A Scene-Adaptive Multi-Component Gas Rapid Detection Device and Method

Abstract

The present invention discloses a scene - adaptive multi - component gas rapid detection device and method, which relates to the technical field of gas detection. The device includes: collecting multi - dimensional features of the target application scenario of the target transformer oil to obtain target scenario feature information; using the target scenario feature information as input information to obtain output information, screening the historical gas detection records of the same type of transformer oil of the target transformer oil to obtain target historical gas detection records; extracting the first historical gas absorption information, performing a union operation on the gas components to obtain a target gas component set; reading a predetermined wavelength strategy to obtain a target wavelength range; activating a mid - infrared detection device to perform gas detection. The present invention solves the technical problems of the prior art, such as low accuracy, low efficiency, and inability to adapt to complex scenarios in the gas detection of transformer oil, and achieves the technical effects of improving the accuracy of gas detection in transformer oil, optimizing the detection efficiency, and enhancing the adaptability.

Description

Technical Field

[0001] The present invention relates to the technical field of gas detection, and particularly to a scene-adaptive multi-component gas rapid detection device and method. Background Art

[0002] In today's power system, as a key power equipment, the stable operation state of a transformer is directly related to the safe and reliable operation of the entire power grid. Transformer oil plays important roles such as insulation, cooling, and arc quenching in a transformer. However, during the operation of the transformer, due to the influence of various internal factors (such as overheating, discharge, etc.) and external factors (such as humidity, pollution, etc.), the transformer oil will decompose and undergo chemical reactions, generating various gases. Detecting and analyzing the gases generated in the transformer oil is an important means to evaluate the operation state of the transformer and diagnose potential faults.

[0003] The prior art has technical problems such as low accuracy, low efficiency, and inability to adapt to complex scenarios in the detection of gases in transformer oil. Summary of the Invention

[0004] The present application provides a scene-adaptive multi-component gas rapid detection device and method for solving the technical problems of low accuracy, low efficiency, and inability to adapt to complex scenarios in the prior art in the detection of gases in transformer oil.

[0005] In view of the above problems, the present application provides a scene-adaptive multi-component gas rapid detection device and method.

[0006] In the first aspect of the present application, a scene-adaptive multi-component gas rapid detection device is provided, and the device includes:

[0007] Scenario feature information acquisition module, which collects multi-dimensional features of the target application scenario of the target transformer oil based on predetermined scenario features to obtain target scenario feature information; output information acquisition module, which is used to use the target scenario feature information as the input information of the transformer oil utility prediction model to obtain output information, where the output information includes target utility; historical gas detection record acquisition module, which screens the historical gas detection records of the same type of transformer oil of the target transformer oil based on the target utility to obtain target historical gas detection records; first historical gas absorption information acquisition module, which is used to extract the first historical record in the target historical gas detection records and obtain the first historical gas absorption information in the first historical record, and the first historical gas absorption information includes multiple gas components with absorption amount identifiers; gas component set acquisition module, which performs union operation on the multiple gas components with absorption amount identifiers to obtain a target gas component set; wavelength interval acquisition module, which reads a predetermined wavelength strategy and analyzes the target gas component set based on the predetermined wavelength strategy to obtain a target wavelength interval; gas detection module, which activates a mid-infrared detection device to detect the gas of the target transformer oil in the target wavelength interval.

[0008] In the second aspect of the present application, a multi-component gas rapid detection method adaptable to scenarios is provided, and the method includes:

[0009] Collect multi-dimensional features of the target application scenario of the target transformer oil based on predetermined scenario features to obtain target scenario feature information; use the target scenario feature information as the input information of the transformer oil utility prediction model to obtain output information, where the output information includes target utility; screen the historical gas detection records of the same type of transformer oil of the target transformer oil based on the target utility to obtain target historical gas detection records; extract the first historical record in the target historical gas detection records and obtain the first historical gas absorption information in the first historical record, and the first historical gas absorption information includes multiple gas components with absorption amount identifiers; perform union operation on the multiple gas components with absorption amount identifiers to obtain a target gas component set; read a predetermined wavelength strategy and analyze the target gas component set based on the predetermined wavelength strategy to obtain a target wavelength interval; activate a mid-infrared detection device to detect the gas of the target transformer oil in the target wavelength interval.

[0010] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0011] A scene feature information acquisition module collects multi-dimensional features of the target application scenario of the target transformer oil based on predetermined scene features to obtain target scene feature information; an output information acquisition module uses the target scene feature information as input information of a transformer oil utility prediction model to obtain output information; a historical gas detection record acquisition module screens the historical gas detection records of the same type of transformer oil of the target transformer oil based on the target utility to obtain target historical gas detection records; a first historical gas absorption information acquisition module extracts the first historical records from the target historical gas detection records and obtains the first historical gas absorption information in the first historical records; a gas component set acquisition module is used to perform a union operation on multiple gas components with absorption amount identifiers to obtain a target gas component set; a wavelength interval acquisition module is used to read a predetermined wavelength strategy to obtain a target wavelength interval; a gas detection module detects the gas in the target transformer oil. The technical effects of improving the accuracy of gas detection in transformer oil, optimizing the detection efficiency, and enhancing the adaptability are achieved. Description of the Drawings

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0013] Figure 1 It is a schematic structural diagram of a scene-adaptive multi-component gas rapid detection device provided by an embodiment of the present application;

[0014] Figure 2 It is a schematic flow diagram of a scene-adaptive multi-component gas rapid detection method provided by an embodiment of the present application.

