodiment of the application is shown in FIG. 1.

[0053] Figure 2 A structural schematic diagram of a preprocessing module provided by an embodiment of the application is shown in FIG. 2.

[0054] Figure 3 A structural schematic diagram of a dynamic allocation module provided by an embodiment of the application is shown in FIG. 3. DETAILED DESCRIPTION

[0055] The terms "first", "second", "third", "fourth" and the like in the description and claims of the application and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of these terms herein is to be construed as interchangeable in order to distinguish between the similar objects. It is also to be understood that the terms "comprising", "having", "including" and "containing" are to be construed as open-ended terms (i.e., meaning "including, but not limited to") unless otherwise noted. Thus, these terms are to be interpreted as specifying the presence of stated features or steps but not precluding the presence of one or more other features, steps or elements, whether or not the other features, steps or elements are specifically recited.

[0056] For the purpose of facilitating understanding, the specific flow of an embodiment of the application is described below, with reference to Figure 1 A structural schematic diagram of a cloud computing-based exploration electronic measurement data analysis system provided by an embodiment of the application is shown in FIG. 1.

[0057] The preprocessing module is configured to collect, by an exploration device, multi-source exploration electronic measurement data including electromagnetic wave signals, seismic wave signals and resistivity signals, perform uniform format conversion and data index establishment on different types of data in a cloud computing environment, denoise and separate effective frequency band subbands of different signals by using a WPT decomposition strategy, and obtain preprocessed electronic measurement data.

[0058] The classification module is configured to input the preprocessed electronic measurement data into a CNN-LSTM hybrid model, dynamically adjust a classification threshold according to signal distribution by using a self-adaptive particle swarm optimization algorithm, perform feature extraction and classification on the preprocessed electronic measurement data, and output a classification result.

[0059] A dynamic allocation module is configured to construct a PPO-DRL model, taking signal types, data volumes, and node loads as states, and taking resource allocation strategies as actions, to dynamically allocate computing resources.

[0060] A result generation module is configured to analyze the classified data based on the classification results output by the CNN-LSTM hybrid model, to generate exploration results, and to present the exploration results in a visual manner.

[0061] In this embodiment, please refer to Figure 2 The preprocessing module includes:

[0062] A transmission sub-module is configured to collect multi-source exploration electronic measurement data including electromagnetic wave signals, seismic wave signals, and resistivity signals by an exploration device, to store and transmit the data in an original format to distributed storage nodes in a cloud computing environment.

[0063] A format conversion sub-module is configured to start a format conversion engine in the cloud computing environment, to formulate mapping rules for different original format data, and to perform unified format conversion on different types of data.

[0064] A construction index sub-module is configured to extract metadata based on the converted unified format data, to construct a multi-level index using a B+ tree index algorithm, and to store index information in an index server in the cloud computing environment, wherein the metadata includes collection time, measuring point latitude and longitude, signal type, and data volume.

[0065] A wavelet packet decomposition sub-module is configured to configure wavelet packet decomposition parameters for different signal types in the unified format, wherein the electromagnetic wave signals are decomposed by 5 layers using a db6 wavelet base, the seismic wave signals are decomposed by 3 layers using a sym8 wavelet base, and the resistivity signals are decomposed by 2 layers using a db4 wavelet base.

[0066] A screening sub-module is configured to perform threshold denoising on each frequency band sub-band obtained by decomposition, to screen effective sub-bands according to signal characteristics, and to output preprocessed electronic measurement data, wherein the effective sub-bands are low frequency sub-bands of 0.001-62.5 Hz for electromagnetic wave signals, main frequency sub-bands of 12.5-100 Hz for seismic wave signals, and sub-bands without 50 or 60 Hz power frequency interference for resistivity signals.

[0067] In this embodiment, the format conversion sub-module includes: converting the seismic wave signals in SEG-Y format into JSON format containing trace header information and sampling point values; converting the electromagnetic wave signals in EDIS format into JSON format containing frequency, amplitude, and phase; and converting the resistivity signals in CSV format into JSON format containing measuring point coordinates and resistivity values.

[0068] In the embodiment, the index construction submodule includes: the first-level index is divided according to the collection time interval, the second-level index is divided according to the measurement point geographical region, and the third-level index is divided according to the signal type.

