A method and system for predicting raw height of a pressure-bearing guide rail

Abstract

The present application relates to a kind of pressure original guide high with height prediction method and system, specifically to the technical field of coal mine safety production.The method includes obtaining sample data;Sample data is divided into training sample and test sample;The correlation between height and the correlation factor value sequence is calculated using grey relational analysis method;According to the correlation degree, the correlation factor is filtered to obtain the filtered correlation factor;The weight of each filtered correlation factor is calculated using entropy weight method;According to the correlation factor value sequence corresponding to each filtered correlation factor, the standardized filtered correlation factor value sequence is obtained;The value of each index is calculated according to the standardized filtered correlation factor value sequence and weight;Based on PCA-PSO-SVR, the value of each index and test sample, the original guide high height prediction model is obtained.The present application can overcome the shortcomings of empirical formula and neural network in small sample prediction, and the result has high accuracy.

Description

Technical Field

[0001] This invention relates to the field of coal mine safety production technology, and in particular to a method and system for predicting the height of the original pressure-bearing guide belt. Background Technology

[0002] The original conduction zone refers to the height to which water in a confined aquifer rises upward along the fault fracture zone or split that runs through the aquifer and the above-ground impermeable layer under the action of water pressure and capillary negative pressure. It is called the original conduction zone because it is not affected by external human factors. The original high-conductivity zone has the following characteristics: ① The rock exhibits a plastic or semi-plastic state due to water dissolution; ② Its distribution is uneven and discontinuous, and its development is influenced by multiple factors, making it difficult to discern a regular pattern; ③ The rock strata with the original high-conductivity zone no longer possess water-blocking properties, their water-resisting capacity decreases, and they exhibit a certain degree of permeability; ④ The original high-conductivity zone of the aquitard has a significant reduction effect on the aquifer head, and the actual head pressure acting on the effective aquitard is much smaller than the aquifer head pressure; ⑤ Stress dissolution and capillary negative pressure are the driving forces for the development of the original high-conductivity zone, and capillary negative pressure is closely related to the surface tension coefficient of the liquid and the capillary diameter. That is, the larger the surface tension coefficient, the denser and finer the fractures, which is more conducive to the development of the original high-conductivity zone. With the deepening of coal mining year by year, the water hazard problem of the Ordovician limestone confined aquifer in the North China coalfield is becoming increasingly serious, with 10% of coal mines being threatened to varying degrees by water inrush from the confined aquifer. The development height of the original conductive strip is a factor that directly affects whether or not water inrush occurs in the floor. It plays a crucial role in judging the risk of water inrush in the floor. The greater the development height of the original conductive strip and the smaller the thickness of the aquitard, the greater the possibility of water inrush in the floor. Therefore, the accurate prediction method of the development height of the original conductive strip has a great contribution and significance to guiding the safe production of mines.

[0003] Patent 201410503595.2 discloses a method for predicting the original elevation zone of the base plate. First, a BP neural network model is established to divide the complexity of the geological structure into four levels: simple, relatively simple, relatively complex, and complex. Then, based on the geological and hydrological conditions of the prediction area, the water-rich areas and their water-rich properties in the prediction area are delineated. Next, based on the overlapping areas determined by the relatively complex and complex geological structures and the water-rich areas, the original elevation zone of the base plate in the prediction area is delineated. Finally, a BP neural network prediction model is established using MATLAB to predict the original elevation of the base plate in the delineated elevation zone. However, the BP neural network method has the following drawbacks: 1) Local minima problem: From a mathematical perspective, the traditional BP neural network is a local search optimization method. It is trying to solve a complex nonlinear problem. The weights of the network are gradually adjusted along the direction of local improvement. This will cause the algorithm to get stuck in local extrema, and the weights will converge to the local minimum point, resulting in network training failure. Furthermore, BP neural networks are highly sensitive to initial network weights. Initializing the network with different weights often leads to convergence to different local minima, which is the fundamental reason why many researchers obtain different results each time they train. 2) The convergence speed of the BP neural network algorithm is slow. 3) The choice of BP neural network structure varies. 4) The contradiction between application examples and network size. 5) The contradiction between the predictive ability and training ability of BP neural networks. 6) The sample dependence problem of BP neural networks. Moreover, the BP neural network prediction model in patent 201410503595.2 has too few influencing factors, is not comprehensive enough, and does not classify them according to the degree of influence of different influencing factors on the development height of the conduction band. Instead, it predicts the height based on the fact that the degree of influence of the selected influencing factors is the same, resulting in inaccurate height prediction. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for predicting the height of the original pressure conductor belt, which can effectively overcome the shortcomings of empirical formulas and neural networks in small sample prediction, and the results obtained have high accuracy.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A method for predicting the height of the original pressure-bearing conductor strip includes:

[0007] Acquire sample data, which includes the heights of multiple original guide zones and multiple sequences of related factor values; each sequence of related factor values ​​includes the values ​​of the same related factors for all original guide zones; the related factors include coal seam mining depth, unit water inflow from the pressure-bearing floor, aquifer thickness, floor permeability coefficient, working face slope length, advance speed, mining height, coal seam floor damage variables, fault strength index, fault fractal dimension, floor pressure, liquid surface tension coefficient, fracture coefficient, and density of water in the floor aquifer;

[0008] The sample data is divided into training samples and test samples;

[0009] For any original guide height band height and any related factor value sequence in the training sample, the grey relational analysis method is used to calculate the correlation degree between the original guide height band height and the related factor value sequence.

