Coking coal oil safety disposal intelligent monitoring method and system

Abstract

The invention relates to the technical field of coking coal oil safety monitoring, and discloses an intelligent monitoring method and system for coking coal oil safety disposal, and the method comprises the steps: collecting environment parameter data of a coking coal oil storage environment, and carrying out the preprocessing; performing parameter data offset estimation on the preprocessed coking coal oil environmental parameter data by using a parameter data offset detection model; correcting the preprocessed coking coal oil environmental parameter data according to a parameter data offset estimation result; performing feature extraction on the corrected coking coal oil environmental parameter data; and constructing a coking coal oil monitoring abnormity identification model to carry out coking coal oil environment monitoring. According to the method, the data sequence is divided into multiple batches, the offset estimation coefficient is calculated according to the extreme value difference of adjacent batches and the abnormal data value difference, data sequence correction is carried out, the continuity and stability of the data sequence are improved, feature vector extraction is carried out by adopting a feature decomposition mode combined with a genetic algorithm, and the accuracy of feature vector extraction is improved. And coking coal oil environment monitoring and safe disposal are realized.

Description

Technical Field

[0001] The present invention relates to the field of coking oil safety monitoring, and in particular to a coking oil safety disposal intelligent monitoring method and system. Background Art

[0002] Coal tar is an important industrial raw material, mainly produced by coal in the high-temperature distillation process, and is a by-product of the coal chemical industry. The chemical composition of coke oil is complex, including toxic and harmful substances such as benzene, naphthalene, phenol, pyridine, etc., and it has certain corrosiveness, flammability and toxicity. Therefore, in coal chemical enterprises, the safe disposal of coke oil has become a focus of attention. If not properly handled, the leakage, volatilization or combustion of coke oil will pose a great threat to the environment and personnel, especially during storage and transportation, which may cause major safety accidents. With the development of industrial intelligence, more and more intelligent technologies are being applied to the monitoring and management of hazardous chemicals. For example, the Internet of Things technology can realize real-time monitoring of the storage environment of coke oil, and artificial intelligence technology can predict potential safety hazards through data analysis and improve the safety management level of coke oil. In the storage and transportation environment of coke oil, key parameters such as temperature, humidity, and pressure have a direct impact on the safety of coke oil. As an important part of the intelligent monitoring system, the accuracy and reliability of sensors are crucial. At present, many sensor technologies are greatly affected by the environment in practical applications. For example, in high temperature, high humidity, and corrosive environments, the sensitivity of the sensor may decrease, or even cause measurement data deviations, causing the monitoring system to fail or give false alarms, thereby affecting the safety monitoring effect of coking oil. Summary of the invention

[0003] In view of this, the present invention provides an intelligent monitoring method for safe disposal of coking coal oil, which ensures that stable and accurate monitoring data can be provided in various complex environments through multi-level sensor data fusion analysis, thereby realizing real-time monitoring of the safety status of coking coal oil, early warning of hidden dangers, and rapid response to accidents.

[0004] To achieve the above object, the present invention provides a method for intelligent monitoring of safe disposal of coking oil, comprising the following steps:

[0005] S1: Collecting environmental parameter data of the coking oil storage environment and preprocessing it to obtain preprocessed coking oil environmental parameter data;

[0006] S2: construct a parameter data offset detection model, and use the parameter data offset detection model to estimate the parameter data offset of the preprocessed coking oil environmental parameter data to obtain the parameter data offset estimation result, wherein the data offset detection adopts a multi-batch data statistical analysis method;

[0007] S3: correcting the pre-processed coking oil environmental parameter data according to the parameter data offset estimation result to obtain corrected coking oil environmental parameter data;

[0008] S4: extracting features from the corrected coking oil environmental parameter data to obtain a coking oil environmental feature vector;

[0009] S5: Construct a coking oil monitoring anomaly recognition model to perform coking oil environmental monitoring. The coking oil monitoring anomaly recognition model takes the coking oil environmental feature vector as input and the monitoring anomaly recognition result as output, and performs safety disposal according to the monitoring anomaly recognition result. The coking oil monitoring anomaly recognition model uses a clustering algorithm to perform anomaly judgment.

[0010] As a further improvement method of the present invention:

[0011] Optionally, the S1 step includes:

[0012] A variety of sensors are used to collect environmental parameter data of the coking oil storage environment, wherein the environmental parameters include environmental temperature, environmental humidity, hydrostatic pressure at the bottom of the coking oil storage container, liquid level of the coking oil in the storage container, and concentration of benzene substances in the coking oil. In the embodiment of the present invention, the concentration of benzene substances is measured by gas chromatography, and the collected environmental parameter data is in the form of:

[0013] x={x u =(x u (1),x u (2),...,x u (n),...,x u (N))|u∈[1,5]}

[0014] in:

[0015] x represents environmental parameter data;

[0016] x u The data sequence representing the u-th environmental parameter, u∈[1,5], the first to fifth environmental parameters are the environmental temperature, the environmental humidity, the hydrostatic pressure at the bottom of the tar oil storage container, the liquid level of the tar oil in the storage container, and the concentration of benzene in the tar oil;

[0017] x u (1),x u (2),...,x u (n),...,x u (N) represents the data value of the u-th environmental parameter at N monitoring moments, n∈[1,N];

[0018] The environmental parameter data x is preprocessed to obtain the preprocessed coking oil environmental parameter data, wherein the preprocessing process is as follows:

[0019] S11: Filter the data sequence in the environmental parameter data x to obtain a filtered data sequence, where the data sequence x u The filtering result is x′ u , data value x u The filtering formula for (n) is:

[0020]

[0021] in:

[0022] x′ u (n) represents the data value x u (n) The filtering result;

[0023] exp(·) represents an exponential function with a natural constant as the base;

[0024] ε represents the position coefficient;

[0025] x′ u =(x′ u (1),x′ u (2),...,x′ u (n),...,x′ u (N));

[0026] S12: Calculate the distance between any two data values ​​in the filtered data sequence;

[0027] S13: Calculate the abnormal information coefficient of any data value in the filtered data sequence, and mark the data value whose abnormal information coefficient is higher than a preset threshold as an abnormal data value, and mark other data values ​​as non-abnormal data values;

[0028] S14: Select the maximum and minimum values ​​of the non-abnormal data values ​​from the filtered data sequence, and normalize all the data values ​​in the filtered data sequence to obtain a normalized data sequence, where the normalized data sequence of the u-th environmental parameter is X u =(X u (1),X u (2),...,X u (n),...,X u (N)), X u (n) represents the data value x u (n) The processing result after filtering and normalization;

[0029] All normalized data sequences are taken as the preprocessed tar oil environmental parameter data.

