Petroleum attribute analysis method and system based on artificial intelligence

Abstract

The invention belongs to the technical field of petroleum attribute analysis, and discloses a petroleum attribute analysis method and system based on artificial intelligence. The method comprises the following steps: constructing a petroleum attribute analysis model, a residual prediction model and a network correction model; according to the real-time petroleum attribute data to be analyzed, using the petroleum attribute analysis model to obtain an initial real-time petroleum attribute predicted value; according to the real-time petroleum attribute data to be analyzed, using the residual prediction model to obtain a real-time residual prediction value; obtaining a final real-time petroleum attribute prediction value according to the initial real-time petroleum attribute prediction value and the real-time residual prediction value; and performing network correction on the petroleum attribute analysis model by using a network correction model according to the real-time residual prediction value to obtain a corrected petroleum attribute analysis model. According to the method, the problems of low efficiency, insufficient accuracy, low intelligent degree and lack of a model evolution mechanism in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of petroleum attribute analysis, and in particular relates to a petroleum attribute analysis method and system based on artificial intelligence. Background Art

[0002] Petroleum property analysis is an important part of the petroleum industry. By analyzing the types, components, quality and other information of petroleum products in petroleum properties, we can understand and predict the state and characteristics of petroleum, and provide an important basis for petroleum production, pricing and sales.

[0003] The existing petroleum property analysis technology has the following defects: Inefficiency: Existing technologies often rely on manual analysis of oil properties, which takes a lot of time and effort, is inefficient, and cannot meet the modern oil industry's demand for fast and accurate analysis. Insufficient accuracy: In the existing technology, manual analysis of oil properties is easily affected by subjective factors, lacks accuracy, and it is difficult to ensure the reliability of the analysis results; Low intelligence: With the development of artificial intelligence algorithms in the field of oil attribute analysis technology, technical solutions using artificial intelligence models for oil attribute analysis have emerged. Existing artificial intelligence models have low intelligence and cannot automatically learn and adapt to new data, and cannot meet the needs of the modern oil industry for intelligence and automation; Lack of model evolution mechanism: Existing artificial intelligence models lack a model evolution mechanism and cannot correct the model in real time, which affects the accuracy and stability of the model. Summary of the invention

[0004] In order to solve the problems of low efficiency, insufficient accuracy, low intelligence and lack of model evolution mechanism in the prior art, the present invention aims to provide a petroleum property analysis method and system based on artificial intelligence.

[0005] The technical solution adopted by the present invention is: A petroleum attribute analysis method based on artificial intelligence comprises the following steps: Based on some historical oil attribute data, artificial intelligence algorithms are used to build oil attribute analysis models, residual prediction models, and network correction models; According to the real-time petroleum attribute data to be analyzed, the petroleum attribute analysis model is used to perform petroleum attribute analysis to obtain an initial real-time petroleum attribute prediction value; According to the real-time petroleum attribute data to be analyzed, residual prediction is performed using a residual prediction model to obtain a real-time residual prediction value; Performing difference calculation based on the initial real-time oil attribute prediction value and the real-time residual prediction value to obtain the final real-time oil attribute prediction value; According to the real-time residual prediction value, the network correction model is used to perform network correction on the petroleum attribute analysis model to obtain the corrected petroleum attribute analysis model.

[0006] Furthermore, based on some historical petroleum attribute data, an artificial intelligence algorithm is used to construct a petroleum attribute analysis model, a residual prediction model, and a network correction model, including the following steps: Collecting some historical petroleum attribute data, and preprocessing some historical petroleum attribute data to obtain some preprocessed historical petroleum attribute data; Based on some pre-processed historical oil attribute data, an artificial intelligence algorithm is used to build an oil attribute analysis model and generate some historical residual data; Based on several historical petroleum attribute data and corresponding historical residual data, an artificial intelligence algorithm is used to build a residual prediction model; Based on a number of historical residual data, an artificial intelligence algorithm is used to build a network correction model and generate a number of historical network correction experiences.