[0010] The relevant factors are filtered based on the correlation between the height of each original guide band and the value sequence of each relevant factor to obtain the filtered relevant factors;

[0011] The weights of each relevant factor after screening are calculated using the entropy weight method;

[0012] The fuzzy comprehensive evaluation method was used to perform dimensionless processing on the relevant factor value sequences corresponding to each screened relevant factor to obtain the standardized screened relevant factor value sequences.

[0013] The index system is determined based on the screened relevant factors, and the values ​​of each index in the index system corresponding to the height of each original guide band in the training sample are calculated based on the standardized value sequence of the screened relevant factors and the weight of each screened relevant factor.

[0014] Based on PCA-PSO-SVR, the values ​​of each indicator in the indicator system corresponding to the height of each original guide strip in the training samples, and the test samples, an original guide strip height prediction model is obtained. The original guide strip height prediction model is used to predict the original guide strip height.

[0015] Optionally, for any original guide height band height and any related factor value sequence in the training sample, the grey relational analysis method is used to calculate the correlation degree between the original guide height band height and the related factor value sequence, specifically including:

[0016] For any original guide height band height and any related factor value sequence in the training sample, the fuzzy comprehensive evaluation method is used to perform dimensionless processing on the original guide height band height and the related factor value sequence to obtain the standardized related factor value sequence and the standardized original guide height band height.

[0017] Calculate the absolute difference between the standardized sequence of relevant factor values ​​and the standardized original height of the guide band;

[0018] The correlation coefficient between the height of the original guide zone and the sequence of related factor values ​​is obtained based on the absolute difference.

[0019] The correlation coefficient is used to calculate the correlation degree between the height of the original guide zone and the sequence of related factor values.

[0020] Optionally, for any original guide height band height and any related factor value sequence in the training samples, the fuzzy comprehensive evaluation method is used to perform dimensionless processing on the original guide height band height and the related factor value sequence to obtain standardized related factor value sequences and standardized original guide height band heights, specifically including:

[0021] The average value of the relevant factor value sequence in the training sample and the average value of the original guide height strip in the training sample are calculated based on the value sequence of each relevant factor in the training sample and the height of each original guide height strip in the training sample.

[0022] Based on the average value of the relevant factor value sequence, the average height of the original guide band, the value sequence of each relevant factor in the training sample, and the height of each original guide band in the training sample, the standard deviation of the value sequence of the relevant factor in the training sample and the standard deviation of the height of the original guide band in the training sample are obtained.

[0023] For any original guide height band height and any related factor value sequence in the training sample, the related factor value sequence and the original guide height band height are standardized according to the average value of the related factor value sequence in the training sample, the average value of the original guide height band height in the training sample, the standard deviation of the related factor value sequence in the training sample, and the standard deviation of the original guide height band height in the training sample, to obtain the standardized related factor value sequence and the standardized original guide height band height.

[0024] Optionally, the method for obtaining the original guideline height prediction model based on PCA-PSO-SVR, the values ​​of each indicator in the indicator system corresponding to the height of each original guideline band in the training samples, and the test samples specifically includes:

[0025] Principal component analysis is used to calculate the weight of each index based on the value of each index in the index system corresponding to the height of each original guide height band in the training sample.

[0026] For any original guide height in the training sample, the values ​​of each indicator in the indicator system corresponding to the height of the original guide height are weighted according to the weight of each indicator to obtain the weighted indicator.

[0027] Using PSO-SVR, the original guide zone height prediction model is obtained based on the weighted indicators and the test samples.

[0028] Optionally, the PSO-SVR method is used to obtain the original guide zone height prediction model based on the weighted indicators and the test samples, specifically including:

[0029] Initialize the particle swarm; the particle swarm includes multiple sets of parameters for the SVR model; each set of parameters includes a penalty factor coefficient and a kernel function;

[0030] Substitute the weighted index into the SVR model corresponding to the parameter to obtain the fitness of the SVR model corresponding to the parameter.

[0031] Based on the test samples, a first judgment result is obtained by determining whether the target model is the optimal target model; the target model is the SVR model corresponding to the parameter with the highest fitness.

[0032] If the first judgment result is yes, then the target model is determined to be the original guide zone height prediction model;

[0033] If the first judgment result is negative, then update the particle swarm and return to the step "substitute the weighted index into the SVR model corresponding to the parameter to obtain the fitness of the SVR model corresponding to the parameter".