[0015] Description of the reference numerals: scene feature information acquisition module 10, output information acquisition module 20, historical gas detection record acquisition module 30, first historical gas absorption information acquisition module 40, gas component set acquisition module 50, wavelength interval acquisition module 60, gas detection module 70. Detailed Embodiments

[0016] The present application provides a scene-adaptive multi-component gas rapid detection device and method for solving the technical problems of low accuracy, low efficiency, and inability to adapt to complex scenarios in the prior art for gas detection in transformer oil.

[0017] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts belong to the scope of protection of the present application.

[0018] Embodiment 1

[0019] As Figure 1 shown, the present application provides a scene - adaptive multi - component gas rapid detection device, and the device includes:

[0020] A scene feature information acquisition module 10, which collects multi - dimensional features of the target application scenario of the target transformer oil based on predetermined scene features to obtain target scene feature information.

[0021] Specifically, the scene feature information acquisition module 10 undertakes an important task. According to the pre - set predetermined scene features, it comprehensively and deeply collects features of the target application scenario of the target transformer oil from multiple dimensions. Information is collected from aspects such as the environmental conditions where the transformer is located (such as temperature, humidity, altitude), the operating load of the transformer, the service life, and the maintenance history. Through the comprehensive analysis and integration of these multi - dimensional features, the target scene feature information that can accurately describe the characteristics of this application scenario is finally obtained. This multi - dimensional feature collection helps to more comprehensively and accurately understand the working environment and operating state of the transformer oil, provides a very rich and reliable data basis for the accurate analysis, evaluation of the performance of the transformer oil and the prediction of potential problems, helps to discover problems in time, take preventive measures, and ensure the safe and stable operation of the transformer.

[0022] An output information acquisition module 20, which is used to use the target scene feature information as the input information of the transformer oil utility prediction model to obtain output information, where the output information includes the target utility.

[0023] Specifically, the output information acquisition module 20 uses the target scenario feature information previously collected and organized by the scenario feature information acquisition module 10 as the input data for the transformer oil utility prediction model. This transformer oil utility prediction model is a complex mathematical model trained and optimized with a large amount of data, capable of conducting in-depth analysis and calculation based on the input information. When the target scenario feature information is input into the model, the model will use its internal complex algorithms and logics to process and deduce these data. Finally, the model outputs a series of information, and these output information includes the key target utility. The target utility covers important aspects such as the insulation performance evaluation of transformer oil, the heat dissipation effect prediction, and the service life estimation, which can accurately reflect the actual utility and potential value of transformer oil in a specific scenario, realizing the effective utilization of the application scenario feature information of transformer oil and improving the accuracy and reliability of the transformer oil utility prediction.

[0024] A historical gas detection record acquisition module 30, and the historical gas detection record acquisition module 30 screens the historical gas detection records of the same type of transformer oil of the target transformer oil based on the target utility to obtain the target historical gas detection records.

[0025] Specifically, according to the key information of the target utility obtained previously, a targeted screening work is carried out on the historical gas detection records of the same type of transformer oil of the target transformer oil. For example, if the target utility of the currently evaluated target transformer oil is for lubrication, then this module will only select the gas detection records of the transformer oil that was also used for lubrication historically. This screening method has clear purpose and pertinence. Through this precise screening, the historical detection records that are not relevant to the current target utility can be excluded, thus focusing on the data with direct reference value. This helps to improve the accuracy and efficiency of subsequent analysis and avoid the interference of irrelevant data.

[0026] A first historical gas absorption information acquisition module 40, and the first historical gas absorption information acquisition module 40 is used to extract the first historical record from the target historical gas detection records and obtain the first historical gas absorption information in the first historical record. The first historical gas absorption information includes various gas components with absorption amount identifiers.

[0027] Specifically, in-depth mining is carried out on the target historical gas detection records obtained through a series of previous screening steps. In specific operations, various factors are comprehensively considered, such as the chronological order of detections, the similarity of the environmental conditions during detections to the current ones, the types and contents of gas components, etc. From numerous historical records, the part that can provide the most valuable reference for the current analysis is determined, that is, the first historical record. Subsequently, the module conducts a detailed analysis of this first historical record. It carefully sorts out various gas component information contained therein. The gas components involved are rich and diverse and have significant representativeness, covering hydrogen (H2), oxygen (O2), methane (CH4), ethylene (C2H6), ethane (C2H4), acetylene (C2H2), carbon monoxide (CO), and carbon dioxide (CO2), etc. For each gas component, not only its presence is simply recorded, but its absorption amount is accurately marked. This means that it can clearly tell us the specific quantity of each gas absorbed in a specific historical detection. The accurate acquisition and detailed recording of the absorption amounts of each gas component provide detailed basic data support for comprehensively, deeply, and accurately evaluating the state and performance of transformer oil in the subsequent process, predicting possible problems or faults, making the entire detection and analysis process more scientific, accurate, and reliable, and enhancing its adaptability.

[0028] Gas component set acquisition module 50, and the gas component set acquisition module 50 is used to perform a union operation on the multiple gas components with absorption amount identifiers to obtain a target gas component set.

[0029] Specifically, a union operation is performed on the multiple gas components with absorption amount identifiers obtained previously, and these gas components that appear in all different sources are summarized without omission and duplication. For example, if there are hydrogen, oxygen, and methane in some historical gas absorption information, ethylene, ethane, and carbon monoxide in others, and acetylene and carbon dioxide in some parts. The gas component set acquisition module 50 will integrate all the gas components covered by these different parts. After such an operation, a comprehensive target gas component set will finally be formed, and this set covers all the gas components that have ever appeared in the relevant detection records. It provides a broad and solid foundation for the subsequent in-depth analysis of gas components in transformer oil, the accurate determination of wavelength ranges, and precise gas detection work, effectively ensuring the comprehensiveness and accuracy of the detection, and helping to more effectively evaluate the state and performance of transformer oil.