[0007] Furthermore, the petroleum attribute analysis model is constructed based on the linear regression algorithm; The residual prediction model is built based on the decision tree algorithm; The network correction model is built based on the reinforcement learning algorithm.

[0008] Furthermore, the petroleum attribute analysis model is built based on deep learning algorithms; The residual prediction model is built based on the ensemble decision tree algorithm; The network correction model is built based on the reinforcement learning algorithm.

[0009] Furthermore, the petroleum attribute analysis model is constructed based on the LSTM-Elman algorithm; The residual prediction model is built based on the RF algorithm; The network correction model is built based on the DQN algorithm.

[0010] Furthermore, based on a number of pre-processed historical oil attribute data, an artificial intelligence algorithm is used to construct an oil attribute analysis model and generate a number of historical residual data, including the following steps: Performing feature engineering selection on the preprocessed historical petroleum attribute data, and setting a first input feature quantity and a first prediction feature quantity of the preprocessed historical petroleum attribute data; According to the first input feature quantity and the first predicted feature quantity, an initial petroleum attribute analysis model is constructed using the LSTM-Elman algorithm; The pre-processed historical petroleum attribute data were divided into a deep learning training set and a deep learning test set in a ratio of 7:3; Input the deep learning training set, optimize the initial petroleum attribute analysis model, obtain the optimized petroleum attribute analysis model, and store the corresponding historical network parameters and historical residual data; The deep learning test set is input to perform model testing on the optimized petroleum attribute analysis model to obtain the first model test accuracy; If the test accuracy of the first model is greater than the first accuracy threshold, the final petroleum attribute analysis model and some historical residual data are output; otherwise, the optimization training continues.

[0011] Furthermore, based on a number of historical petroleum attribute data and corresponding historical residual data, an artificial intelligence algorithm is used to construct a residual prediction model, including the following steps: According to the historical petroleum attribute data, the second characteristic input quantity is set, and according to the historical residual data, the second characteristic prediction quantity is set; According to the second feature input amount and the second feature prediction amount, an initial residual prediction model is constructed using the RF algorithm; Perform data association on a number of historical petroleum attribute data and corresponding historical residual data to obtain a number of residual prediction samples; Divide a number of residual prediction samples into an ensemble decision tree training set and an ensemble decision tree test set in a ratio of 7:3; Input the integrated decision tree training set, optimize the initial residual prediction model and obtain the optimized residual prediction model; Input the integrated decision tree test set to perform model testing on the optimized residual prediction to obtain the second model test accuracy; If the test accuracy of the second model is greater than the second accuracy threshold, the final residual prediction model is output, otherwise, the optimization training continues.

[0012] Furthermore, based on some historical residual data, an artificial intelligence algorithm is used to construct a network correction model and generate some historical network correction experiences, including the following steps: Generate questions based on the network correction strategy, set up the simulation environment for the DQN algorithm, and build the agent and experience replay pool; Collect the historical network parameters of the oil attribute analysis model corresponding to each historical residual data, use the historical network parameters as the state, define the state space of the DQN algorithm, and construct the input layer of the deep Q network; According to the differences between historical network parameters, set corresponding actions, define the action space of the DQN algorithm, and construct the output layer of the deep Q network; According to the input layer and the output layer, several hidden layers are set to construct the corresponding deep Q network, connecting the input layer to the state space and the output layer to the action space; Set the reward function of the DQN algorithm, and use the reward function to generate the corresponding reward value based on the historical residual data; Based on the state space, action space and corresponding reward values, the deep Q network and agent are optimized and trained, a network correction model is constructed, and several historical network correction experiences are generated; The generated historical network correction experiences are stored in the experience replay pool.