[0034] A system for predicting the height of a pressure-bearing original conductor belt includes:

[0035] The acquisition module is used to acquire sample data, which includes the heights of multiple original guide height zones and multiple sequences of related factor values. Each sequence of related factor values ​​includes the values ​​of the same related factors for all original guide height zones. The related factors include coal seam mining depth, unit water inflow from the pressure floor, aquifer thickness, floor permeability coefficient, working face slope length, advance speed, mining height, coal seam floor damage variables, fault strength index, fault fractal dimension, floor pressure, liquid surface tension coefficient, fracture coefficient, and density of water in the floor aquifer.

[0036] A training sample and test sample generation module is used to divide the sample data into training samples and test samples;

[0037] The correlation calculation module is used to calculate the correlation between the height of any original guide band and the sequence of related factor values ​​in the training sample using grey relational analysis.

[0038] The relevant factor filtering module is used to filter relevant factors based on the correlation between the height of each original guide band and the value sequence of each relevant factor to obtain the filtered relevant factors;

[0039] The weight calculation module is used to calculate the weights of each relevant factor after screening using the entropy weight method.

[0040] The standardization module is used to perform dimensionless processing on the sequence of relevant factor values ​​corresponding to each screened relevant factor using the fuzzy comprehensive evaluation method to obtain the standardized sequence of screened relevant factor values.

[0041] The indicator calculation module is used to determine the indicator system based on the screened relevant factors, and to calculate the value of each indicator in the indicator system corresponding to the height of each original guide height band in the training sample based on the standardized value sequence of the screened relevant factors and the weight of each screened relevant factor.

[0042] The original guide zone height prediction model determination module is used to obtain the original guide zone height prediction model based on PCA-PSO-SVR, the values ​​of each indicator in the indicator system corresponding to the height of each original guide zone in the training samples, and the test samples. The original guide zone height prediction model is used to predict the height of the original guide zone.

[0043] Optionally, the correlation calculation module specifically includes:

[0044] The standardization unit is used to perform dimensionless processing on the height of any original guide height band and the sequence of related factor values ​​in the training sample using the fuzzy comprehensive evaluation method, so as to obtain the standardized sequence of related factor values ​​and the standardized height of the original guide height band.

[0045] An absolute difference calculation unit is used to calculate the absolute difference between the standardized sequence of relevant factor values ​​and the height of the standardized original guide strip.

[0046] The correlation coefficient calculation unit is used to obtain the correlation coefficient between the height of the original guide band and the sequence of related factor values ​​based on the absolute difference.

[0047] The correlation calculation unit is used to calculate the correlation between the height of the original guide band and the sequence of related factor values ​​based on the correlation coefficient.

[0048] Optionally, the standardized unit specifically includes:

[0049] The average value calculation subunit is used to calculate the average value of the relevant factor value sequence in the training sample and the average value of the original guide height strip in the training sample based on the value sequence of each relevant factor in the training sample and the height of each original guide height strip in the training sample;

[0050] The standard deviation calculation subunit is used to obtain the standard deviation of the relevant factor value sequence in the training sample and the standard deviation of the original guide height band in the training sample based on the average value of the relevant factor value sequence, the average height of the original guide height band, the relevant factor value sequence in the training sample, and the height of the original guide height band in the training sample.

[0051] The standardization subunit is used to standardize the height of any original guide height band and any related factor value sequence in the training sample based on the average value of the related factor value sequence in the training sample, the average height of the original guide height band in the training sample, the standard deviation of the related factor value sequence in the training sample, and the standard deviation of the height of the original guide height band in the training sample, to obtain the standardized related factor value sequence and the standardized height of the original guide height band.

[0052] Optionally, the original guide strip height prediction model determination module specifically includes:

[0053] The weight calculation unit is used to calculate the weight of each indicator based on the value of each indicator in the indicator system corresponding to the height of each original guide height band in the training sample using the principal component analysis method.

[0054] The weighting unit is used to assign weights to the values ​​of each indicator in the indicator system corresponding to the height of any original guide height in the training sample according to the weights of each indicator, so as to obtain the weighted indicator.

[0055] The original guide zone height prediction model determination unit is used to obtain the original guide zone height prediction model based on the weighted index and the test sample using PSO-SVR.

[0056] Optionally, the original guide strip height prediction model determination unit specifically includes:

[0057] An initialization subunit is used to initialize the particle swarm; the particle swarm includes multiple sets of parameters for the SVR model; each set of parameters includes a penalty factor coefficient and a kernel function;

[0058] The fitness calculation subunit is used to substitute the weighted index into the SVR model corresponding to each group of parameters to obtain the fitness of the SVR model corresponding to each group of parameters.

[0059] The judgment subunit is used to determine whether the target model is the optimal target model based on the test sample to obtain a first judgment result; the target model is the SVR model corresponding to the parameter with the highest fitness.

[0060] The first result determination subunit is used to determine that the target model is the original guide zone height prediction model if the first judgment result is yes.