[0030] Optionally, constructing a parameter data offset detection model in step S2 includes:

[0031] Constructing a parameter data offset detection model, wherein the parameter data offset detection model takes the preprocessed coking oil environmental parameter data as input and takes the parameter data offset estimation result as output, and the parameter data offset detection model includes a sequence division layer, a difference calculation layer and an offset estimation layer;

[0032] The sequence partitioning layer is used to divide the normalized data sequence into continuous data batches;

[0033] The difference calculation layer is used to calculate the statistical differences between consecutive data batches and convert the statistical differences into offset probabilities;

[0034] The offset estimation layer is used to combine the offset probability and calculate the offset estimation result of each data batch.

[0035] Optionally, the parameter data offset detection model is used to estimate the parameter data offset of the preprocessed coking oil environmental parameter data to obtain the parameter data offset estimation result, including:

[0036] The parameter data offset detection model is used to estimate the parameter data offset of the preprocessed coke oil environmental parameter data, and the offset estimation results of any normalized data sequence are obtained. The offset estimation results of all normalized data sequences are integrated into the parameter data offset estimation results, where the normalized data sequence X u The calculation process of the offset estimation result is as follows:

[0037] S21: The sequence partitioning layer normalizes the data sequence X u Divide into consecutive batches of data:

[0038]

[0039] in:

[0040] represents the mth data batch obtained by division, wherein each data batch contains 3 consecutive normalized data values;

[0041] X u (3m) represents the data sequence X u The 3mth data value in;

[0042] S22: The difference calculation layer calculates the statistical differences between consecutive data batches, where the data batches and The statistical difference between

[0043]

[0044] in:

[0045] Represents data batches and The difference between the extreme values, max(·) means the maximum value is obtained through traversal, and min(·) means the minimum value is obtained through traversal;

[0046] Represents data batches and The difference in outliers between It indicates that the mean of abnormal data values ​​after normalization in the data batch is calculated;

[0047] S23: Convert the statistical differences of data batches into shift probabilities, where the statistical differences The conversion formula is:

[0048]

[0049] in:

[0050] Indicates statistical differences The corresponding offset probability;

[0051] S24: Combined with the offset probability, the offset estimation result of each data batch is calculated; wherein the data batch The offset estimation result is:

[0052]

[0053] in:

[0054] Represents data batches The offset estimation result of

[0055] δ(X u (3m)) represents the normalized data sequence X u Medium data value X u (3m) offset estimation coefficient;

[0056] Represents data batches The standard deviation of Represents data batches The mean of

[0057] S25: Integrate the offset estimation results of all data batches as the normalized data sequence X u The offset estimation result.

[0058] Optionally, the S3 step includes:

[0059] According to the parameter data offset estimation result, the pre-processed coking oil environmental parameter data is corrected to obtain the correction result of any normalized data sequence, and the correction result is used to form the corrected coking oil environmental parameter data, wherein the normalized data sequence X u The correction process is:

[0060] S31: Get normalized data sequence X u ;

[0061] S32: Normalized data sequence X u Any data value in is corrected, where the data value X u The correction formula for (n) is:

[0062] Y u (n) = X u (n)·δ(X u (n))

[0063] in:

[0064] Y u (n) represents the data value X u (n) The correction result;

[0065] δ(X u (n)) represents the data value X u (n) The corresponding offset estimation coefficient;

[0066] S33: Construct normalized data sequence X u The corrected result Y u =(Y u (1),Y u (2),...,Y u (n),...,Y u (N)).

[0067] Optionally, the S4 step includes:

[0068] The corrected coking oil environmental parameter data is subjected to feature extraction to obtain the coking oil environmental feature vector F, wherein the feature extraction process is as follows:

[0069] S41: The corrected coke oil environmental parameter data is used to form a data matrix C:

[0070] C=[Y1,Y2,Y3,Y4,Y5] T

[0071] in:

[0072] Y1, Y2, Y3, Y4, and Y5 represent the corrected data sequences corresponding to the five environmental parameters, and T represents transposition;

[0073] S42: Using a genetic algorithm to solve the influence weights of environmental parameters, weighting the data matrix C is processed to form a data weighted matrix C′:

[0074] C′=[w1Y1,w2Y2,w3Y3,w4Y4,w5Y5] T

[0075] in:

[0076] w1, w2, w3, w4, and w5 are the influence weights of the five environmental parameters respectively;

[0077] S43: Perform eigendecomposition on the data weighted matrix C′ to obtain the five largest eigenvalues ​​of the data weighted matrix C′ and the corresponding eigenvectors, and use the five eigenvectors to form a coking oil environment eigenvector F:

[0078] F=[α1,α2,α3,α4,α5]C

[0079] in:

[0080] α1, α2, α3, α4, and α5 represent the eigenvectors corresponding to the five largest eigenvalues ​​of the data weighted matrix C′.