[0013] Furthermore, the historical petroleum attribute data includes historical petroleum product type parameters, historical petroleum composition parameters, historical petroleum quality parameters, and historical petroleum price index; The real-time petroleum attribute data includes real-time petroleum product type parameters, real-time petroleum composition parameters, real-time petroleum quality parameters and real-time petroleum price index.

[0014] An artificial intelligence-based petroleum attribute analysis system is used to implement a petroleum attribute analysis method. The system includes a model building unit, a petroleum attribute analysis unit, a residual prediction unit, a difference calculation unit and a network correction unit which are connected in sequence.

[0015] The beneficial effects of the present invention are: The present invention discloses an artificial intelligence-based petroleum attribute analysis method and system. By using artificial intelligence algorithms to construct a petroleum attribute analysis model, a residual prediction model and a network correction model, the automation and intelligence of petroleum attribute analysis are realized, and the analysis efficiency is greatly improved; automated and intelligent data analysis is performed, reducing the influence of manual experience and professional knowledge on the analysis results, and improving the objectivity and accuracy of the analysis results; a variety of artificial intelligence algorithms such as deep learning algorithms, integrated decision tree algorithms and reinforcement learning algorithms are integrated, and the automatic learning and adaptation of the model to new data are realized, and the intelligence level of the model is improved; a model evolution mechanism is implemented, and the petroleum attribute analysis model can be continuously iterated and evolved according to the execution of each petroleum attribute analysis task, so that the accuracy of the petroleum attribute analysis model is continuously improved.

[0016] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flowchart of the petroleum attribute analysis method based on artificial intelligence in the present invention.

[0018] Figure 2 It is a structural block diagram of the petroleum attribute analysis system based on artificial intelligence in the present invention. DETAILED DESCRIPTION

[0019] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.