[0061] The second result determination subunit is used to update the particle swarm if the first judgment result is negative, and return to the step "substitute the weighted index into the SVR model corresponding to each group of parameters to obtain the fitness of the SVR model corresponding to each group of parameters".

[0062] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: Based on considering as many relevant factors as possible that affect the development of the original conductive strip, the present invention screens them and obtains a prediction model of the original conductive strip height based on PCA-PSO-SVR, so as to make the prediction process more consistent with reality, so as to overcome the shortcomings of empirical formulas and neural networks in small sample prediction, improve the final prediction accuracy, and have good feasibility in predicting the original conductive strip under pressure. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 A flowchart illustrating a method for predicting the height of a pressure-bearing original conductive strip, provided in an embodiment of the present invention;

[0065] Figure 2 The flowchart shows the index system for establishing a prediction model based on the MATLAB platform to screen factors that are strongly correlated with the original conduction height band.

[0066] Figure 3 Flowchart for establishing the original guide zone height prediction model;

[0067] Figure 4 A graph showing how fitness changes with the number of generations.

[0068] Figure 5 Fit a curve to the training samples;

[0069] Figure 6 A curve is fitted to the test sample. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0072] like Figure 1 As shown, this embodiment of the invention provides a method for predicting the height of the original pressure-bearing conductor strip, including:

[0073] Step 101: Obtain sample data, which includes the heights of multiple original guide height zones and multiple sequences of related factor values. Each sequence of related factor values ​​includes the values ​​of the same related factors for all original guide height zones; the related factors include coal seam mining depth, unit water inflow from the pressure-bearing floor, aquifer thickness, floor permeability coefficient, working face slope length, advance speed, mining height, coal seam floor damage variables, fault strength index, fault fractal dimension, floor pressure, liquid surface tension coefficient, fracture coefficient, and density of water in the floor aquifer.

[0074] Step 102: Divide the sample data into training samples and test samples.

[0075] Step 103: For any original height of the guide band and any sequence of related factor values ​​in the training sample, use grey relational analysis to calculate the correlation between the original height of the guide band and the sequence of related factor values.

[0076] Step 104: Based on the correlation between the height of each original guide band and the value sequence of each related factor, the related factors are filtered to obtain the filtered related factors.

[0077] Step 105: Calculate the weights of each relevant factor after screening using the entropy weight method.

[0078] Step 106: Use the fuzzy comprehensive evaluation method to perform dimensionless processing on the relevant factor value sequences corresponding to each screened relevant factor to obtain the standardized screened relevant factor value sequences.

[0079] Step 107: Determine the index system based on the screened relevant factors, and calculate the value of each index in the index system corresponding to the height of each original guide height band in the training sample based on the standardized value sequence of the screened relevant factors and the weight of each screened relevant factor.

[0080] Step 108: Based on PCA-PSO-SVR, the values ​​of each index in the index system corresponding to the height of each original guideway in the training samples, and the test samples, an original guideway height prediction model is obtained. This original guideway height prediction model is used to predict the height of the original guideway. The literature "Research on Annual Runoff Forecasting of Danjiangkou Reservoir Based on PCA-PSO-SVR" incorporates principal component analysis (PCA) and particle swarm optimization (PSO) algorithms into the SVR model, establishing a PCA-PSO-SVR forecasting model. Redundant information and noise are removed, the main features between factors are extracted, and the optimal parameter combination of the model is selected as the input to the regression support vector machine (SVR) model. The Danjiangkou Reservoir, the water source of the South-to-North Water Diversion Project, is selected as the study area, and the model is validated using inflow data from Danjiangkou from 1981 to 2016.

[0081] In practical applications, for any original guide height band height and any related factor value sequence in the training sample, the grey relational analysis method is used to calculate the correlation degree between the original guide height band height and the related factor value sequence, specifically including:

[0082] For any original height of the guide band and any sequence of related factor values ​​in the training samples, the fuzzy comprehensive evaluation method is used to perform dimensionless processing on the original height of the guide band and the sequence of related factor values ​​to obtain the standardized sequence of related factor values ​​and the standardized height of the original guide band.

[0083] Calculate the absolute difference between the standardized sequence of relevant factor values ​​and the height of the original standardized guide band.

[0084] The correlation coefficient between the height of the original guide zone and the sequence of related factor values ​​is obtained based on the absolute difference.

[0085] The correlation coefficient is used to calculate the correlation degree between the height of the original guide zone and the sequence of related factor values.

[0086] In practical applications, for any original guide height band height and any related factor value sequence in the training samples, the fuzzy comprehensive evaluation method is used to perform dimensionless processing on the original guide height band height and the related factor value sequence to obtain standardized related factor value sequences and standardized original guide height band heights, specifically including:

[0087] The average value of the relevant factor value sequence in the training sample and the average value of the original guide height strip in the training sample are calculated based on the value sequence of each relevant factor in the training sample and the height of each original guide height strip in the training sample.