[0081] Optionally, in step S42, solving the influence weights of the environmental parameters using a genetic algorithm includes:

[0082] S421: Initialize D groups of chromosomes, where each group of chromosomes has 5 gene segments, representing the influence weights of 5 environmental parameters respectively. The dth group of chromosomes generated by initialization is:

[0083]

[0084] in:

[0085] represents the dth group of chromosomes generated by initialization, Represents chromosome 5 genes in

[0086] S422: Set the current number of iterations of the chromosome to t, the initial value of t is 0, and the maximum value is Max, then the result of the tth iteration of the dth group of chromosomes is

[0087] S423: Setting the chromosome genetic objective function G1(·), substituting the iterated chromosome into the chromosome genetic objective function G1(·), and obtaining the chromosome genetic objective function value of the chromosome, where the chromosome The corresponding chromosome genetic objective function value is

[0088]

[0089] in:

[0090] Indicates the use of chromosomes The data matrix C is weighted and feature-decomposed to form the coking oil environment feature vector;

[0091] ||·|| represents the L1 norm, ||·||2 represents the L2 norm; in the embodiment of the present invention, express The richness of feature information in

[0092] S424: Set the chromosome mutation objective function G2(·), substitute the iterated chromosome into the chromosome mutation objective function G1(·), and obtain the chromosome mutation objective function value of the chromosome, where chromosome The corresponding chromosome mutation objective function value for:

[0093]

[0094] in:

[0095] Indicates the use of chromosomes A sequence of five eigenvectors obtained by weighting and eigendecomposing the data matrix C;

[0096] S425: Calculate the probability of chromosome mutation according to the chromosome mutation objective function value; The probability of mutation is:

[0097]

[0098] in:

[0099] Represents chromosome The probability of mutation to undergo mutation;

[0100] S426: extracting the mutated chromosome as the mutated chromosome, and the chromosome whose chromosome genetic objective function value is higher than that of the mutated chromosome as the genetic chromosome, and performing cross processing on some gene segments in the mutated chromosome and corresponding gene segments of the genetic chromosome;

[0101] S427: Let d=d+1, and return to step S425 until the maximum number of iterations is reached, and the chromosome with the highest chromosome genetic objective function value is taken as the solution result to extract the influence weight of the environmental parameters.

[0102] Optionally, in step S5, constructing a coking oil monitoring anomaly recognition model to perform coking oil environmental monitoring includes:

[0103] Constructing a coking oil monitoring anomaly recognition model, wherein the coking oil monitoring anomaly recognition model takes the extracted coking oil environment feature vector as input and takes the monitoring anomaly recognition result as output, wherein the coking oil monitoring anomaly recognition model includes an input layer, a feature convolution layer, a feature weight calculation layer and a mapping layer;

[0104] The input layer is used to receive the coking oil environment feature vector;

[0105] The feature convolution layer is a convolutional neural network model structure, which is used to perform multi-level convolution processing on the coke oil environment feature vector;

[0106] The feature weight calculation layer includes the attention unit and the residual unit in the residual network, which is used to calculate the weight of the convolution feature;

[0107] The mapping layer performs clustering and fusion processing on the convolution features based on the weights, and maps the clustering and fusion features into monitoring anomaly recognition results;

[0108] The coking oil monitoring anomaly recognition model is used to monitor the coking oil environment feature vector, and the coking oil environment monitoring process is as follows:

[0109] S51: the input layer receives the coking oil environment feature vector F;

[0110] S52: The feature convolution layer performs multi-level convolution processing on the coke oil environment feature vector F to obtain multi-level convolution features:

[0111] F(k)=W k×k *F,k∈[1,K]

[0112] in:

[0113] F(k) represents the convolution feature of the coking oil environment feature vector F at the k-scale level, and K represents the preset maximum scale level; W k×k represents the convolution weight matrix of k rows and k columns, and * represents the convolution operator;

[0114] S53: The feature weight calculation layer calculates the weight of the convolution feature, where the weight of F(k) is:

[0115]

[0116] in:

[0117] tanh(F(k-1)) represents the residual information at the k-1 scale level;

[0118] weight(k) represents the weight of F(k);

[0119] S54: The mapping layer performs weighted processing on the convolution features based on the weights, and clusters the weighted convolution features to obtain several clusters, extracts the cluster centers of each cluster, and splices them to obtain the cluster fusion feature G, and maps the cluster fusion feature G to the monitoring anomaly recognition result:

[0120]

[0121] in:

[0122] W represents the mapping matrix;

[0123] V represents the abnormal monitoring identification result. The higher the V value, the higher the leakage concentration of coke oil and the more serious the environmental pollution.

[0124] Take safe measures based on the monitoring anomaly identification results.

[0125] Optionally, the abnormal information coefficient of any data value in the filtered data sequence is calculated in step S13, including:

[0126] Data value x′ u The calculation process of the abnormal information coefficient of (n) is:

[0127] S131: Traverse to get x′ u The mid-range data value x′ u (n) The nearest s-th data value, and calculate the distance between the two as the data value x′ u (n) is the nearest neighbor distance, and x′ u The data value x′ u (n) The data value whose distance is less than the nearest neighbor distance constitutes the data value x′ u (n)’s nearest neighbor data set Near(x′ u (n));

[0128] S132: Calculate and obtain data value x′ u Density of (n):

[0129]

[0130] in:

[0131] ρ(x′ u (n)) represents the data value x′ u (n) density;

[0132] σ u Represents the data sequence x′ u The standard deviation of

[0133] Near(x′ u (n),e) represents the nearest neighbor data set Near(x′ u The e-th data value in (n));

[0134] S133: Calculate and obtain data value x′ u The abnormal information coefficient S(x′) of (n) u (n)):

[0135]

[0136] in:

[0137] ρ(Near(x′ u (n),e)) represents the nearest neighbor data set Near(x′ u The e-th data value Near(x′) in (n) u (n),e) density.

[0138] In order to solve the above problems, the present invention provides an intelligent monitoring system for safe disposal of coking oil, the system comprising:

[0139] A data acquisition module is used to collect environmental parameter data of the coking oil storage environment and perform preprocessing to obtain the preprocessed coking oil environmental parameter data, construct a parameter data offset detection model, use the parameter data offset detection model to estimate the parameter data offset of the preprocessed coking oil environmental parameter data, and correct the preprocessed coking oil environmental parameter data according to the estimated parameter data offset estimation result;

[0140] A feature extraction module is used to extract features from the corrected coking oil environmental parameter data to obtain a coking oil environmental feature vector;

[0141] The environmental monitoring device is used to construct a coking oil monitoring anomaly recognition model for coking oil environmental monitoring.