[0069] In the embodiment, the classification module includes:

[0070] The first convolutional submodule is configured to input the preprocessed electronic measurement data into the CNN-LSTM hybrid model, and the first convolutional layer is configured to perform sliding convolution by using 32 3*3 convolutional kernels and extract local frequency domain features by using a ReLU activation function.

[0071] The second convolutional submodule is configured to perform deep feature extraction by using 64 3*3 convolutional kernels in the second convolutional layer, and enhance the capture of signal detail patterns.

[0072] The capture submodule is configured to perform 2*2 maximum pooling on the feature map output by the convolutional layer, input the pooled features into an LSTM layer with 128 hidden units, learn the dynamic change rule of the signal in the time dimension, and capture the time sequence dependency.

[0073] The weight distribution submodule is configured to introduce a Bahdanau attention mechanism, distribute weights to the time sequence features output by the LSTM layer, and highlight the feature contribution of the time node.

[0074] The classification submodule is configured to output the classification results and the corresponding confidence degrees of the electromagnetic signal, the seismic signal and the resistivity signal by using a fully connected layer and a softmax classifier.

[0075] In the embodiment, the classification module further includes:

[0076] The initialization submodule is configured to set 50 particles as threshold candidate solutions, each particle includes three dimensions, and the positions and velocities of the particles are randomly initialized.

[0077] The first calculation submodule is configured to use the classification error rate as an adaptive function, apply the threshold value of the current particle to the classification result, and calculate the adaptive value of each particle.

[0078] The update submodule is configured to compare the current adaptive value of each particle with the historical optimal value, update the individual optimal position if the current value is better, compare the adaptive values of all particles, and update the global optimal position.

[0079] The iteration submodule is configured to update the velocity and position according to the individual optimal position and the global optimal position, guide the particles to move to a better solution, stop after 50 iterations, and use the threshold value corresponding to the global optimal position as the final classification threshold, wherein the final classification threshold is used for confidence judgment of real-time signal classification.

[0080] In this embodiment, the core parameters of the particle swarm are configured: 50 particles are set as the candidate solutions of the classification threshold, and the 3 dimensions of each particle correspond to the classification threshold of the electromagnetic wave signal, the seismic wave signal and the resistivity signal, respectively, i.e. the value of each dimension represents the confidence determination standard of the corresponding signal; then the position of each particle is randomly initialized, and the position value range is limited between [0.5, 1.0]; at the same time, the speed of each particle is randomly initialized, and the speed value range is limited between [-0.1, 0.1], finally forming an initial particle swarm containing 50 three-dimensional particles;

[0081] Taking the classification error rate as the only fitness function, the signal classification result output by the CNN-LSTM hybrid model is first obtained from the classification module, and then the threshold value corresponding to the 3 dimensions of the current particle is applied to the confidence determination of the three types of signals: if the confidence of a signal is higher than the threshold value of the corresponding dimension, it is determined as correct classification; if it is lower than the threshold value, it is determined as misclassification; then the number of misclassified samples under the threshold value of the particle and the total number of samples are counted, the ratio of the two is calculated, and the ratio is taken as the fitness value of the current particle. After the fitness values of all 50 particles are calculated, the fitness data set of the particle swarm is formed;

[0082] The optimization is executed in the order of individual update→group update: for each particle, the optimal fitness value stored in its history is called, which is the first fitness value after the particle is initialized. The current calculated fitness value is compared with the historical optimal value. If the current fitness value is smaller, i.e. the classification error rate is lower, the current position of the particle is updated as the new individual optimal position, and the historical optimal fitness value is updated synchronously. After the individual optimal update of all particles is completed, the current fitness values of all particles in the whole particle swarm are collected, the particle position corresponding to the minimum fitness value is selected, and the position is updated as the new global optimal position, ensuring that the particle swarm always converges to the direction with lower classification error rate;

[0083] Firstly, the particle velocity is updated based on the particle swarm optimization rule: the inertia weight, the cognitive factor and the social factor are introduced, the new speed of each particle is calculated by combining the current speed of the particle, the difference between the individual optimal position and the current position, and the difference between the global optimal position and the current position, so as to avoid the influence of too large or too small speed on the convergence efficiency; then the new position of each particle is calculated according to the new speed, so as to ensure that the new position is still within the reasonable threshold range of [0.5, 1.0]; subsequently, the first calculation submodule is called to recalculate the fitness values of all particles, and the update submodule is called to update the individual optimal position and the global optimal position, and the iteration is repeated according to the above process until the operation is stopped after 50 iterations; finally, the global optimal position at the end of iteration is extracted, and the three dimension values of the position are taken as the final classification thresholds of the electromagnetic wave, the seismic wave and the resistivity signal respectively, and in the real-time signal classification process, the type of the signal is determined by comparing the signal confidence with the corresponding final classification threshold, and the confidence verification is completed.