[0088] Based on the average value of the relevant factor value sequence, the average height of the original guide band, the value sequence of each relevant factor in the training sample, and the height of each original guide band in the training sample, the standard deviation of the value sequence of the relevant factor in the training sample and the standard deviation of the height of the original guide band in the training sample are obtained.

[0089] For any original guide height band height and any related factor value sequence in the training sample, the related factor value sequence and the original guide height band height are standardized according to the average value of the related factor value sequence in the training sample, the average value of the original guide height band height in the training sample, the standard deviation of the related factor value sequence in the training sample, and the standard deviation of the original guide height band height in the training sample, to obtain the standardized related factor value sequence and the standardized original guide height band height.

[0090] In practical applications, the original guideline height prediction model, based on PCA-PSO-SVR, the values ​​of each indicator in the indicator system corresponding to the height of each original guideline band in the training samples, and the test samples, specifically includes:

[0091] Principal component analysis is used to calculate the weight of each index based on the value of each index in the index system corresponding to the height of each original guide height band in the training sample.

[0092] For any original height of the guide band in the training sample, the values ​​of each indicator in the indicator system corresponding to the height of the original guide band are weighted according to the weight of each indicator to obtain the weighted indicator.

[0093] Using PSO-SVR, the original guide zone height prediction model is obtained based on the weighted indicators and the test samples.

[0094] In practical applications, the PSO-SVR method, based on the weighted indicators and the test samples, yields the original guide zone height prediction model, specifically including:

[0095] Initialize the particle swarm; the particle swarm includes multiple sets of parameters for the SVR model; each set of parameters includes a penalty factor coefficient and a kernel function.

[0096] Substitute the weighted indices into the SVR models corresponding to each set of parameters to obtain the fitness of the SVR models corresponding to each set of parameters.

[0097] Based on the test samples, a first judgment result is obtained by determining whether the target model is the optimal target model; the target model is the SVR model corresponding to the parameter with the highest fitness.

[0098] If the first judgment result is yes, then the target model is determined to be the original guide zone height prediction model.

[0099] If the first judgment result is negative, then update the particle swarm (using formulas 21 and 22) and return to the step "substitute the weighted index into the SVR model corresponding to each group of parameters to obtain the fitness of the SVR model corresponding to each group of parameters".

[0100] This embodiment provides a more specific method for predicting the height of the original pressure-bearing conductor strip, and the specific steps are as follows:

[0101] Step 1: Using the MATLAB platform, screen for factors strongly correlated with the original conduction height band, and establish an index system for the prediction model based on these factors, such as... Figure 2 As shown, the specific process is as follows:

[0102] ① Identify areas in the study area with original high-conductivity zones and collect relevant sample data. The data for each sample includes the height of the original high-conductivity zone, the coal seam mining depth, the unit water inflow rate exposed in the pressure-bearing floor, the thickness of the aquifer, the floor permeability coefficient, the working face slope length, the advance speed, the mining height, the coal seam floor damage variables, the fault strength index, the fault fractal dimension, the floor pressure of the pressure water, the liquid surface tension coefficient, the fracture coefficient, and the density of the water in the floor aquifer.

[0103] ② Divide the collected sample data into training samples and test samples, and use fuzzy theory to perform dimensionless processing on each relevant factor and data of the training samples.

[0104] Let Y1, Y2, Y3, ..., Y n The original height sequence of the conduction band, Y n Y represents the height sequence of the nth original high-frequency band. n ={y n}, y n Represents the height of the nth original guide height band, X1, X2, X3, ..., X m Let X1 be the sequence of relevant factors, representing the coal seam mining depth sequence for all original elevation zones, X1 = {x1}, where x1 represents the coal seam mining depth value for all original elevation zones. X2 is the sequence of water inflow corresponding to all heights... and so on. i (1<i≤n), X j The time lengths for (1 < j ≤ m) are the same, meaning the number of data points is the same. Let Y be another variable. i Given the parent sequence and m related factor sequences X j It is a subsequence.

[0105] definition

[0106] in, These are the average values ​​of the original data for the relevant factor sequence and the original height sequence of the guide zone, respectively.

[0107]

[0108] Where s1 and s2 are the original data standard deviations of the relevant factor sequence and the original height sequence of the guide band, respectively.

[0109]

[0110] Where, x j ',y j These are the standardized sequences of the relevant factor sequence and the original guide height sequence, respectively.

[0111] ③ The grey relational analysis method was used to calculate the correlation degree between each relevant factor and the original height of the guide strip, and the relevant factors with a correlation degree greater than 0.8 were selected as the basis for establishing the model index system.

[0112] Calculate the correlation coefficient:

[0113] Let the standardized parent sequence be Y. i ', subsequence X j ', Y i 'with X j The absolute difference is Δ ij k, then:

[0114] Δ ij k = |Y i '(k)-X j '(k)| Formula 4

[0115] Among them, Y′ i (k) is the k-th element of the parent sequence, X′ j If (k) is the k-th element in the subsequence, then the parent sequence Y... i 'with subsequence X j The correlation coefficient of ' is:

[0116]

[0117] Where, let Δ max Indicates Δ ij The maximum value of k, Δ min Indicates Δ ij The minimum value of k, β is the standardized coefficient, usually β = 0.5.