[0142] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0143] A memory storing at least one instruction;

[0144] Communication interface, enabling electronic equipment to communicate; and

[0145] The processor executes the instructions stored in the memory to implement the above-mentioned intelligent monitoring method for safe disposal of coking coal oil.

[0146] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned intelligent monitoring method for safe disposal of coke oil.

[0147] Compared with the prior art, the present invention proposes an intelligent monitoring method for safe disposal of coking oil, which has the following advantages:

[0148] Firstly, the scheme of the present invention adopts a variety of sensors to collect environmental parameter data of the coking oil storage environment, combines the density difference between the data value and other adjacent data values ​​in different data sequences, calculates the abnormal information coefficient of the data value, identifies the abnormal data value in the data sequence, and divides the data sequence into batches, calculates the extreme value difference and abnormal data value difference of adjacent batches, takes them as the statistical difference of data batches and converts them into offset probability. The greater the statistical difference between adjacent batches, the greater the probability that the data batch will be offset, and then generates offset estimation coefficients for different data values, realizes data offset estimation and correction processing under multiple batches, and improves the consistency and stability of the data sequence.

[0149] At the same time, this scheme adopts the method of feature decomposition to extract features from the corrected coking oil environmental parameter data, and in the process of feature decomposition, according to the amplitude of iterative update of the influence weight and the effect of iterative update, respectively construct the chromosome mutation objective function and the chromosome inheritance objective function, wherein the iterative update effect includes the similarity between the final coking oil environmental feature vector and the original data matrix and the richness of the characteristic information of the coking oil environmental feature vector. The chromosome is used as the influence weight of different environmental parameters, and the mutation probability and gene inheritance mode of different chromosomes are calculated according to the chromosome mutation objective function value and the chromosome inheritance objective function value, so as to reduce the randomness of chromosome selection in the genetic algorithm, improve the iteration efficiency, and use the optimal influence weight obtained by multiple iterations to extract the coking oil environmental feature vector with richer characteristic information, realize coking oil environmental monitoring, and carry out safety disposal according to the monitoring results. BRIEF DESCRIPTION OF THE DRAWINGS

[0150] Figure 1 A schematic diagram of a process for an intelligent monitoring method for safe disposal of coking coal oil provided in one embodiment of the present invention;

[0151] Figure 2 A functional module diagram of a coking oil safe disposal intelligent monitoring system provided by an embodiment of the present invention;

[0152] Figure 2In: 100 intelligent monitoring system for safe disposal of coke oil, 101 data acquisition module, 102 feature extraction module, 103 environmental monitoring device;

[0153] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0154] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0155] The embodiment of the present application provides an intelligent monitoring method for safe disposal of coking oil. The execution subject of the intelligent monitoring method for safe disposal of coking oil includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided in the embodiment of the present application. In other words, the intelligent monitoring method for safe disposal of coking oil can be executed by software or hardware installed in a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0156] Embodiment 1:

[0157] A method for intelligent monitoring of safe disposal of coking oil comprises the following steps:

[0158] S1: Collecting environmental parameter data of a coking oil storage environment and preprocessing it to obtain preprocessed coking oil environmental parameter data.

[0159] The S1 step includes:

[0160] A variety of sensors are used to collect environmental parameter data of the coking oil storage environment, where the environmental parameters include ambient temperature, ambient humidity, hydrostatic pressure at the bottom of the coking oil storage container, liquid level of the coking oil in the storage container, and concentration of benzene in the coking oil. The collected environmental parameter data are in the form of:

[0161] x={x u =(x u (1),x u (2),...,x u (n),...,x u (N))|u∈[1,5]}

[0162] in:

[0163] x represents environmental parameter data;

[0164] x uThe data sequence representing the u-th environmental parameter, u∈[1,5], the first to fifth environmental parameters are the environmental temperature, the environmental humidity, the hydrostatic pressure at the bottom of the tar oil storage container, the liquid level of the tar oil in the storage container, and the concentration of benzene in the tar oil;

[0165] x u (1),x u (2),...,x u (n),...,x u (N) represents the data value of the u-th environmental parameter at N monitoring moments, n∈[1,N];

[0166] The environmental parameter data x is preprocessed to obtain the preprocessed coking oil environmental parameter data, wherein the preprocessing process is as follows:

[0167] S11: Filter the data sequence in the environmental parameter data x to obtain a filtered data sequence, where the data sequence x u The filtering result is x′ u , data value x u The filtering formula for (n) is:

[0168]

[0169] in:

[0170] x′ u (n) represents the data value x u (n) The filtering result;

[0171] exp(·) represents an exponential function with a natural constant as the base;

[0172] ε represents the position coefficient;

[0173] x′ u =(x′ u (1),x′ u (2),...,x′ u (n),...,x′ u (N));

[0174] S12: Calculate the distance between any two data values ​​in the filtered data sequence;

[0175] S13: Calculate the abnormal information coefficient of any data value in the filtered data sequence, and mark the data value whose abnormal information coefficient is higher than a preset threshold as an abnormal data value, and mark other data values ​​as non-abnormal data values;

[0176] S14: Select the maximum and minimum values ​​of the non-abnormal data values ​​from the filtered data sequence, and normalize all the data values ​​in the filtered data sequence to obtain a normalized data sequence, where the normalized data sequence of the u-th environmental parameter is X u =(X u (1),X u (2),...,X u (n),...,X u (N)), X u (n) represents the data value x u (n) The processing result after filtering and normalization;

[0177] All normalized data sequences are taken as the preprocessed tar oil environmental parameter data.

[0178] S2: construct a parameter data offset detection model, and use the parameter data offset detection model to estimate the parameter data offset of the preprocessed coking oil environmental parameter data to obtain the parameter data offset estimation result.