[0020] Embodiment 1: like Figure 1 As shown, this embodiment provides a petroleum attribute analysis method based on artificial intelligence, comprising the following steps: S1: Based on some historical oil attribute data, using artificial intelligence algorithms, construct oil attribute analysis models, residual prediction models and network correction models, including the following steps: S1-1: collecting some historical petroleum attribute data, and preprocessing some historical petroleum attribute data to obtain some preprocessed historical petroleum attribute data; The historical petroleum attribute data include historical petroleum product type parameters, historical petroleum composition parameters, historical petroleum quality parameters and historical petroleum price index; The petroleum product category parameter includes parameter values ​​representing various types of petroleum products, each parameter value refers to a corresponding petroleum product, such as oil, kerosene, diesel, fuel oil, refinery dry gas, solvent oil, lubricating oil, naphtha, petroleum asphalt, etc.; Petroleum component parameters include physical and chemical property parameters, chemical composition, etc. Physical and chemical property parameters include density parameters, viscosity, freezing point parameters, boiling point range parameters, solubility parameters, etc. Chemical composition includes carbon content, hydrogen content, sulfur content, nitrogen content, oxygen content and trace metal element content, etc.; The oil price index includes the price evaluation index of oil prices in the collection period. The larger the oil price index is, the higher the oil price is in the oil trading market in the same period. Preprocessing includes format conversion, data cleaning, normalization, and data dimension reduction to improve data quality and provide data support for subsequent model training; S1-2: Based on some pre-processed historical oil attribute data, a deep learning algorithm is used to construct an oil attribute analysis model and generate some historical residual data, including the following steps: S1-2-1: performing feature engineering selection on the preprocessed historical petroleum attribute data, and setting a first input feature quantity and a first prediction feature quantity of the preprocessed historical petroleum attribute data; In this embodiment, the first input feature quantity is a petroleum product type parameter, a petroleum component parameter, and a petroleum quality parameter, and the first prediction feature quantity is a petroleum price index; S1-2-2: Based on the first input feature quantity and the first predicted feature quantity, an initial petroleum attribute analysis model is constructed using the LSTM-Elman algorithm; In this embodiment, the oil attribute analysis model is constructed based on the Long Short-Term Memory (LSTM) algorithm; the LSTM network can learn to retain important information in a long time series and ignore unimportant information through a gate mechanism, so that the LSTM network has significant advantages when processing sequence data. The LSTM network can capture the long-term dependencies in the data and extract the deep data features of the oil attribute data, thereby improving the accuracy of the oil attribute analysis. The Elman network can identify the deep data features of the oil attribute data through pre-training and perform oil attribute prediction; S1-2-3: Divide a number of pre-processed historical petroleum attribute data into a deep learning training set and a deep learning test set in a ratio of 7:3; S1-2-4: Input the deep learning training set, optimize the initial petroleum attribute analysis model, obtain the optimized petroleum attribute analysis model, and store the corresponding historical network parameters and historical residual data; S1-2-5: input the deep learning test set to perform model testing on the optimized petroleum attribute analysis model to obtain the first model test accuracy; S1-2-6: If the test accuracy of the first model is greater than the first accuracy threshold, the final petroleum attribute analysis model and a number of historical residual data are output; otherwise, optimization training is continued; S1-3: Based on several historical petroleum attribute data and corresponding historical residual data, an integrated decision tree algorithm is used to construct a residual prediction model, including the following steps: S1-3-1: Set the second feature input according to the historical petroleum attribute data, and set the second feature prediction according to the historical residual data; S1-3-2: Based on the second feature input and the second feature prediction, use the RF algorithm to construct an initial residual prediction model; In this embodiment, the residual prediction model is constructed based on the Random Forest (RF) algorithm; the residual prediction model screens the key features of the input petroleum attribute data through the internal Classification And Regression Tree (CART), extracts several key features related to residual prediction, and optimizes and trains the residual prediction model based on the several key features of the selected petroleum attribute data, using several historical petroleum attribute data and corresponding historical residual data, and adjusts the parameters of the residual prediction model, such as the number of trees, the depth of the tree, the feature selection method, etc., to optimize the model performance. The constructed residual prediction model can use the trained RF structure to perform residual prediction based on the new petroleum attribute data and several key features; S1-3-3: perform data association on a number of historical petroleum attribute data and corresponding historical residual data to obtain a number of residual prediction samples; S1-3-4: Divide a number of residual prediction samples into an ensemble decision tree training set and an ensemble decision tree test set in a ratio of 7:3; S1-3-5: Input the integrated decision tree training set, optimize the initial residual prediction model and obtain the optimized residual prediction model, including the following steps: S1-3-5-1: Use the RF structure of the initial residual prediction model to extract the feature contribution of M candidate features of the residual prediction samples in the integrated decision tree training set; The formula is: ; In the formula, For the Feature contribution of candidate features; For the The candidate features are in the random forest The feature contribution of each tree; is the CART tree indicator; is an alternative characteristic indicator; is the total number of CARTs; ; In the formula, CART tree node for random forest m ,node and nodes The Gini index of CART tree node m Medium Category The proportion of is the total number of categories; m , , is the node indicator; is the category indicator; S1-3-5-2: normalize the feature contributions of the M candidate features to obtain corresponding normalized feature contributions; The formula is: ; In the formula, It is the contribution of the feature after normalization; J is the total number of candidate features; S1-3-5-3: Generate feature selection standard values ​​for several candidate features based on the normalized feature contribution; The formula is: ; In the formula, For the Feature selection criterion values ​​for candidate features; For the The normalized feature contribution of the candidate features; is an alternative characteristic indicator; S1-3-5-4: According to the feature selection standard value, the candidate features are sorted in descending order, and the first M candidate features are selected as key features to obtain M key features; S1-3-5-5: Based on