[0118] Calculate the correlation:

[0119] Calculate the correlation between the parent sequence and the child sequence:

[0120]

[0121] In the formula, λ ij A represents the degree of correlation between the parent sequence and the child sequence; A represents the length of the comparison sequence.

[0122] ④ Calculate the weight of each relevant factor after screening using the entropy weight method.

[0123] This example uses the `weight` function in MATLAB to solve the problem. The calculation process is as follows:

[0124] Determine the number of evaluation objects and indicators, create a multi-object, multi-indicator evaluation matrix, and based on the relevant factors with a correlation degree greater than 0.8 selected in ③, establish the multi-object, multi-indicator evaluation matrix in ④:

[0125]

[0126] in accordance with The evaluation matrix R′ is dimensionless to obtain the standardized matrix R = (γ ij ) m×n .

[0127] in accordance with Normalizing the standardized matrix R yields the entropy value H of the i-th evaluation index. i :

[0128] Then the entropy weight ω of the i-th evaluation index i It can be represented as:

[0129]

[0130] ⑤ The screened relevant factors are dimensionless, and the weighted relevant factor data are divided into three indicators: water abundance index, tectonic index, and pressure index. The calculation method of the index is as follows:

[0131] E = QW Q +MW M +KW K Formula 12

[0132] G = SW S +FW F +DW D Formula 13

[0133] Y = PW P +MW M +σW σ Formula 14

[0134] In the formula: E represents the water-bearing index; G represents the structural index; Y represents the pressure index; Q, M, K, S, F, D, P, M, and σ represent the standardized values ​​of unit water inflow, aquifer thickness, floor permeability coefficient, coal seam floor damage variable, fault strength index, fault fractal dimension, floor confined water pressure, liquid surface tension coefficient, and fracture coefficient, respectively; W Q W M W K W S W F W D W P W M W σ These represent the weights of the aforementioned factors.

[0135] ⑥ The weights of the three indices—water abundance index, tectonic index, and pressure index—are determined using the contribution rates obtained from PCA. Specifically, this embodiment uses the PCA function on the Matlab platform to calculate the weights of the water abundance index, tectonic index, and pressure index. The calculation process is as follows:

[0136] Each sample is treated as a row vector, and the same discriminant index of multiple samples is combined vertically to form a sample matrix. Assuming there are n samples, each sample contains 3 discriminant indices, namely the water abundance index, the tectonic index, and the pressure index (corresponding to the height of the same original conduction zone), forming an n-row, 3-column sample matrix.

[0137] The sample matrix is ​​then standardized to eliminate the influence of dimensions. The standardization formula is as follows:

[0138]

[0139] Among them B i For normalized data; b i The original data before normalization (the i-th column of the sample matrix); min(b i ) represents the minimum value among all original data before normalization; max(b) i The first digit represents the maximum value among all original data before normalization; the final standardized sample matrix is: B = (B1, B2, ..., B...). P ) n×p .

[0140] The covariance matrix of the sample obtained is ∑=∑(s ij ) p×p Formula 16

[0141] in:

[0142] In the formula, s ij Let b be the element in the i-th row and j-th column of the covariance matrix.ki Let be the element in the k-th row and i-th column of the sample matrix B; Let be the average value of the i-th column of the sample matrix B. Let x be the average value of the j-th row of the sample matrix B. ki Let be the element in the k-th row and i-th column of the sample matrix B.

[0143] Using singular value decomposition, we can find the eigenvalues ​​λ of the sample data covariance matrix ∑. i and eigenvectors

[0144] Constructing a projection matrix using eigenvectors for eigenvalues ​​λ i Sort the data and select the eigenvectors corresponding to the eigenvalues ​​to construct the projection matrix:

[0145] In the formula, To sort the k-th eigenvalue λ k The corresponding eigenvector; Y is the projection matrix.

[0146] The contribution rate of each eigenvalue is calculated using the projection matrix. The calculation process is as follows:

[0147] The principal component composite score coefficient is determined as follows:

[0148]

[0149] The contribution rate corresponding to each eigenvalue is determined as follows:

[0150] The contribution rate of each calculated eigenvalue is used as the weight of the three indicators: water abundance index, tectonic index, and pressure index. The three indicators are then assigned weights.

[0151] Step 2: Establish the original guide zone height prediction model, such as Figure 3 The specific process shown is as follows:

[0152] ① Using the weighted water abundance index EW, tectonic index GW, pressure index YW and original conduction zone height as training sample data, complete the initial parameter setting of PSO, and use PSO to search for the optimal parameters C (penalty factor) and g (kernel function parameter) of SVR.

[0153] Specifically: Input the initial parameters and training sample data into the SVR model to calculate the fitness.