[0179] The parameter data offset detection model is constructed in step S2, including:

[0180] Constructing a parameter data offset detection model, wherein the parameter data offset detection model takes the preprocessed coking oil environmental parameter data as input and takes the parameter data offset estimation result as output, and the parameter data offset detection model includes a sequence division layer, a difference calculation layer and an offset estimation layer;

[0181] The sequence partitioning layer is used to divide the normalized data sequence into continuous data batches;

[0182] The difference calculation layer is used to calculate the statistical differences between consecutive data batches and convert the statistical differences into offset probabilities;

[0183] The offset estimation layer is used to combine the offset probability and calculate the offset estimation result of each data batch.

[0184] The method of using the parameter data offset detection model to estimate the parameter data offset of the pre-processed coking oil environmental parameter data to obtain the parameter data offset estimation result includes:

[0185] The parameter data offset detection model is used to estimate the parameter data offset of the preprocessed coke oil environmental parameter data, and the offset estimation results of any normalized data sequence are obtained. The offset estimation results of all normalized data sequences are integrated into the parameter data offset estimation results, where the normalized data sequence X u The calculation process of the offset estimation result is as follows:

[0186] S21: The sequence partitioning layer normalizes the data sequence X u Divide into consecutive batches of data:

[0187]

[0188] in:

[0189] represents the mth data batch obtained by division, wherein each data batch contains 3 consecutive normalized data values;

[0190] X u (3m) represents the data sequence X u The 3mth data value in;

[0191] S22: The difference calculation layer calculates the statistical differences between consecutive data batches, where the data batches and The statistical difference between

[0192]

[0193] in:

[0194] Represents data batches and The difference between the extreme values, max(·) means the maximum value is obtained through traversal, and min(·) means the minimum value is obtained through traversal;

[0195] Represents data batches and The difference in outliers between It indicates that the mean of abnormal data values ​​after normalization in the data batch is calculated;

[0196] S23: Convert the statistical differences of data batches into shift probabilities, where the statistical differences The conversion formula is:

[0197]

[0198] in:

[0199] Indicates statistical differences The corresponding offset probability;

[0200] S24: Combine the offset probability and calculate the offset estimation result of each data batch; where data batch X u m The offset estimation result is:

[0201]

[0202] in:

[0203] Represents data batches The offset estimation result of

[0204] δ(X u (3m)) represents the normalized data sequence X u Medium data value X u (3m) offset estimation coefficient;

[0205] Represents data batches The standard deviation of Represents data batches The mean of

[0206] S25: Integrate the offset estimation results of all data batches as the normalized data sequence X u The offset estimation result.

[0207] S3: Correcting the pre-processed coking oil environmental parameter data according to the parameter data offset estimation result to obtain corrected coking oil environmental parameter data.

[0208] The S3 step includes:

[0209] According to the parameter data offset estimation result, the pre-processed coking oil environmental parameter data is corrected to obtain the correction result of any normalized data sequence, and the correction result is used to form the corrected coking oil environmental parameter data, wherein the normalized data sequence X u The correction process is:

[0210] S31: Get normalized data sequence X u ;

[0211] S32: Normalized data sequence X u Any data value in is corrected, where the data value X u The correction formula for (n) is:

[0212] Y u (n) = X u (n)·δ(X u (n))

[0213] in:

[0214] Y u (n) represents the data value X u (n) The correction result;

[0215] δ(X u(n)) represents the data value X u (n) The corresponding offset estimation coefficient;

[0216] S33: Construct normalized data sequence X u The corrected result Y u =(Y u (1),Y u (2),...,Y u (n),...,Y u (N)).

[0217] S4: Extract features from the corrected coking oil environmental parameter data to obtain a coking oil environmental feature vector.

[0218] The S4 step includes:

[0219] The corrected coking oil environmental parameter data is subjected to feature extraction to obtain the coking oil environmental feature vector F, wherein the feature extraction process is as follows:

[0220] S41: The corrected coke oil environmental parameter data is used to form a data matrix C:

[0221] C=[Y1,Y2,Y3,Y4,Y5] T

[0222] in:

[0223] Y1, Y2, Y3, Y4, and Y5 represent the corrected data sequences corresponding to the five environmental parameters, and T represents transposition;

[0224] S42: Using a genetic algorithm to solve the influence weights of environmental parameters, weighting the data matrix C is processed to form a data weighted matrix C′:

[0225] C′=[w1Y1,w2Y2,w3Y3,w4Y4,w5Y5] T

[0226] in:

[0227] w1, w2, w3, w4, and w5 are the influence weights of the five environmental parameters respectively;

[0228] S43: Perform eigendecomposition on the data weighted matrix C′ to obtain the five largest eigenvalues ​​of the data weighted matrix C′ and the corresponding eigenvectors, and use the five eigenvectors to form a coking oil environment eigenvector F:

[0229] F=[α1,α2,α3,α4,α5]C

[0230] in:

[0231] α1, α2, α3, α4, and α5 represent the eigenvectors corresponding to the five largest eigenvalues ​​of the data weighted matrix C′.

[0232] In step S42, a genetic algorithm is used to solve the influence weight of the environmental parameters, including:

[0233] S421: Initialize D groups of chromosomes, where each group of chromosomes has 5 gene segments, representing the influence weights of 5 environmental parameters respectively. The dth group of chromosomes generated by initialization is:

[0234]

[0235] in:

[0236] represents the dth group of chromosomes generated by initialization, Represents chromosome 5 genes in

[0237] S422: Set the current number of iterations of the chromosome to t, the initial value of t is 0, and the maximum value is Max, then the result of the tth iteration of the dth group of chromosomes is

[0238] S423: Setting the chromosome genetic objective function G1(·), substituting the iterated chromosome into the chromosome genetic objective function G1(·), and obtaining the chromosome genetic objective function value of the chromosome, where the chromosome The corresponding chromosome genetic objective function value is

[0239]

[0240] in:

[0241] Indicates the use of chromosomes The data matrix C is weighted and feature-decomposed to form the coking oil environment feature vector;