[0030] Wavelength range acquisition module 60, and the wavelength range acquisition module 60 is used to read a predetermined wavelength strategy and analyze the target gas component set based on the predetermined wavelength strategy to obtain a target wavelength range.

[0031] Specifically, the main function of the wavelength interval acquisition module 60 is to read a predetermined wavelength strategy, which contains a series of rules and criteria on how to analyze the target gas component set. Based on these rules and criteria, in-depth analysis of the target gas component set is carried out. During the analysis process, various factors such as the absorption amount of gas components, the characteristics of absorption peaks, and the wavelength rules in historical data are considered. By comprehensively evaluating these factors, the wavelength interval that can most effectively detect and analyze the target gas components is finally determined, that is, the target wavelength interval. For example, the predetermined wavelength strategy stipulates that priority is given to considering the wavelength range of gas components with larger absorption amounts, or focuses on wavelength regions with specific absorption peak shapes and intensities. The wavelength interval acquisition module 60 will perform detailed calculations and comparisons on the target gas component set according to these regulations, so as to obtain an accurate and practical target wavelength interval. This target wavelength interval will provide crucial guidance for subsequent gas detection and analysis work, ensuring the accuracy and efficiency of detection.

[0032] The gas detection module 70 is used to activate the mid-infrared detection device to perform gas detection on the target transformer oil in the target wavelength interval.

[0033] Specifically, the gas detection module 70 exerts its control and coordination functions to accurately activate the mid-infrared detection device. The activation process here is not just simply starting the device, but also includes a series of initialization settings for the device to ensure that it is in the best detection preparation state. The activated mid-infrared detection device will perform gas detection on the target transformer oil according to the previously determined target wavelength interval. This target wavelength interval is obtained through careful analysis by multiple previous modules and is considered to be a specific wavelength range that can most effectively detect the target gas components. For example, assuming the target wavelength interval is determined to be from 1000 nanometers to 1500 nanometers, under the control of the gas detection module 70, the mid-infrared detection device will emit infrared light in this wavelength interval and receive the light signals reflected, absorbed, or scattered after passing through the target transformer oil. By analyzing and processing these light signals, the mid-infrared detection device can accurately detect key information such as the presence and concentration levels of various gas components in the target transformer oil. Such precise detection operations can provide accurate and reliable data support for evaluating the quality, performance, and potential failure risks of transformer oil, helping to take corresponding maintenance and improvement measures in a timely manner to ensure the normal operation and safety of the transformer.

[0034] In a possible implementation manner, the output information acquisition module 20 further includes:

[0035] Obtain the first transformer oil scenario record, where the first transformer oil scenario record includes the first scenario feature information of the first transformer oil scenario and the first transformer oil utility; obtain the utility prediction data set based on the first data set composed of the first scenario feature information and the first transformer oil utility; perform supervised learning and testing on the utility prediction data set to obtain the transformer oil utility prediction model.

[0036] Specifically, in the process of constructing the transformer oil utility prediction model, it is first necessary to obtain the first transformer oil scenario record. This record is a collection of rich information. The first scenario feature information therein is very detailed, covering specific data and descriptions in many aspects such as the geographical location of the transformer, climatic conditions (temperature, humidity, altitude, etc.), the type and specifications of the transformer, the working voltage and current levels, and the load change law. In terms of the first transformer oil utility, it clearly points out its specific performance in insulation, heat dissipation, and lubricant. In terms of insulation, the numerical value of the maximum voltage that the transformer oil can withstand without breakdown in a specific scenario will be detailedly recorded; in terms of heat dissipation, there will be efficiency data on the effective reduction of the internal temperature of the transformer by the transformer oil under different working loads and environmental temperatures; when used as a lubricant, it will cover its lubrication effect on relevant components, such as specific achievements in reducing friction loss and extending the service life of components.

[0037] Deeply analyze and sort out the first scenario feature information and the first transformer oil utility in the first data set. Clean these data to remove noise and outliers to ensure the quality and reliability of the data. Through data mining and feature engineering methods, extract more representative and predictive features from the original data. For example, for continuous variables such as temperature, humidity, and load in the scenario feature information, perform numerical transformation or discretization; for categorical variables such as insulation, heat dissipation, and lubricant in the transformer oil utility, perform encoding and quantization. Combine and construct these processed and extracted features to form a new data set. This new data set focuses more on reflecting the potential relationship and law between the scenario features and the transformer oil utility. The utility prediction data set obtained in this way provides a more targeted and valuable data basis for establishing an accurate and effective transformer oil utility prediction model in the future.

[0038] In the process of constructing a transformer oil utility prediction model, performing supervised learning on the generated utility prediction data set is a key step. Supervised learning means that we already know the input features (i.e., the first scenario feature information) and the corresponding output results (i.e., the transformer oil utility) of some samples in the data set. Using these known "answers", the learning algorithm will try to find the potential relationships and patterns between the input features and the output results. Based on various algorithms such as neural network algorithms and decision tree algorithms, it will automatically adjust the internal parameters and weights according to the samples in the data set to establish a model that can accurately predict the output results of unknown samples, and obtain a reliable model that can accurately predict the transformer oil utility.