[0084] In the embodiment, please refer to Figure 3 The dynamic allocation module comprises:

[0085] The integration submodule is configured to integrate the signal type, the data volume, the real-time load rate of the computing node and the signal importance weight into a state vector as the input of the PPO-DRL model;

[0086] The division submodule is configured to divide the action into three dimensions of computing node selection, CPU core number allocation and memory allocation, and the reward value of the reward function is calculated by weighting the load deviation rate, the delay rate and the classification error rate;

[0087] The second calculation submodule is configured to calculate the action advantage value by using the generalized advantage estimation, and update the policy network and the value network through the reward function;

[0088] The output submodule is configured to input the state vector of the signal to be processed into the PPO-DRL model, output the optimal action, and schedule the resources according to the scheme in the cloud computing environment, so as to realize the dynamic allocation of the computing resources.

[0089] In the embodiment, the signal type is encoded according to the preset rule, for example, the electromagnetic signal is recorded as 1, the seismic signal is recorded as 2, and the resistivity signal is recorded as 3, the data volume is converted into a unified unit and is normalized to the interval [0, 1], the load rate percentage of each computing node is collected in real time through the monitoring interface of the cloud computing environment, the signal weight is preset according to the importance of the exploration target, for example, the resistivity is 0.4, the seismic is 0.35, and the electromagnetic is 0.25; then the processed signal type code, the normalized data volume, the node load rate and the signal weight are arranged in sequence to form a state vector with fixed dimensions, for example, a four-dimensional vector, and the vector is standardized to make the mean value of each dimension 0 and the variance 1, and finally the state vector is taken as the input data of the PPO-DRL model;

[0090] The computing node selection dimension includes all available computing nodes in the cloud, such as 5 nodes numbered 1 to 5, for the model to select a node from; the CPU core number allocation dimension sets discrete selectable values covering different processing requirements; the memory allocation dimension sets discrete selectable values matching the size of the data amount; at the same time, the sub-module defines the calculation method of the reward function: the load deviation rate is 1 minus the absolute deviation proportion of the actual load of the computing node and the target load, the delay rate is 1 minus the absolute deviation proportion of the actual processing time and the theoretical processing time, and the classification error rate is 1 minus the proportion of the number of misclassified samples and the total number of samples; finally, the above three indicators are weighted and summed according to the weight of 6:3:1 to obtain the reward value measuring the pros and cons of the action, providing a feedback signal for model training;

[0091] The trajectory data generated by the interaction between the PPO-DRL model and the environment includes state, action, reward, and next state, and the generalized advantage estimation algorithm is used to calculate the action advantage value: based on the reward value of the current state and the value estimation value of the next state, a decay coefficient λ such as 0.95 is introduced to weight and sum the multi-step advantage, obtaining the advantage estimation of each action relative to the average level; then, the advantage value is used to update the policy network: through the clipping objective function of PPO, the policy update amplitude is limited to ensure the stability of the training, so that the probability ratio of the new policy and the old policy is within a reasonable range; at the same time, the value network is updated based on the mean square error loss function, so that the value network estimates the state value more accurately, and by alternately updating the policy network and the value network, the decision-making ability of the model is continuously optimized;

[0092] After receiving the relevant information of the signal to be processed, the integration sub-module generates the corresponding state vector, and inputs the vector into the PPO-DRL model trained to convergence; the model processes the state vector through the policy network, outputs the probability distribution of each action, and selects the action combination with the highest probability as the optimal action including the specific computing node number, CPU core number, and memory size; then, the optimal action is converted into a resource scheduling instruction executable in the cloud computing environment, and the specified number of CPU cores and memory are allocated to the selected computing node through the resource manager of the cloud, and the data of the signal to be processed is transmitted to the node; the resource manager monitors the node resource usage in real time to ensure accurate execution of the scheduling instruction, and finally realizes dynamic allocation of computing resources to meet the processing requirements of different signals.