[0242] ||·|| represents the L1 norm, ||·||2 represents the L2 norm; in the embodiment of the present invention, express The richness of feature information in

[0243] S424: Set the chromosome mutation objective function G2(·), substitute the iterated chromosome into the chromosome mutation objective function G1(·), and obtain the chromosome mutation objective function value of the chromosome, where chromosome The corresponding chromosome mutation objective function value for:

[0244]

[0245] in:

[0246] Indicates the use of chromosomes A sequence of five eigenvectors obtained by weighting and eigendecomposing the data matrix C;

[0247] S425: Calculate the probability of chromosome mutation according to the chromosome mutation objective function value; The probability of mutation is:

[0248]

[0249] in:

[0250] Represents chromosome The probability of mutation to undergo mutation;

[0251] S426: extracting the mutated chromosome as the mutated chromosome, and the chromosome whose chromosome genetic objective function value is higher than that of the mutated chromosome as the genetic chromosome, and performing cross processing on some gene segments in the mutated chromosome and corresponding gene segments of the genetic chromosome;

[0252] S427: Let d=d+1, and return to step S425 until the maximum number of iterations is reached, and the chromosome with the highest chromosome genetic objective function value is taken as the solution result to extract the influence weight of the environmental parameters.

[0253] S5: Construct a coking oil monitoring anomaly recognition model to perform coking oil environmental monitoring. The coking oil monitoring anomaly recognition model takes the coking oil environmental feature vector as input and the monitoring anomaly recognition result as output, and performs safety disposal according to the monitoring anomaly recognition result.

[0254] In the step S5, a coking oil monitoring anomaly recognition model is constructed to perform coking oil environmental monitoring, including:

[0255] Constructing a coking oil monitoring anomaly recognition model, wherein the coking oil monitoring anomaly recognition model takes the extracted coking oil environment feature vector as input and takes the monitoring anomaly recognition result as output, wherein the coking oil monitoring anomaly recognition model includes an input layer, a feature convolution layer, a feature weight calculation layer and a mapping layer;

[0256] The input layer is used to receive the coking oil environment feature vector;

[0257] The feature convolution layer is a convolutional neural network model structure, which is used to perform multi-level convolution processing on the coke oil environment feature vector;

[0258] The feature weight calculation layer includes the attention unit and the residual unit in the residual network, which is used to calculate the weight of the convolution feature;

[0259] The mapping layer performs clustering and fusion processing on the convolution features based on the weights, and maps the clustering and fusion features into monitoring anomaly recognition results;

[0260] The coking oil monitoring anomaly recognition model is used to monitor the coking oil environment feature vector, and the coking oil environment monitoring process is as follows:

[0261] S51: the input layer receives the coking oil environment feature vector F;

[0262] S52: The feature convolution layer performs multi-level convolution processing on the coke oil environment feature vector F to obtain multi-level convolution features:

[0263] F(k)=W k×k *F,k∈[1,K]

[0264] in:

[0265] F(k) represents the convolution feature of the coking oil environment feature vector F at the k-scale level, and K represents the preset maximum scale level; W k×k represents the convolution weight matrix of k rows and k columns, and * represents the convolution operator;

[0266] S53: The feature weight calculation layer calculates the weight of the convolution feature, where the weight of F(k) is:

[0267]

[0268] in:

[0269] tanh(F(k-1)) represents the residual information at the k-1 scale level;

[0270] weight(k) represents the weight of F(k);

[0271] S54: The mapping layer performs weighted processing on the convolution features based on the weights, and clusters the weighted convolution features to obtain several clusters, extracts the cluster centers of each cluster, and splices them to obtain the cluster fusion feature G, and maps the cluster fusion feature G to the monitoring anomaly recognition result:

[0272]

[0273] in:

[0274] W represents the mapping matrix;

[0275] V represents the abnormal monitoring identification result. The higher the V value, the higher the leakage concentration of coke oil and the more serious the environmental pollution.

[0276] Take safe measures based on the monitoring anomaly identification results.

[0277] Embodiment 2:

[0278] like Figure 2 , which is a functional module diagram of a coking oil safe disposal intelligent monitoring system provided in one embodiment of the present invention, which can implement the coking oil safe disposal intelligent monitoring method in Example 1.

[0279] The intelligent monitoring system 100 for safe disposal of coking coal oil of the present invention can be installed in an electronic device. According to the functions to be implemented, the intelligent monitoring system for safe disposal of coking coal oil can include a data acquisition module 101, a feature extraction module 102 and an environmental monitoring device 103. The module of the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by a processor of an electronic device and can complete fixed functions, which are stored in the memory of the electronic device.

[0280] The data acquisition module 101 is used to collect environmental parameter data of the coking oil storage environment and perform preprocessing to obtain the preprocessed coking oil environmental parameter data, construct a parameter data offset detection model, use the parameter data offset detection model to estimate the parameter data offset of the preprocessed coking oil environmental parameter data, and correct the preprocessed coking oil environmental parameter data according to the estimated parameter data offset estimation result;

[0281] A feature extraction module 102 is used to extract features from the corrected coking oil environmental parameter data to obtain a coking oil environmental feature vector;

[0282] The environmental monitoring device 103 is used to construct a coking oil monitoring anomaly recognition model to perform coking oil environmental monitoring.

[0283] In detail, the modules in the intelligent monitoring system 100 for safe disposal of coke oil in the embodiment of the present invention are used in the same manner as described above. Figure 1 The same technical means are used as the intelligent monitoring method for safe disposal of coking coal oil described in the text and can produce the same technical effects, so they will not be repeated here.