[0039] In a possible implementation manner, the output information acquisition module 20 further includes:

[0040] Partition the utility prediction data set to obtain a data partition result, where the data partition result includes a first data set and a second data set; perform supervised training on the first data set based on the neural network principle to obtain a first utility prediction model; perform supervised training on the second data set based on the gradient boosting decision tree principle to obtain a second utility prediction model; the first utility prediction model analyzes the first scenario feature information to obtain a first predicted utility, the second utility prediction model analyzes the first scenario feature information to obtain a second predicted utility, and the first predicted utility and the second predicted utility form meta-prediction data; use the first transformer oil utility as the meta-predicted utility, and form a meta-data set with the meta-prediction data; perform training on the meta-data set to obtain the transformer oil utility prediction model.

[0041] Specifically, in the process of constructing a transformer oil utility prediction model, partitioning the utility prediction data set is an important preliminary step. This partitioning process will be carried out according to certain rules or strategies. For example, it can be randomly sampled to randomly divide the utility prediction data set into two parts, thus obtaining a first data set and a second data set. Or, it can be partitioned according to certain characteristics of the data. For example, according to different types of transformer oil usage scenarios (such as industrial scenarios, civilian scenarios), or according to the length of transformer oil usage time, the data is classified to obtain different data sets. The purpose of this partitioning is to be able to use different methods or models to perform targeted training on different data sets in the follow-up, so as to comprehensively utilize the advantages of multiple methods and improve the accuracy and generalization ability of the final transformer oil utility prediction model.

[0042] The first dataset is subjected to supervised training based on the principle of neural network to obtain the first utility prediction model. A neural network is a computational model that mimics the connection pattern between neurons in a biological brain. During the training process, each sample in the first dataset contains the input first scenario feature information and the true value (i.e., label) of the corresponding first transformer oil utility. A neural network consists of multiple neurons, which form different layers through complex connections, including an input layer, hidden layers, and an output layer. The input layer receives the first scenario feature information, and then the data undergoes a series of weighted calculations and non-linear transformations in the hidden layers. During training, the neural network makes predictions on the input data according to the current weights and parameters, and compares the prediction results with the true first transformer oil utility to calculate the error. Through the backpropagation algorithm, the error propagates backward from the output layer to the input layer, and the connection weights between neurons are adjusted according to the error to make the next prediction result closer to the true value. This process is repeated continuously, and the neural network gradually learns the complex relationship between the first scenario feature information and the first transformer oil utility, and continuously optimizes its own parameters until the error of the model on the training data reaches an acceptable level or the training reaches the preset number of iterations. The finally obtained fully trained neural network model is the first utility prediction model that can predict the corresponding first transformer oil utility according to the input new first scenario feature information.

[0043] The second dataset is subjected to supervised training based on the principle of gradient boosting decision tree, aiming to construct the second utility prediction model. Gradient boosting decision tree is an ensemble learning method that gradually improves the performance of the model by continuously constructing new decision trees to fit the residuals of the previous tree. During the training process, for each sample in the second dataset, it contains the input features (i.e., the first scenario feature information) and the corresponding true utility value (label). First, a simple model is initialized, usually a very shallow decision tree. Then, new decision trees are constructed in sequence, and the construction of each new tree is to fit the difference (i.e., the residual) between the prediction result of the previous model and the true label. By continuously adding new trees and adjusting their parameters, the predicted value of the entire model gradually approaches the true value. When constructing a decision tree, some criteria (such as information gain, Gini impurity, etc.) are used to select the optimal features and split points to maximize the prediction ability and purity of the tree. After multiple iterations and optimizations, when the error of the model on the second dataset reaches a satisfactory level or reaches the preset training stop condition, the training ends, and the second utility prediction model that can predict the utility according to the input first scenario feature information is obtained.

[0044] After obtaining the first utility prediction model and the second utility prediction model, the same first scenario feature information is input into these two models for analysis. The first utility prediction model will process and calculate the input first scenario feature information according to the patterns and relationships it has learned, so as to obtain the first predicted utility. At the same time, the second utility prediction model will also perform operations on the same first scenario feature information based on its own training results and internal mechanisms to obtain the second predicted utility. These two predicted utilities generated by different models, namely the first predicted utility and the second predicted utility, will be combined together to form meta-prediction data. In this way, by integrating the prediction results of the two models, it is possible to describe the utility situation of transformer oil in a specific scenario more comprehensively and accurately, providing richer and more reliable data support for subsequent analysis and decision-making.

[0045] Set the known first transformer oil utility as the meta-predicted utility. This actual utility value is obtained through precise measurement or authoritative verification, representing the true and exact utility performance of transformer oil in a specific scenario. Integrate this meta-predicted utility with the previously obtained meta-prediction data. The meta-prediction data is jointly composed of the first predicted utility and the second predicted utility respectively obtained by two different models, namely the first utility prediction model and the second utility prediction model. When we combine the meta-predicted utility and the meta-prediction data, a new and comprehensive meta-dataset is created. For example, the meta-prediction data contains numerical estimates (the first predicted utility and the second predicted utility) of the insulation performance and heat dissipation performance of transformer oil under specific temperature, humidity, and load conditions based on model predictions. The first transformer oil utility is the accurate record of the comprehensive utility such as insulation and heat dissipation obtained through strict experimental tests or long-term actual operation observations under exactly the same scenario conditions. Combining these two organically in the meta-dataset provides a rich data set that can not only reflect the model prediction results but also be directly compared with the actual utility. Through such a combination, it is possible to more intuitively discover the differences and similarities between the model predictions and the actual situation, thereby optimizing and improving the prediction model targeted, improving the accuracy and reliability of its predictions, and providing more powerful support for the performance evaluation and optimization of transformer oil.