[0093] In this embodiment, the generation result module includes:

[0094] The distribution sub-module is configured to distribute data to corresponding computing nodes according to the classification results output by the CNN-LSTM hybrid model through the task scheduler of the cloud computing environment, and distribute electromagnetic signals to resistivity inversion nodes, distribute seismic signals to velocity modeling nodes, and distribute resistivity signals to lithology identification nodes.

[0095] The generating sub-module is configured to calculate the resistivity values at different depths of the underground by the electromagnetic signal node through an inversion algorithm, to construct a seismic wave velocity model of the underground medium by the seismic signal node, and to identify the lithology of the stratum according to the resistivity difference by the resistivity signal node, so as to generate a preliminary analysis result;

[0096] The summarizing sub-module is configured to summarize the preliminary results of the nodes, to perform spatial coordinate matching based on the GPS information of the exploration points, and to generate a multi-dimensional exploration report;

[0097] The converting sub-module is configured to convert the aggregated multi-dimensional exploration report into a WebGL compatible format, to store the report as grid data in layers according to the depths, and to convert the attribute information into a texture map;

[0098] The displaying sub-module is configured to load the converted data through the 3D platform constructed by WebGL, to display the signal classification log and the resource allocation record, and to realize the visualization of the exploration results.

[0099] In the embodiment, for the classification and aggregation problems in multi-dimensional exploration signal cross-domain parallel statistics, the existing methods usually rely on fixed thresholds or manual intervention to distinguish signals, which is difficult to deal with the problems of signal aliasing, frequency band interlacing, strong time-varying nature, etc. in the exploration environment, resulting in large statistical bias and significant cross-domain interference. Therefore, an adaptive classification statistical model based on signal physical hierarchy is proposed in this paper. The model realizes hierarchical classification according to the physical properties of different signal sources such as electromagnetic waves, seismic waves, and resistivity, which fundamentally overcomes the statistical bias problem caused by signal mixing in traditional methods. In the classification process, the system intelligently allocates each type of signal to a dedicated statistical module according to the high-dimensional parameters such as the original characteristics, frequency band distribution, and time sequence characteristics of the signal. The adaptive mechanism can dynamically optimize the classification threshold according to the real-time signal distribution and automatically schedule data resources in the parallel architecture, ensuring that each type of signal obtains accurate and independent statistical processing. This method significantly improves the cross-domain computing efficiency, effectively suppresses the mutual interference between signal types, and greatly improves the accuracy and reliability of exploration data processing.

[0100] In this embodiment, by the essential differences in physical properties, frequency band range and time sequence characteristics of different types of signals such as electromagnetic waves, seismic waves and resistivity, a layered adaptive signal classification mechanism is constructed, which effectively solves the feature confusion and error accumulation problems caused by signal mixed processing in traditional methods; the model realizes fine classification based on the signal physical level, ensures that each type of signal is processed in an independent module, and suppresses the mutual interference between multi-modal signals from the source, significantly improving the accuracy of data statistics; in the cross-domain parallel computing architecture, the adaptive classification mechanism can dynamically optimize the threshold according to the real-time signal distribution, and automatically schedule the computing resources, realizing the differentiated processing of different signal types; for example, the high-frequency seismic signal and the low-frequency electromagnetic signal are respectively matched with appropriate computing nodes and iteration strategies, avoiding redundant calculation and improving the resource utilization efficiency; the system not only significantly speeds up the processing speed of large-scale exploration data, but also improves the accuracy and reliability of the inversion and interpretation link, providing key technical support for high-precision data analysis in complex geophysical exploration environment.