[0284] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0285] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0286] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0287] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An intelligent monitoring method for safe disposal of coking oil, characterized in that: The method comprises: S1: Collecting environmental parameter data of the coking oil storage environment and preprocessing it to obtain preprocessed coking oil environmental parameter data; S2: constructing a parameter data offset detection model, and using the parameter data offset detection model to estimate the parameter data offset of the preprocessed coking oil environmental parameter data to obtain a parameter data offset estimation result; S3: correcting the pre-processed coking oil environmental parameter data according to the parameter data offset estimation result to obtain corrected coking oil environmental parameter data; S4: extracting features from the corrected coking oil environmental parameter data to obtain a coking oil environmental feature vector; S5: Construct a coking oil monitoring anomaly recognition model to perform coking oil environmental monitoring. The coking oil monitoring anomaly recognition model takes the coking oil environmental feature vector as input and the monitoring anomaly recognition result as output, and performs safety disposal according to the monitoring anomaly recognition result.

2. The intelligent monitoring method for safe disposal of coking oil according to claim 1, characterized in that: The S1 step includes: Collecting environmental parameter data of the coking oil storage environment, wherein the environmental parameters include environmental temperature, environmental humidity, hydrostatic pressure at the bottom of the coking oil storage container, liquid level of the coking oil in the storage container, and concentration of benzene in the coking oil; The collected environmental parameter data is in the form of: x={x u =(x u (1),x u (2),...,x u (n),...,x u (N))|u∈[1,5]} in: x represents environmental parameter data; x u The data sequence representing the u-th environmental parameter, u∈[1,5], the first to fifth environmental parameters are the environmental temperature, the environmental humidity, the hydrostatic pressure at the bottom of the tar oil storage container, the liquid level of the tar oil in the storage container, and the concentration of benzene in the tar oil; x u (1),x u (2),...,x u (n),...,x u (N) represents the data value of the u-th environmental parameter at N monitoring moments, n∈[1,N]; The environmental parameter data x is preprocessed to obtain the preprocessed coking oil environmental parameter data, wherein the preprocessing process is as follows: S11: Filter the data sequence in the environmental parameter data x to obtain a filtered data sequence, where the data sequence x u The filtering result is x ′ u ; S12: Calculate the distance between any two data values ​​in the filtered data sequence; S13: Calculate the abnormal information coefficient of any data value in the filtered data sequence, and mark the data value whose abnormal information coefficient is higher than a preset threshold as an abnormal data value, and mark other data values ​​as non-abnormal data values; S14: Select the maximum and minimum values ​​of the non-abnormal data values ​​from the filtered data sequence, and normalize all the data values ​​in the filtered data sequence to obtain a normalized data sequence, where the normalized data sequence of the u-th environmental parameter is X u =(X u (1),X u (2),...,X u (n),...,X u (N)), X u (n) represents the data value x u (n) The processing result after filtering and normalization; All normalized data sequences are taken as the preprocessed tar oil environmental parameter data.

3. A method for intelligent monitoring of safe disposal of coking oil according to claim 2, characterized in that: The parameter data offset detection model is constructed in step S2, including: Constructing a parameter data offset detection model, wherein the parameter data offset detection model takes the preprocessed coking oil environmental parameter data as input and takes the parameter data offset estimation result as output, and the parameter data offset detection model includes a sequence division layer, a difference calculation layer and an offset estimation layer; The sequence partitioning layer is used to divide the normalized data sequence into continuous data batches; The difference calculation layer is used to calculate the statistical differences between consecutive data batches and convert the statistical differences into offset probabilities; The offset estimation layer is used to combine the offset probability and calculate the offset estimation result of each data batch.

4. A method for intelligent monitoring of safe disposal of coking oil as claimed in claim 3, characterized in that: The method of using the parameter data offset detection model to estimate the parameter data offset of the pre-processed coking oil environmental parameter data to obtain the parameter data offset estimation result includes: The parameter data offset detection model is used to estimate the parameter data offset of the preprocessed coke oil environmental parameter data, and the offset estimation results of any normalized data sequence are obtained. The offset estimation results of all normalized data sequences are integrated into the parameter data offset estimation results, where the normalized data sequence X u The calculation process of the offset estimation result is as follows: S21: The sequence partitioning layer normalizes the data sequence X u Divide into consecutive batches of data: in: represents the mth data batch obtained by division, wherein each data batch contains 3 consecutive normalized data values; X u (3m) represents the data sequence X u The 3mth data value in; S22: The difference calculation layer calculates the statistical differences between consecutive data batches, where the data batches and The statistical difference between S23: Convert the statistical differences of data batches into shift probabilities, where the statistical differences The corresponding shift probability is S24: Combined with the offset probability, the offset estimation result of each data batch is calculated; wherein the data batch The offset estimation result is: in: Represents data batches The offset estimation result of δ(X u (3m)) represents the normalized data sequence X u Medium data value X u (3m) offset estimation coefficient; Represents data batches The standard deviation of Represents data batches The mean of S25: Integrate the offset estimation results of all data batches as the normalized data sequence X u The offset estimation result.

5. A method for intelligent monitoring of safe disposal of coking oil according to claim 4, characterized in that: The S3 step includes: The pre-processed coking oil environmental parameter data is corrected according to the parameter data offset estimation result to obtain the correction result of any normalized data sequence, and the correction result is used to form the corrected coking oil environmental parameter data, wherein the normalized data sequence X u The correction process is: S31: Get normalized data sequence X u ; S32: Normalized data sequence X u Any data value in is corrected, where the data value X u The correction formula for (n) is: Y u (n)=X u (n)·δ(X u (n)) in: Y u (n) represents the data value X u (n) The correction result; δ(X u (n)) represents the data value X u (n) The corresponding offset estimation coefficient; S33: Construct normalized data sequence X u The corrected result Y u =(Y u (1),Y u (2),...,Y u (n),...,Y u (N)).

6. A method for intelligent monitoring of safe disposal of coking oil according to claim 5, characterized in that: The S4 step includes: The corrected coking oil environmental parameter data is subjected to feature extraction to obtain the coking oil environmental feature vector F, wherein the feature extraction process is as follows: S41: Constructing a data matrix C with the corrected coking oil environmental parameter data; S42: Genetic algorithm is used to solve the influence weights of environmental parameters, and weighted processing is performed on the data matrix C to form a data weighted matrix C ′ ; S43: Weighting matrix C for data ′ Perform eigendecomposition to obtain the data weight matrix C ′ The five largest eigenvalues ​​and the corresponding eigenvectors are used to form the coking oil environment eigenvector F.