[0046] After obtaining the meta-dataset containing meta-prediction utility and meta-prediction data, it is further trained to optimize and improve the transformer oil utility prediction model. During the training process, the model deeply analyzes various data relationships and patterns in the meta-dataset. By continuously adjusting the parameters and weights of the model, the model can better fit the data and improve the prediction accuracy. At the same time, training also helps the model reduce the situation of overfitting or underfitting. Overfitting will cause the model to overfit the training data and perform poorly on new data; underfitting cannot fully capture the rules in the data. Through the training of the meta-dataset, a balance can be found to enable the model to have good generalization ability. After sufficient training, a more accurate and reliable transformer oil utility prediction model is finally obtained, which can provide more valuable prediction results for practical applications.

[0047] In a possible implementation manner, the output information acquisition module 20 further includes:

[0048] The meta-dataset is subjected to supervised training to obtain a meta-predictor; the first utility prediction model and the second utility prediction model are used as initial predictors; the transformer oil utility prediction model is constructed by combining the initial predictor and the meta-predictor.

[0049] Specifically, in the process of constructing the transformer oil utility prediction model, supervised training of the meta-dataset to obtain a meta-predictor is a key step. Supervised training means that during the training process, for each sample in the meta-dataset, we clearly know its corresponding correct output (i.e., the true transformer oil utility value). First, prepare the meta-dataset, which contains various feature information related to the transformer oil scenario and the corresponding utility labels. Then, select a suitable machine learning algorithm or model architecture for training. Taking a neural network as an example, the training process is as follows: The data enters the network through the input layer, undergoes complex calculations and processing in the hidden layer, and finally produces a prediction result at the output layer. By continuously comparing the prediction result with the true utility label, a loss function (such as mean squared error, cross-entropy, etc.) is used to measure the gap between the predicted value and the true value. Based on the value of the loss function, the connection weights between neurons in the network are adjusted through the backpropagation algorithm to make the next prediction result closer to the true value. After multiple iterative trainings, the model gradually learns the complex relationship between the features and utilities in the meta-dataset, continuously optimizes its own parameters until the value of the loss function reaches a relatively small stable state or meets the preset stop condition. The finally obtained fully trained model is the meta-predictor that can accurately predict the utility based on the input new transformer oil scenario feature information.

[0050] In the overall framework of constructing a transformer oil utility prediction model, the previously trained first utility prediction model and second utility prediction model are designated as the initial predictors. The first utility prediction model and the second utility prediction model are trained based on different principles and data, and each has its specific advantages and scope of application. Designating them as the initial predictors means that when constructing a more perfect prediction model subsequently, these two existing models will be used as the starting point and foundation to improve the performance and accuracy of the overall prediction model.

[0051] In the process of constructing a transformer oil utility prediction model, the initial predictors (i.e., the first utility prediction model and the second utility prediction model) are built with the meta-predictor. This building process is not a simple combination but a carefully designed fusion process. A hierarchical structure can also be constructed. First, the initial predictors make a preliminary prediction, and then the meta-predictor corrects and optimizes based on the results of the initial predictors, so as to obtain the final transformer oil utility prediction. By reasonably building the initial predictors and the meta-predictor and giving full play to their respective advantages, a more accurate, reliable, and robust transformer oil utility prediction model can be constructed, providing more valuable prediction results for practical applications.

[0052] In a possible implementation manner, the gas component set acquisition module 50 further includes:

[0053] Extract the first gas component from the multiple gas components with absorption amount identifiers, where the first gas component corresponds to a first absorption amount; obtain a second historical record, where the second historical record includes second historical gas absorption information; match the second absorption amount of the first gas component in the second historical gas absorption information; take the average of the first absorption amount and the second absorption amount as the first target absorption amount of the first gas component; based on the first correspondence between the first gas component and the first target absorption amount, perform a union operation to obtain the target gas component set.

[0054] Specifically, when faced with multiple gas components with clear absorption amount identifiers, a targeted extraction operation is carried out to determine the first gas component therefrom. Each of these gas components has a specific absorption amount identifier, which means that the amount of each gas absorbing other substances or energy under specific conditions can be quantified and identified. The first gas component specially selected in this process necessarily corresponds to a specific absorption amount value, and the absorption amount associated with the first gas component is called the first absorption amount.

[0055] When conducting research or analysis related to gas absorption, a second historical record needs to be obtained. This record is the data accumulated and saved over a certain period in the past. The second historical record contains second historical gas absorption information, which details the absorption of various gases under specific conditions in the past.

[0056] After obtaining the second historical record containing rich gas absorption information, specific searching and matching operations need to be carried out in the second historical gas absorption information therein. Find the absorption amount corresponding to the first gas component extracted previously, and this absorption amount is called the second absorption amount. For example, assuming the first gas component is oxygen, then search for the absorption amount data related to oxygen in the second historical gas absorption information. This requires screening and comparing a large amount of data to accurately find the corresponding absorption amount value based on the identification of the gas component or a specific code. Through this matching operation, the first gas component of current concern can be associated with the corresponding information in the historical data, providing important data support for subsequent analysis and calculation.

[0057] After obtaining the first absorption amount of the first gas component and the second absorption amount matched in the second historical gas absorption information, in order to more accurately determine a comprehensive absorption amount of the first gas component, a method of taking the average value is adopted. Add the first absorption amount and the second absorption amount, and then divide by 2. The result obtained is their average value, and this average value is defined as the first target absorption amount of the first gas component. For example, assuming the first absorption amount is 80 units and the second absorption amount is 100 units, then their average value is (80 + 100) / 2 = 90 units, and this 90 units is the first target absorption amount of the first gas component. By taking the average value, the absorption amount obtained from the current measurement or calculation and the absorption amount in the historical data can be comprehensively considered, reducing the deviation and uncertainty of a single measurement or historical data, and making the obtained first target absorption amount more representative and reliable.