[0101] 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 embodiments, and the above embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit 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. A cloud computing based exploration electronic measurement data analysis system, characterized in that, The system comprises: A preprocessing module for collecting multi-source exploration electronic measurement data including electromagnetic wave signals, seismic wave signals and resistivity signals by exploration equipment, performing unified format conversion and data index establishment on different types of data through a cloud computing environment, denoising and separating effective frequency band subbands of different signals by adopting a WPT decomposition strategy, and obtaining preprocessed electronic measurement data; A classification module for inputting the preprocessed electronic measurement data into a CNN-LSTM hybrid model, dynamically adjusting a classification threshold according to signal distribution by adopting a self-adaptive particle swarm optimization algorithm, performing feature extraction and classification on the preprocessed electronic measurement data, and outputting a classification result; A dynamic allocation module for constructing a PPO-DRL model, taking signal types, data volumes and node loads as states, and taking a resource allocation strategy as an action to dynamically allocate computing resources; A result generation module for analyzing classified data based on the classification result output by the CNN-LSTM hybrid model, generating an exploration result, and presenting the result in a visual manner; The classification module comprises: A first convolutional submodule for inputting the preprocessed electronic measurement data into the CNN-LSTM hybrid model, performing sliding convolution on the first convolutional layer by using 32 3x3 convolutional kernels, and extracting local frequency domain features by a ReLU activation function; A second convolutional submodule for deepening feature extraction by using 64 3x3 convolutional kernels on the second convolutional layer, and enhancing the capture of signal detail patterns; A capture submodule for performing 2x2 maximum pooling on feature maps output by the convolutional layer, inputting the pooled features into an LSTM layer containing 128 hidden units, learning the dynamic change rule of signals in the time dimension, and capturing time sequence dependency; A weight allocation submodule for introducing a Bahdanau attention mechanism, allocating weights to time sequence features output by the LSTM layer, and highlighting the feature contribution of time nodes; A classification submodule for outputting classification results including electromagnetic signals, seismic signals and resistivity signals and corresponding confidence levels through a fully connected layer and a softmax classifier; The classification module further comprises: An initialization submodule for setting 50 particles as threshold candidate solutions, each particle containing 3 dimensions, and randomly initializing the position and speed of the particles; A first calculation submodule for taking the classification error rate as a fitness function, applying the threshold value of the current particle to the classification result, and calculating the fitness value of each particle; An update submodule for comparing the current fitness value of each particle with the historical optimal value, updating the individual optimal position if the current value is better, comparing the fitness values of all particles, and updating the global optimal position; An iteration submodule for updating the speed and position according to the individual optimal and global optimal positions, guiding the particles to move to better solutions, stopping after 50 iterations, and taking the threshold value corresponding to the global optimal position as the final classification threshold, wherein the final classification threshold is used for confidence judgment in real-time signal classification.

2. The cloud computing based exploration electronic measurement data analysis system of claim 1, wherein, The preprocessing module comprises: The transmission sub-module is configured to collect multi-source exploration electronic measurement data including electromagnetic wave signals, seismic wave signals and resistivity signals by the exploration equipment, store the data in the original format and transmit the data to the distributed storage nodes of the cloud computing environment; The format conversion sub-module is configured to start a format conversion engine in the cloud computing environment, formulate mapping rules for different original format data, and perform unified format conversion on different types of data; The index construction sub-module is configured to extract metadata based on the converted unified format data, construct a multi-level index by using a B+ tree index algorithm, and store the index information in an index server of the cloud computing environment, wherein the metadata includes collection time, measurement point latitude and longitude, signal type and data volume; The wavelet packet decomposition sub-module is configured to configure wavelet packet decomposition parameters for different signal types in the unified format, wherein the electromagnetic wave signals are decomposed by using a db6 wavelet base with 5 layers, the seismic wave signals are decomposed by using a sym8 wavelet base with 3 layers, and the resistivity signals are decomposed by using a db4 wavelet base with 2 layers; The screening sub-module is configured to perform threshold denoising on each frequency band sub-band obtained by decomposition, screen effective sub-bands according to signal characteristics, and output preprocessed electronic measurement data, wherein the effective sub-bands are low frequency sub-bands of 0.001–62.5 Hz for the electromagnetic wave signals, main frequency sub-bands of 12.5–100 Hz for the seismic wave signals, and sub-bands without 50 or 60 Hz power frequency interference for the resistivity signals.

3. The cloud computing based exploration electronic measurement data analysis system of claim 2, wherein, The format conversion sub-module includes: The SEG-Y format seismic wave signal is converted into a JSON format containing trace header information and sampling point values; The EDIS format electromagnetic wave signal is converted into a JSON format containing frequency, amplitude and phase; The CSV format resistivity signal is converted into a JSON format containing measurement point coordinates and resistivity values.

4. The cloud computing based exploration electronic measurement data analysis system of claim 2, wherein, The index construction sub-module includes: The first-level index is divided according to the collection time interval, the second-level index is divided according to the measurement point geographical region, and the third-level index is divided according to the signal type.