[0154] In the existing technology, SVR specifically refers to:

[0155] Let the data samples be n-dimensional vectors, and the training dataset be {(xi, yi), ..., (xl, yl)}, then the regression function used to fit the sample data is:

[0156] f(x)=ω×Φ(x)+b Formula 21

[0157] In the formula, the undetermined parameters ω and b represent the weight vector and bias, respectively.

[0158] Introducing C (C > 0), the above equation can be expressed as the following constrained optimization problem:

[0159]

[0160]

[0161] Where ε is the insensitive loss function; its function is to ignore the error within a certain range above and below the true value; ξ is the relaxation factor.

[0162] The above problem is a convex quadratic optimization problem. By introducing Lagrange multipliers, we can transform the constrained optimization problem into its dual problem, resulting in:

[0163]

[0164]

[0165] k(x i ,x)=exp{-|xx i | 2 / 2σ 2} Formula 26

[0166] 1 / σ 2 =g Formula 27

[0167] in, For kernel functions; α and α * For the corresponding support vectors.

[0168] Solving Equation 25 yields the corresponding α and α'. * The optimal fitting function can then be determined as:

[0169]

[0170] The collected test data was used as the raw data for model building. First, the initial population parameters were determined using the particle swarm optimization algorithm.

[0171] ② The test samples are fed into the established SVR model with the lowest fitness to calculate the mean squared error (MSE). Then, the learning sample fitting map is obtained through sample mapping calculation and linear fitting training. The learning effect is evaluated based on the mean squared error (MSE) of the fit. If the MSE approaches 0, the learning effect is better, indicating that the established model has good generalization ability and meets the requirements. This prediction model can then be used to predict the height of the original guide zone. When making predictions again, the predicted height of the original guide zone can be directly obtained by inputting the original data. If the average value of the MSE is large, the learning effect is poor. The original population is iteratively generated into a new population, the fitness is recalculated, and the model is retrained. This process is repeated until a good result is achieved. (Because each particle has a memory function, the effect will be better each time. A new model is obtained and tested again until the optimal model is obtained.)

[0172] The new parameter population generated by POS is specifically calculated according to the following formula:

[0173]

[0174]

[0175] c1 and c2 are acceleration factors, usually initially set to 1.49; r1 and r2 are rand random functions in the range [0, 1]; w is called the inertia weight, usually initially set to 0.8; and k represents the number of iterations.

[0176] Finally, after extensive training, the optimal parameters C and g are found. (C controls the error of the training samples; the larger C is, the better the training effect, but the lower the generalization ability. The width coefficient g reflects the correlation between support vectors. The larger g is, the looser the relationship between support vectors. The smaller g is, the greater the influence between support vectors.)

[0177] like Figure 4 As shown in the figure, the curve of fitness changing with the number of generations in the PSO (Particle Swarm Optimization)-SVR (Support Vector Regression) prediction model can better observe the changes in fitness during the iteration process.

[0178] like Figure 5 As shown, the training sample fitting curve in the PSO (Particle Swarm Optimization)-SVR (Support Vector Regression) prediction model can better observe the error between the true value and the predicted value of the training sample.

[0179] like Figure 6 As shown, the test sample fitting curve in the PSO (Particle Swarm Optimization)-SVR (Support Vector Regression) prediction model can better observe the error between the true value and the predicted value of the test sample.

[0180] This invention also provides a prediction system for the original height of the pressure-bearing conductor strip corresponding to the above method, comprising:

[0181] The acquisition module is used to acquire sample data, which includes the heights of multiple original guide height zones and multiple sequences of related factor values. Each sequence of related factor values ​​includes the values ​​of the same related factors for all original guide height zones. The related factors include coal seam mining depth, unit water inflow rate exposed in the pressure-bearing floor, aquifer thickness, floor permeability coefficient, working face slope length, advance speed, mining height, coal seam floor damage variables, fault strength index, fault fractal dimension, floor pressure, liquid surface tension coefficient, fracture coefficient, and density of water in the floor aquifer.

[0182] The training sample and test sample generation module is used to divide the sample data into training samples and test samples.

[0183] The correlation calculation module is used to calculate the correlation between the height of any original guide band and the sequence of related factor values ​​in the training samples using grey relational analysis.

[0184] The relevant factor filtering module is used to filter relevant factors based on the correlation between the height of each original guide band and the value sequence of each relevant factor to obtain the filtered relevant factors.

[0185] The weight calculation module is used to calculate the weights of each relevant factor after screening using the entropy weight method.

[0186] The standardization module is used to perform dimensionless processing on the sequence of relevant factor values ​​corresponding to each screened relevant factor using the fuzzy comprehensive evaluation method to obtain the standardized sequence of screened relevant factor values.

[0187] The indicator calculation module is used to determine the indicator system based on the screened relevant factors, and to calculate the value of each indicator in the indicator system corresponding to the height of each original guide height band in the training sample based on the standardized value sequence of the screened relevant factors and the weight of each screened relevant factor.