[0058] The first corresponding relationship between the first gas component and the first target absorption amount is clarified. Based on this one-to-one correspondence relationship, the target gas component set is constructed through the union operation. Take the first gas component and its corresponding first target absorption amount as a basic combination unit. In the union operation, other gas components and their corresponding target absorption amounts will also be integrated in the same way. For example, in addition to the first gas component, there are the second gas component, the third gas component, etc., and their respective corresponding target absorption amounts. By combining all these combinations of gas components and corresponding absorption amounts together, the final target gas component set is formed. This set contains all the gas components we are concerned about and the target absorption amount information obtained through calculation and processing, enabling us to comprehensively and systematically understand and analyze the absorption situation of these gas components.

[0059] In a possible implementation manner, the wavelength interval acquisition module 60 further includes:

[0060] Obtain a descending list of target gas components of the target gas component set based on the predetermined wavelength strategy; extract the components with a predetermined sorting threshold in the descending list of target gas components to obtain target screened gas components; extract the first gas from the target screened gas components, where the first gas has a first absorption peak; perform visualization processing on the first absorption peak to obtain a target visualized infrared wavelength map; determine the target wavelength interval according to the target visualized infrared wavelength map.

[0061] Specifically, according to the given predetermined wavelength strategy, that is, arranging in descending order according to the absorption amount of gas components, the components with more absorption amount are considered more critical and more in need of detection. For each gas component in the target gas component set, obtain its corresponding absorption amount data. Sort these gas components from largest to smallest according to the absorption amount. The obtained descending list of target gas components can clearly show which gas components have a larger absorption amount, thus highlighting the components that need to be focused on during the detection process. Such sorting helps to more effectively allocate detection resources and preferentially perform rapid and accurate detection and analysis on the gas components with more absorption amount and greater criticality.

[0062] After obtaining the descending list of target gas components arranged in descending order of absorption amount, it is necessary to extract the components that reach the predetermined sorting threshold from it to obtain the target screened gas components. The predetermined sorting threshold is a preset position or standard used to determine which components are screened out. For example, assume there are 10 gas components in the descending list of target gas components, and the predetermined sorting threshold is set to the top 5. Then, the top 5 gas components will be extracted and become the target screened gas components. Or, the predetermined sorting threshold is set with a specific value of the absorption amount. For example, the threshold is set to an absorption amount greater than or equal to 80 units, then all gas components with an absorption amount reaching or exceeding 80 units will be extracted as the target screened gas components. In this way, according to specific requirements and conditions, the gas components that need to be focused on can be selectively screened out from the descending list for further in-depth analysis and processing.

[0063] From the already determined target screened gas components, further select the first gas among them and define it as the first gas. The first gas has a first absorption peak. The absorption peak refers to the position where the absorption degree of a substance to light of a specific wavelength reaches the maximum value in spectral analysis. The extraction and identification of the first gas and its first absorption peak are of great significance for in-depth study of the characteristics of the gas, quantitative analysis, and related detection and monitoring work.

[0064] After determining the first absorption peak of the first gas, in order to more intuitively display and analyze the characteristics of this absorption peak, it is necessary to perform visualization processing on it. Using relevant data analysis and graph drawing techniques, convert and plot the data such as the wavelength and absorption intensity corresponding to the first absorption peak to form a clear and intuitive image, that is, the target visualization infrared wavelength map. Such visualization processing can help researchers more quickly and accurately understand the key characteristics such as the position, width, and intensity of the first absorption peak, thus providing an intuitive basis for further research and application. For example, in gas detection, the optimal detection wavelength range can be determined according to this wavelength map to improve the accuracy and efficiency of detection.

[0065] After obtaining the target visualization infrared wavelength map, it is necessary to carefully analyze it to determine the target wavelength range. Observe the shape and position of the absorption peak in the wavelength map. The width of the absorption peak and the slope changes on both sides provide important information. If the absorption peak is relatively sharp and obvious, select a certain wavelength range extending to both sides centered on the peak vertex as the target wavelength range. If the absorption peak is relatively wide or there are multiple sub-peaks, multiple factors need to be considered comprehensively to determine the target wavelength range. For example, pay attention to the area with higher absorption intensity, or delimit the range according to the experimental accuracy requirements and background interference conditions. It is also possible to compare and statistically analyze the wavelength maps of multiple similar samples to determine a more general and reliable target wavelength range. Through the above methods, accurately determine the target wavelength range according to the target visualization infrared wavelength map, providing a key parameter basis for subsequent gas detection, component analysis, etc.

[0066] In a possible implementation manner, the wavelength interval acquisition module 60 further includes:

[0067] Read a predetermined unit interval; based on the predetermined unit interval, segment the target visualization infrared wavelength map to obtain a target segmentation result; based on the number of absorption peaks, perform comparative analysis on multiple wavelength intervals in the target segmentation result to obtain the target wavelength interval.