5. The cloud computing based exploration electronic measurement data analysis system of claim 1, wherein, The dynamic allocation module includes: The integration sub-module is configured to integrate the signal type, data volume, real-time load rate of the computing node, and signal importance weight into a state vector as an input of the PPO-DRL model; The division sub-module is configured to divide the action into three dimensions of computing node selection, CPU core number allocation and memory allocation, and calculate the reward value of the reward function by weighting the load bias rate, delay rate and classification error rate; The second calculation sub-module is configured to calculate the action advantage value by using generalized advantage estimation, and update the policy network and the value network by using the reward function; The output sub-module is configured to input the state vector of the signal to be processed into the PPO-DRL model, output an optimal action, and schedule resources according to the scheme by the cloud computing environment to realize dynamic allocation of computing resources.

6. The cloud computing based exploration electronic measurement data analysis system of claim 1, wherein, The result generation module includes: The distribution sub-module is configured to distribute the data to the corresponding computing nodes according to the classification results output by the CNN-LSTM hybrid model through the task scheduler of the cloud computing environment. The generating submodule is configured to calculate resistivity values at different depths of the underground by an inversion algorithm of the electromagnetic signal node, to construct a seismic wave velocity model of the underground medium by the seismic signal node, and to identify stratum lithology according to resistivity differences by the resistivity signal node, and to generate a preliminary analysis result; The aggregating submodule is configured to aggregate the preliminary results of the nodes, to perform spatial coordinate matching based on GPS information of the exploration points, and to generate a multi-dimensional exploration report; The converting submodule is configured to convert the aggregated multi-dimensional exploration report into a WebGL compatible format, to store the report as grid data in layers according to depths, and to convert attribute information into a texture map; The displaying submodule is configured to load the converted data through a 3D platform constructed by WebGL, and to display signal classification logs and resource allocation records, thereby realizing visualization of the exploration results.

7. The cloud computing based exploration electronic measurement data analysis system as claimed in claim 6, wherein, The electromagnetic signal is distributed to the resistivity inversion node, the seismic signal is distributed to the velocity modeling node, and the resistivity signal is distributed to the lithology identification node.

8. A method of implementing a cloud computing based exploration electronic measurement data analysis system as claimed in claim 1, wherein, The method comprises the following steps: The multi-source exploration electronic measurement data including electromagnetic wave signals, seismic wave signals and resistivity signals are collected by an exploration device, the different types of data are converted into a unified format and indexed by a cloud computing environment, the different signals are denoised and separated into effective frequency band subbands by using a WPT decomposition strategy, and the preprocessed electronic measurement data are obtained; The preprocessed electronic measurement data are input into a CNN-LSTM hybrid model, and a self-adaptive particle swarm optimization algorithm is used to dynamically adjust classification thresholds according to signal distribution, so as to extract features and classify the preprocessed electronic measurement data, and output classification results: the preprocessed electronic measurement data are input into the CNN-LSTM hybrid model, 32 3×3 convolution kernels are used in the first convolution layer to perform sliding convolution, and local frequency domain features are extracted through a ReLU activation function; 64 3×3 convolution kernels are used in the second convolution layer to deepen feature extraction and enhance the capture of signal detail patterns; 2×2 maximum pooling is performed on the feature maps output by the convolution layer, the pooled features are input into an LSTM layer containing 128 hidden units, the dynamic change law of the signal in the time dimension is learned, and the time sequence dependence relationship is captured; the Bahdanau attention mechanism is introduced to assign weights to the time sequence features output by the LSTM layer, and the feature contribution of the time nodes is highlighted; the classification results and corresponding confidence levels of the electromagnetic signal, the seismic signal and the resistivity signal are output through a fully connected layer and a softmax classifier; 50 particles are set as threshold candidate solutions, each particle contains 3 dimensions, and the positions and velocities of the particles are randomly initialized; the classification error rate is used as the fitness function, the threshold values of the current particles are applied to the classification results, and the fitness values of each particle are calculated; each particle compares its current fitness value with the historical optimal value, and if the current value is better, the individual optimal position is updated; the fitness values of all particles are compared, and the global optimal position is updated; the velocity and position are updated according to the individual optimal and global optimal positions, the particles are guided to move to better solutions, and the iteration is stopped after 50 times; and the threshold value corresponding to the global optimal position is taken as the final classification threshold. The PPO-DRL model is constructed, a signal type, a data amount and a node load are used as states, a resource allocation strategy is used as an action, and computing resources are dynamically allocated. Based on the classification results output by the CNN-LSTM hybrid model, the classified data are analyzed, exploration results are generated, and the exploration results are presented in a visual manner.

Images