M key features and the integrated decision tree training set, the initial residual prediction model is optimized and trained to obtain an optimized residual prediction model; S1-3-6: Input the integrated decision tree test set to perform model testing on the optimized residual prediction to obtain the second model test accuracy; S1-3-7: If the test accuracy of the second model is greater than the second accuracy threshold, the final residual prediction model is output, otherwise, optimization training is continued; S1-4: Based on some historical residual data, use the reinforcement learning algorithm to build a network correction model and generate some historical network correction experience, including the following steps: S1-4-1: Generate questions based on the network correction strategy, set up the simulation environment of the DQN algorithm, and build the agent and experience replay pool; In this embodiment, the network correction model is constructed based on the Deep Q Network (DQN); the network correction model includes an agent, a deep Q network, and an experience replay pool. The agent is an entity that interacts with the model and the environment, and is used to select an action according to the current state and observe the state and reward after the action. The strategy of the agent is determined by learning the Q function. The deep Q network is used to approximate the Q function. The Q function is used to take the expected return of a specific action in a given state. The experience replay pool is used to store the historical experience of the interaction between the model and the environment. S1-4-2: Collect the historical network parameters of the oil attribute analysis model corresponding to each historical residual data, use the historical network parameters as the state, define the state space of the DQN algorithm, and construct the input layer of the deep Q network; S1-4-3: According to the differences between historical network parameters, set corresponding actions, define the action space of the DQN algorithm, and construct the output layer of the deep Q network; S1-4-4: According to the input layer and the output layer, set several hidden layers, build the corresponding deep Q network, connect the input layer to the state space, and connect the output layer to the action space; S1-4-5: Set the reward function of the DQN algorithm, and use the reward function to generate the corresponding reward value based on the historical residual data; S1-4-6: Based on the state space, action space and corresponding reward values, optimize the training of the deep Q network and the agent, build a network correction model, and generate some historical network correction experiences; S1-4-7: Store the generated historical network correction experiences into the experience playback pool; S2: Based on the real-time petroleum attribute data to be analyzed, using the petroleum attribute analysis model, petroleum attribute analysis is performed to obtain an initial real-time petroleum attribute prediction value, including the following steps: S2-1: collecting real-time petroleum attribute data to be analyzed, and preprocessing the real-time petroleum attribute data to obtain preprocessed real-time petroleum attribute data; In this embodiment, the real-time petroleum attribute data includes real-time petroleum product type parameters, real-time petroleum composition parameters, and real-time petroleum quality parameters; S2-2: input the preprocessed real-time petroleum attribute data into the petroleum attribute analysis model, and use the LSTM network to extract the real-time data features of the preprocessed real-time petroleum attribute data; S2-3: Based on the real-time data characteristics, the Elman network is used to analyze the oil properties and obtain the initial real-time oil property prediction value; S3: According to the real-time petroleum attribute data to be analyzed, residual prediction is performed using a residual prediction model to obtain a real-time residual prediction value, including the following steps: S3-1: Input the real-time data features of the pre-processed real-time petroleum attribute data extracted by the LSTM network into the residual prediction model; S3-2: Use the RF structure trained in the residual prediction model to perform residual prediction based on M key features to obtain real-time residual prediction values; S4: performing a difference operation based on the initial real-time oil attribute prediction value and the real-time residual prediction value to obtain a final real-time oil attribute prediction value; In this embodiment, the real-time oil attribute prediction value is the real-time oil price index prediction value, and the real-time oil price index prediction value is the price evaluation index within the collection period. The larger the real-time oil price index prediction value is, the higher the oil price prediction situation corresponding to the real-time oil attribute data to be analyzed is in the oil trading market of the same period. By comprehensively considering the real-time oil price index prediction value, it is possible to achieve accurate and reasonable pricing of the oil products corresponding to the real-time oil attribute data to be analyzed. S5: According to the real-time residual prediction value, the network correction model is used to perform network correction on the oil attribute analysis model to obtain the corrected oil attribute analysis model, including the following steps: S5-1: Extract the real-time network parameters of the petroleum attribute analysis model, use the real-time network parameters as the state, update the state space of the network correction model, and obtain the updated state space ,in, For the updated Status value, is the state indicator, is the total number of state space dimensions; S5-2: Extract some historical network correction experiences from the experience playback pool of the network correction model, and update the action space of the network correction model based on some historical network correction experiences to obtain an updated action space ,in, For the updated Action value, is the action indicator, is the total number of action space dimensions; S5-3: Use the agent to generate the corresponding real-time reward value based on the real-time residual prediction value and the reward function. The formula is: ; In the formula, To use the updated action The status Adjust to the updated status The normalized real-time reward value; A real-time reward value generated by using a reward function according to the real-time residual prediction value; is the reward mean and reward standard deviation; is the real-time residual prediction value; is the historical residual forecast value; It is a comprehensive indicator; S5-4: Based on the normalized real-time reward value, use the deep Q network to update the Q value of the possible actions in the updated action space to obtain the updated Q value of each possible action; The formula is: ; In the formula, In the updated state Next update action The corresponding updated Q value; For the status Next action The corresponding Q value of the possible action; is the learning rate; is the Q value of the highest possible action; To update the parameters; To use the updated action The status Adjust to the updated status The normalized real-time reward value; S5-5: Repeat the above Q-value updating steps until the number of iterations reaches the iteration number threshold, and use the greedy strategy to take the possible action with the highest updated Q-value as the execution action; S5-6: According to the execution action, the oil attribute analysis model is network-corrected to obtain a corrected oil attribute analysis model.