[0068] Specifically, the predetermined unit interval is a fixed wavelength range interval value set in advance according to various factors. For example, this predetermined unit interval may be set to 50 nanometers, 100 nanometers, or other specific values. The operation of reading this predetermined unit interval is usually to obtain this preset value from a database, configuration file, or specific register that stores relevant settings. The determination of this value is based on various factors such as past experimental experience, the accuracy and resolution of the instrument, the characteristics of the gas to be detected, and the requirements for the accuracy and sensitivity of the detection results. For example, in the study of the infrared absorption characteristics of a certain specific gas, through multiple experiments, it is found that setting the unit interval to 80 nanometers can achieve a better balance between detection accuracy and efficiency. Therefore, 80 nanometers is used as the predetermined unit interval for reading and use in subsequent operations.

[0069] After obtaining the predetermined unit interval, it is applied to the segmentation operation of the target visualization infrared wavelength map. According to the size of the predetermined unit interval, starting from the starting wavelength of the wavelength map, it is divided in turn with the interval length as the step size. Each segmented interval contains information such as the wavelength and absorption intensity within the corresponding wavelength range. These segmented intervals combined together form the target segmentation result. Through this segmentation, the entire wavelength map can be decomposed into multiple smaller parts with a fixed interval length, so as to more carefully analyze and compare the characteristics of different wavelength intervals.

[0070] After obtaining multiple wavelength intervals in the target segmentation result, the next step is to conduct a comparative analysis of these intervals based on the number of absorption peaks. The number of absorption peaks is an important indicator for evaluating the gas absorption characteristics within each wavelength interval. A wavelength interval with a larger number of absorption peaks means that the gas has a more complex and significant absorption behavior of infrared light within that interval. When conducting the comparative analysis, the number of absorption peaks within each wavelength interval is counted one by one. For example, there are three wavelength intervals A, B, and C. Interval A has 2 absorption peaks, interval B has 1 absorption peak, and interval C has 3 absorption peaks. Interval C is relatively more important. By comparing these numbers, the wavelength interval with the largest number of absorption peaks is determined as the target wavelength interval. In this way, subsequent detection and analysis work can be carried out more targeted within this specific wavelength interval, improving the accuracy and efficiency of the detection.

[0071] Embodiment 2

[0072] Based on the same inventive concept as a scene-adaptive multi-component gas rapid detection device in the foregoing embodiment, as Figure 2 shown, the present application provides a scene-adaptive multi-component gas rapid detection method. The method in the embodiment of the present application and the device embodiment are based on the same inventive concept. Among them, the method includes:

[0073] Step S100: Collect multi-dimensional features of the target application scenario of the target transformer oil based on predetermined scenario features to obtain target scenario feature information.

[0074] Step S200: Use the target scenario feature information as input information for the transformer oil utility prediction model to obtain output information, where the output information includes the target utility.

[0075] Step S300: Screen the historical gas detection records of the same type of transformer oil of the target transformer oil based on the target utility to obtain target historical gas detection records.

[0076] Step S400: Extract the first historical record from the target historical gas detection records and obtain the first historical gas absorption information in the first historical record, where the first historical gas absorption information includes multiple gas components with absorption amount identifiers.

[0077] Step S500: Perform a union operation on the multiple gas components with absorption amount identifiers to obtain a target gas component set.

[0078] Step S600: Read a predetermined wavelength strategy and analyze the target gas component set based on the predetermined wavelength strategy to obtain a target wavelength range.

[0079] Step S700: Activate the mid-infrared detection device to detect the gas in the target transformer oil in the target wavelength range.

[0080] Further, step S200 further includes:

[0081] Obtain the first transformer oil scenario record, where the first transformer oil scenario record includes the first scenario feature information and the first transformer oil utility of the first transformer oil scenario; obtain a utility prediction data set according to the first data set composed of the first scenario feature information and the first transformer oil utility; perform supervised learning and verification on the utility prediction data set to obtain the transformer oil utility prediction model.

[0082] Further, step S200 further includes:

[0083] The utility prediction data set is partitioned to obtain a data partition result, which includes a first data set and a second data set; the first data set is supervised and trained based on the neural network principle to obtain a first utility prediction model; the second data set is supervised and trained based on the gradient boosting decision tree principle to obtain a second utility prediction model; the first utility prediction model analyzes the first scenario feature information to obtain a first predicted utility, and the second utility prediction model analyzes the first scenario feature information to obtain a second predicted utility. The first predicted utility and the second predicted utility form meta-prediction data; the first transformer oil utility is used as the meta-predicted utility, and the meta-prediction data forms a meta-data set; the meta-data set is trained to obtain the transformer oil utility prediction model.

[0084] Further, step S200 further includes:

[0085] The meta-data set is supervised and trained to obtain a meta-predictor; the first utility prediction model and the second utility prediction model are used as initial predictors; the transformer oil utility prediction model is constructed by the initial predictors and the meta-predictor.

[0086] Further, step S500 further includes:

[0087] Extract the first gas component from the multiple gas components with absorption amounts, and the first gas component corresponds to a first absorption amount; obtain a second historical record, which includes second historical gas absorption information; match the second absorption amount of the first gas component in the second historical gas absorption information; take the average of the first absorption amount and the second absorption amount as the first target absorption amount of the first gas component; based on the first correspondence between the first gas component and the first target absorption amount, perform a union operation to obtain the target gas component set.

[0088] Further, step S600 further includes:

[0089] Based on the predetermined wavelength strategy, obtain a target gas component descending list of the target gas component set; extract the components with a predetermined sorting threshold in the target gas component descending list to obtain target screened gas components; extract the first gas in the target screened gas components, and the first gas has a first absorption peak; perform visualization processing on the first absorption peak to obtain a target visualized infrared wavelength map; determine the target wavelength range according to the target visualized infrared wavelength map.