7. A method for intelligent monitoring of safe disposal of coking oil according to claim 6, characterized in that: In step S42, a genetic algorithm is used to solve the influence weight of the environmental parameters, including: S421: Initialize D groups of chromosomes, where each group of chromosomes has 5 segments of genes, representing the influence weights of 5 environmental parameters respectively. The dth group of chromosomes generated by initialization is d∈[1,D]; S422: Set the current number of iterations of the chromosome to t, the initial value of t is 0, and the maximum value is Max, then the result of the tth iteration of the dth group of chromosomes is S423: Setting the chromosome genetic objective function G1(·), substituting the iterated chromosome into the chromosome genetic objective function G1(·), and obtaining the chromosome genetic objective function value of the chromosome, where the chromosome The corresponding chromosome genetic objective function value is S424: Set the chromosome mutation objective function G2(·), substitute the iterated chromosome into the chromosome mutation objective function G1(·), and obtain the chromosome mutation objective function value of the chromosome, where chromosome The corresponding chromosome mutation objective function value is S425: According to the chromosome mutation objective function value of the chromosome, the mutation probability of the chromosome is calculated; wherein the chromosome The probability of mutation is: in: Represents chromosome The probability of mutation to undergo mutation; S426: extracting the mutated chromosome as the mutated chromosome, and the chromosome whose chromosome genetic objective function value is higher than that of the mutated chromosome as the genetic chromosome, and performing cross processing on some gene segments in the mutated chromosome and corresponding gene segments of the genetic chromosome; S427: Let d=d+1, return to step S425 until the maximum number of iterations is reached, and the chromosome with the highest chromosome genetic objective function value is used as the solution result, and the influence weight of the environmental parameters is extracted.

8. An intelligent monitoring method for safe disposal of coke kerosene as claimed in claim 1, characterized in that In the S5 step, a coke kerosene monitoring abnormal identification model is constructed for coke kerosene environmental monitoring, including: Constructing a coking oil monitoring anomaly recognition model, wherein the coking oil monitoring anomaly recognition model takes the extracted coking oil environment feature vector as input and takes the monitoring anomaly recognition result as output, wherein the coking oil monitoring anomaly recognition model includes an input layer, a feature convolution layer, a feature weight calculation layer and a mapping layer; The input layer is used to receive the coke kerosene environmental feature vector; The feature convolution layer is a convolutional neural network model structure, which is used to perform multi-level convolution processing on the coking kerosene environmental feature vectors; The feature weight calculation layer includes attention units and residual units in the residual network, which are used to calculate the weights of the convolutional features; The mapping layer performs clustering and fusion processing on the convolutional features based on weights, and maps the clustering and fusion features into monitoring exception recognition results; The coking kerosene environmental characteristic vector is used to monitor coking kerosene environmentally by using the coking kerosene monitoring anomaly identification model, and the coking kerosene environmental monitoring process is: S51: The input layer receives the coking kerosene environmental characteristic vector F; S52: The feature convolution layer performs multi-level convolution processing on the coking kerosene environment feature vector F to obtain multi-level convolution features: F(k)=W k×k *F,k∈[1,K] in: F(k) represents the convolutional feature of the coking kerosene environment characteristic vector F at the k-scale level, and K represents the preset maximum scale level; W k×k represents the convolution weight matrix of k rows and k columns, and * represents the convolution operator; S53: The feature weight calculation layer calculates the weight of the convolutional feature, where the weight of F(k) is weight(k); S54: The mapping layer performs weighted processing on the convolution features based on the weights, and clusters the weighted convolution features to obtain several clusters, extracts the cluster centers of each cluster, and splices them to obtain the cluster fusion feature G, and maps the cluster fusion feature G to the monitoring anomaly recognition result: in: W represents the mapping matrix; V represents the monitoring abnormal identification result. The higher the V value, the higher the leakage concentration of coke kerosene, and the more serious the environmental pollution is; Safe handling is carried out based on the monitoring abnormal identification results.

9. An intelligent monitoring method for safe disposal of coke kerosene as claimed in claim 2, characterized in that, In the S13 step, the abnormal information coefficient of any data value in the filtered data sequence is calculated, including: Data value x ′ u The calculation process of the abnormal information coefficient of (n) is: S131: Traverse to get x ′ u Middle distance data value x ′ u (n) The nearest s-th data value, and calculate the distance between the two as the data value x ′ u (n) is the nearest neighbor distance, and x ′ u The data value x ′ u (n) The data value whose distance is less than the nearest neighbor distance constitutes the data value x ′ u (n)’s nearest neighbor data set Near(x ′ u (n)); S132: Calculate the data value x ′ u The density ρ(x ′ u (n)); S133: Calculate the data value x ′ u The abnormal information coefficient S(x) of (n) ′ u (n)): in: ρ(Near(x ′ u (n),e)) represents the nearest neighbor data set Near(x ′ u The e-th data value Near(x ′ u (n),e) density.

10. An intelligent monitoring system for safe disposal of coking oil, characterized in that: The system comprises: A data acquisition module is used to collect environmental parameter data of the coking oil storage environment and perform preprocessing to obtain the preprocessed coking oil environmental parameter data, construct a parameter data offset detection model, use the parameter data offset detection model to estimate the parameter data offset of the preprocessed coking oil environmental parameter data, and correct the preprocessed coking oil environmental parameter data according to the estimated parameter data offset estimation result; Feature extraction module is used to feature extract the corrected coke kerosene environmental parameter data to obtain the coke kerosene environmental feature vector; An environmental monitoring device is used to construct a coking oil monitoring anomaly recognition model to perform coking oil environmental monitoring, and to safely dispose of the coking oil according to the monitoring and recognition results, so as to realize an intelligent monitoring method for safe disposal of coking oil as described in any one of claims 1-9.

Images