[0021] Embodiment 2: This embodiment is based on the technical solution of Embodiment 1, and the technical features different from Embodiment 1 are: As a preferred method, the petroleum attribute analysis model is constructed based on a linear regression algorithm; Linear regression algorithm refers to a predictive modeling technique that mainly studies the relationship between independent variables and dependent variables. It usually uses a line / curve to fit the data points and calculates the parameters to minimize the distance difference from the curve to the data points. A: Based on some pre-processed historical oil attribute data, a linear regression algorithm is used to build an oil attribute analysis model and generate some historical residual data, including the following steps: A-1: Construct a linear regression training sample set based on some pre-processed historical petroleum attribute data; A-2: Use linear regression algorithm to build an initial petroleum property analysis model; A-3: Input the linear regression training sample set into the initial petroleum attribute analysis model for training analysis to obtain the corresponding training analysis results; A-4: According to the training analysis results, the network parameters of the initial oil attribute analysis model are corrected and adjusted to obtain the final oil attribute analysis model, and the corresponding historical oil attribute prediction values ​​are generated; A-5: Generate corresponding historical residual data according to the historical petroleum attribute data and the corresponding historical petroleum attribute prediction values ​​output by the petroleum attribute analysis model; The residual prediction model is built based on the decision tree algorithm; B: Based on some historical petroleum attribute data and corresponding historical residual data, a residual prediction model is constructed using a decision tree algorithm, including the following steps: B-1: Use the decision tree algorithm to build an initial residual prediction model; B-2: Based on a number of historical petroleum attribute data and corresponding historical residual data, the network parameters of the initial residual prediction model are calibrated and adjusted to obtain the final residual prediction model.