[0090] Further, step S600 further includes:

[0091] Read a predetermined unit interval; segment the target visual infrared wavelength map based on the predetermined unit interval to obtain a target segmentation result; perform a comparative analysis on multiple wavelength intervals in the target segmentation result based on the number of absorption peaks to obtain the target wavelength interval.

[0092] It should be noted that the above-mentioned sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above description of specific embodiments of this specification has been made. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0093] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included within the protection scope of the present application.

[0094] This specification and the drawings are only exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A scene - adaptive multi - component gas rapid detection device, characterized in that, Including: A scene feature information acquisition module, which collects multi-dimensional features of the target application scenario of the target transformer oil based on predetermined scene features to obtain target scene feature information; An output information acquisition module, which is used to use the target scene feature information as the input information of the transformer oil utility prediction model to obtain output information, where the output information includes the target utility; A historical gas detection record acquisition module, which screens the historical gas detection records of the same type of transformer oil of the target transformer oil based on the target utility to obtain target historical gas detection records; A first historical gas absorption information acquisition module, which is used to extract the first historical record in the target historical gas detection record and obtain the first historical gas absorption information in the first historical record, and the first historical gas absorption information includes multiple gas components with absorption amount identifiers; A gas component set acquisition module, which is used to perform a union operation on the multiple gas components with absorption amount identifiers to obtain a target gas component set; A wavelength range acquisition module, which is used to read a predetermined wavelength strategy and analyze the target gas component set based on the predetermined wavelength strategy to obtain a target wavelength range; A gas detection module, which is used to activate the mid-infrared detection device to detect the gas in the target transformer oil in the target wavelength range.

2. The multi-component gas rapid detection device with scene adaptability according to claim 1, characterized in that The output information acquisition module is further used for: Obtaining the first transformer oil scene record, which includes the first scene feature information and the first transformer oil utility of the first transformer oil scene; Obtaining a utility prediction data set according to a first data set composed of the first scene feature information and the first transformer oil utility; Performing supervised learning and testing on the utility prediction data set to obtain the transformer oil utility prediction model.

3. The multi-component gas rapid detection device with scene adaptability according to claim 2, wherein The output information acquisition module is further used for: Dividing the utility prediction data set to obtain a data division result, where the data division result includes a first data set and a second data set; Performing supervised training on the first data set based on the neural network principle to obtain a first utility prediction model; Performing supervised training on the second data set based on the gradient boosting decision tree principle to obtain a second utility prediction model; The first utility prediction model analyzes the first scene feature information to obtain a first predicted utility, the second utility prediction model analyzes the first scene feature information to obtain a second predicted utility, and the first predicted utility and the second predicted utility form meta-prediction data; Taking the first transformer oil utility as the meta-predicted utility and forming a meta-data set with the meta-prediction data; Training the meta-data set to obtain the transformer oil utility prediction model.

4. The multi-component gas rapid detection device with scene adaptability according to claim 3, wherein, The output information acquisition module is further used for: Performing supervised training on the meta-data set to obtain a meta-predictor; Taking the first utility prediction model and the second utility prediction model as initial predictors; The initial predictor and the meta-predictor are used to construct the transformer oil utility prediction model.

5. The multi-component gas rapid detection device with scene adaptability according to claim 1, characterized in that, The gas component set acquisition module is further configured to: Extract a first gas component from the multiple gas components with absorption amounts, where the first gas component corresponds to a first absorption amount; Obtain a second historical record, where the second historical record includes second historical gas absorption information; Match the second absorption amount of the first gas component in the second historical gas absorption information; Take the average of the first absorption amount and the second absorption amount as the first target absorption amount of the first gas component; Based on the first correspondence between the first gas component and the first target absorption amount, perform a union operation to obtain the target gas component set.

6. The multi-component gas rapid detection device with scene adaptability according to claim 1, characterized in that The wavelength interval acquisition module is further configured to: Based on the predetermined wavelength strategy, obtain a descending list of target gas components of the target gas component set; Extract the components with a predetermined sorting threshold from the descending list of target gas components to obtain target screened gas components; Extract a first gas from the target screened gas components, where the first gas has a first absorption peak; Perform visualization processing on the first absorption peak to obtain a target visualized infrared wavelength map; Determine the target wavelength interval according to the target visualized infrared wavelength map.

7. The multi-component gas rapid detection device with scene adaptability according to claim 6, characterized in that The wavelength interval acquisition module is further configured to: Read a predetermined unit interval; Based on the predetermined unit interval, segment the target visualized infrared wavelength map to obtain a target segmentation result; Based on the number of absorption peaks, perform comparative analysis on multiple wavelength intervals in the target segmentation result to obtain the target wavelength interval.

8. A method for rapid detection of multi-component gases with scene adaptability, characterized in that, The method is applied to the device according to any one of claims 1 to 7, and the method includes: Collect multi-dimensional features of the target application scenario of the target transformer oil based on predetermined scenario features to obtain target scenario feature information; Use the target scenario feature information as input information of the transformer oil utility prediction model to obtain output information, where the output information includes a target utility; Based on the target utility, screen the historical gas detection records of the same type of transformer oil as the target transformer oil to obtain target historical gas detection records; Extract a first historical record from the target historical gas detection records, and obtain first historical gas absorption information in the first historical record, where the first historical gas absorption information includes multiple gas components with absorption amounts; Perform a union operation on the multiple gas components with absorption amounts to obtain a target gas component set; Read a predetermined wavelength strategy, and analyze the target gas component set based on the predetermined wavelength strategy to obtain a target wavelength interval; Activate the mid-infrared detection device to perform gas detection on the target transformer oil in the target wavelength interval.

Images