[0022] Embodiment 3: like Figure 2As shown, this embodiment provides an artificial intelligence-based petroleum attribute analysis system for implementing a petroleum attribute analysis method, the system comprising a model building unit, a petroleum attribute analysis unit, a residual prediction unit, a difference calculation unit, and a network correction unit connected in sequence; A model building unit, used to build a petroleum attribute analysis model, a residual prediction model and a network correction model based on a number of historical petroleum attribute data using an artificial intelligence algorithm; The petroleum attribute analysis unit is used to perform petroleum attribute analysis based on the real-time petroleum attribute data to be analyzed using the petroleum attribute analysis model to obtain an initial real-time petroleum attribute prediction value; A residual prediction unit, used to perform residual prediction based on real-time petroleum attribute data to be analyzed using a residual prediction model to obtain a real-time residual prediction value; A difference calculation unit is used to perform a difference calculation based on the initial real-time oil attribute prediction value and the real-time residual prediction value to obtain a final real-time oil attribute prediction value; The network correction unit is used to perform network correction on the petroleum attribute analysis model according to the real-time residual prediction value using the network correction model to obtain the corrected petroleum attribute analysis model.

[0023] The present invention discloses an artificial intelligence-based petroleum attribute analysis method and system. By using artificial intelligence algorithms to construct a petroleum attribute analysis model, a residual prediction model and a network correction model, the automation and intelligence of petroleum attribute analysis are realized, and the analysis efficiency is greatly improved; automated and intelligent data analysis is performed, reducing the influence of manual experience and professional knowledge on the analysis results, and improving the objectivity and accuracy of the analysis results; a variety of artificial intelligence algorithms such as deep learning algorithms, integrated decision tree algorithms and reinforcement learning algorithms are integrated, and the automatic learning and adaptation of the model to new data are realized, and the intelligence level of the model is improved; a model evolution mechanism is implemented, and the petroleum attribute analysis model can be continuously iterated and evolved according to the execution of each petroleum attribute analysis task, so that the accuracy of the petroleum attribute analysis model is continuously improved.

[0024] The present invention is not limited to the above optional implementation modes, and anyone can derive other various forms of products under the enlightenment of the present invention.

[0025] The above specific implementation methods should not be understood as limiting the scope of protection of the present invention. The scope of protection of the present invention should be based on the definition in the claims, and the description can be used to interpret the claims.

Claims

1. A petroleum property analysis method based on artificial intelligence, characterized in that: The steps include: Based on some historical oil attribute data, artificial intelligence algorithms are used to build oil attribute analysis models, residual prediction models, and network correction models; According to the real-time petroleum attribute data to be analyzed, the petroleum attribute analysis model is used to perform petroleum attribute analysis to obtain an initial real-time petroleum attribute prediction value; According to the real-time petroleum attribute data to be analyzed, residual prediction is performed using a residual prediction model to obtain a real-time residual prediction value; Performing difference calculation based on the initial real-time oil attribute prediction value and the real-time residual prediction value to obtain the final real-time oil attribute prediction value; According to the real-time residual prediction value, the network correction model is used to perform network correction on the petroleum attribute analysis model to obtain the corrected petroleum attribute analysis model.

2. The petroleum property analysis method based on artificial intelligence according to claim 1, characterized in that: Based on some historical oil attribute data, an oil attribute analysis model, a residual prediction model and a network correction model are constructed using artificial intelligence algorithms, including the following steps: Collecting some historical petroleum attribute data, and preprocessing some historical petroleum attribute data to obtain some preprocessed historical petroleum attribute data; Based on some pre-processed historical oil attribute data, an artificial intelligence algorithm is used to build an oil attribute analysis model and generate some historical residual data; Based on several historical petroleum attribute data and corresponding historical residual data, an artificial intelligence algorithm is used to build a residual prediction model; Based on a number of historical residual data, an artificial intelligence algorithm is used to build a network correction model and generate a number of historical network correction experiences.

3. The petroleum property analysis method based on artificial intelligence according to claim 2, characterized in that: The petroleum attribute analysis model is constructed based on a linear regression algorithm; The residual prediction model is constructed based on a decision tree algorithm; The network correction model is constructed based on a reinforcement learning algorithm.

4. The petroleum property analysis method based on artificial intelligence according to claim 2, characterized in that: The petroleum attribute analysis model is constructed based on a deep learning algorithm; The residual prediction model is constructed based on an integrated decision tree algorithm; The network correction model is constructed based on a reinforcement learning algorithm.

5. The petroleum property analysis method based on artificial intelligence according to claim 4, characterized in that: The petroleum attribute analysis model is constructed based on the LSTM-Elman algorithm; The residual prediction model is constructed based on the RF algorithm; The network correction model is constructed based on the DQN algorithm.

6. The petroleum property analysis method based on artificial intelligence according to claim 5, characterized in that: Based on some pre-processed historical oil attribute data, an artificial intelligence algorithm is used to build an oil attribute analysis model and generate some historical residual data, including the following steps: Performing feature engineering selection on the preprocessed historical petroleum attribute data, and setting a first input feature quantity and a first prediction feature quantity of the preprocessed historical petroleum attribute data; According to the first input feature quantity and the first predicted feature quantity, an initial petroleum attribute analysis model is constructed using the LSTM-Elman algorithm; The pre-processed historical petroleum attribute data were divided into a deep learning training set and a deep learning test set in a ratio of 7:3; Input the deep learning training set, optimize the initial petroleum attribute analysis model, obtain the optimized petroleum attribute analysis model, and store the corresponding historical network parameters and historical residual data; The deep learning test set is input to perform model testing on the optimized petroleum attribute analysis model to obtain the first model test accuracy; If the test accuracy of the first model is greater than the first accuracy threshold, the final petroleum attribute analysis model and some historical residual data are output; otherwise, the optimization training continues.

7. The petroleum property analysis method based on artificial intelligence according to claim 5, characterized in that: Based on several historical petroleum attribute data and corresponding historical residual data, an artificial intelligence algorithm is used to construct a residual prediction model, which includes the following steps: According to the historical petroleum attribute data, the second characteristic input quantity is set, and according to the historical residual data, the second characteristic prediction quantity is set; According to the second feature input amount and the second feature prediction amount, an initial residual prediction model is constructed using the RF algorithm; Perform data association on a number of historical petroleum attribute data and corresponding historical residual data to obtain a number of residual prediction samples; Divide a number of residual prediction samples into an ensemble decision tree training set and an ensemble decision tree test set in a ratio of 7:3; Input the integrated decision tree training set, optimize the initial residual prediction model and obtain the optimized residual prediction model; Input the integrated decision tree test set to perform model testing on the optimized residual prediction to obtain the second model test accuracy; If the test accuracy of the second model is greater than the second accuracy threshold, the final residual prediction model is output, otherwise, the optimization training continues.

8. The petroleum property analysis method based on artificial intelligence according to claim 6, characterized in that: Based on some historical residual data, an artificial intelligence algorithm is used to build a network correction model and generate some historical network correction experiences, including the following steps: Generate questions based on the network correction strategy, set up the simulation environment for the DQN algorithm, and build the agent and experience replay pool; Collect the historical network parameters of the oil attribute analysis model corresponding to each historical residual data, use the historical network parameters as the state, define the state space of the DQN algorithm, and construct the input layer of the deep Q network; According to the differences between historical network parameters, set corresponding actions, define the action space of the DQN algorithm, and construct the output layer of the deep Q network; According to the input layer and the output layer, several hidden layers are set to construct the corresponding deep Q network, connecting the input layer to the state space and the output layer to the action space; Set the reward function of the DQN algorithm, and use the reward function to generate the corresponding reward value based on the historical residual data; Based on the state space, action space and corresponding reward values, the deep Q network and agent are optimized and trained, a network correction model is constructed, and several historical network correction experiences are generated; The generated historical network correction experiences are stored in the experience replay pool.

9. The petroleum property analysis method based on artificial intelligence according to claim 1, characterized in that: The historical petroleum attribute data include historical petroleum product type parameters, historical petroleum composition parameters, historical petroleum quality parameters and historical petroleum price index; The real-time petroleum attribute data includes real-time petroleum product type parameters, real-time petroleum composition parameters, real-time petroleum quality parameters and real-time petroleum price index.

10. An artificial intelligence-based petroleum property analysis system, used to implement the petroleum property analysis method according to any one of claims 1 to 9, characterized in that: The system comprises a model building unit, a petroleum attribute analysis unit, a residual prediction unit, a difference calculation unit and a network correction unit which are connected in sequence.

Images