[0188] The original guide zone height prediction model determination module is used to obtain the original guide zone height prediction model based on PCA-PSO-SVR, the values ​​of each indicator in the indicator system corresponding to the height of each original guide zone in the training samples, and the test samples. The original guide zone height prediction model is used to predict the height of the original guide zone.

[0189] As an optional implementation, the correlation calculation module specifically includes:

[0190] The standardization unit is used to perform dimensionless processing on the height of any original guide height band and the sequence of related factor values ​​in the training samples using the fuzzy comprehensive evaluation method, so as to obtain the standardized sequence of related factor values ​​and the standardized height of the original guide height band.

[0191] The absolute difference calculation unit is used to calculate the absolute difference between the standardized sequence of relevant factor values ​​and the height of the standardized original guide strip.

[0192] The correlation coefficient calculation unit is used to obtain the correlation coefficient between the height of the original guide band and the sequence of related factor values ​​based on the absolute difference.

[0193] The correlation calculation unit is used to calculate the correlation between the height of the original guide band and the sequence of related factor values ​​based on the correlation coefficient.

[0194] As an optional implementation, the standardized unit specifically includes:

[0195] The average value calculation subunit is used to calculate the average value of the relevant factor value sequence in the training sample and the average value of the original guide height band in the training sample based on the value sequence of each relevant factor in the training sample and the height of each original guide height band in the training sample.

[0196] The standard deviation calculation subunit is used to obtain the standard deviation of the relevant factor value sequence in the training sample and the standard deviation of the original guide height band in the training sample based on the average value of the relevant factor value sequence, the average height of the original guide height band, the relevant factor value sequence in the training sample, and the height of the original guide height band in the training sample.

[0197] The standardization subunit is used to standardize the height of any original guide height band and any related factor value sequence in the training sample based on the average value of the related factor value sequence in the training sample, the average height of the original guide height band in the training sample, the standard deviation of the related factor value sequence in the training sample, and the standard deviation of the height of the original guide height band in the training sample, to obtain the standardized related factor value sequence and the standardized height of the original guide height band.

[0198] As an optional implementation, the original guide strip height prediction model determination module specifically includes:

[0199] The weight calculation unit is used to calculate the weight of each indicator based on the value of each indicator in the indicator system corresponding to the height of each original guide height band in the training sample using principal component analysis.

[0200] The weighting unit is used to assign weights to the values ​​of each indicator in the indicator system corresponding to the height of any original guide height in the training sample, according to the weights of each indicator, so as to obtain the weighted indicator.

[0201] The original guide zone height prediction model determination unit is used to obtain the original guide zone height prediction model based on the weighted index and the test sample using PSO-SVR.

[0202] As an optional implementation, the original guide strip height prediction model determination unit specifically includes:

[0203] An initialization subunit is used to initialize the particle swarm; the particle swarm includes multiple sets of parameters for the SVR model; each set of parameters includes a penalty factor coefficient and a kernel function.

[0204] The fitness calculation subunit is used to substitute the weighted index into the SVR model corresponding to each group of parameters to obtain the fitness of the SVR model corresponding to each group of parameters.

[0205] The judgment subunit is used to determine whether the target model is the optimal target model based on the test sample to obtain a first judgment result; the target model is the SVR model corresponding to the parameter with the highest fitness.

[0206] The first result determination subunit is used to determine the target model as the original guide zone height prediction model if the first judgment result is yes.

[0207] The second result determination subunit is used to update the particle swarm if the first judgment result is negative, and return to the step "substitute the weighted index into the SVR model corresponding to each group of parameters to obtain the fitness of the SVR model corresponding to each group of parameters".

[0208] The present invention has the following technical effects:

[0209] This invention overcomes the limitations of traditional empirical formulas and neural network predictions with small sample sizes, exhibiting strong generalization ability and high accuracy. It demonstrates good feasibility in predicting the height of the original confined aquifer elevation zone. While ensuring prediction accuracy, a prediction model for the original confined aquifer elevation zone is established: a Fuzzy (Fuzzy Comprehensive Evaluation Method) - GRA (Grey Relational Analysis Method) - EWM (Entropy Weight Method) - PCA (Principal Component Analysis Method) - PSO (Particle Swarm Optimization) - SVR (Support Vector Regression) prediction model. This model can deepen the understanding of the development mechanism of the original confined aquifer elevation zone, providing a theoretical basis for the prediction and prevention of water inrush in confined aquifers and for safe coal mine production, and can be applied to field prediction.

[0210] This invention does not rely on simple calculations using traditional formulas. Traditional formulas have a very limited number of coefficients and parameters, failing to comprehensively consider the factors influencing the development of the original conduction zone, leading to significant calculation errors. Furthermore, the correlation between the influencing factors and the numerical value of the conduction zone height in traditional formulas is overly simplistic, resulting in poor generalization. This invention, considering as many relevant factors as possible (14 influencing factors), first uses grey relational analysis to refine and reduce dimensionality, selecting influencing factors with a correlation coefficient greater than 0.8 as the basis for establishing the model's index system. Then, entropy weighting is used to assign weights to the refined influencing factors. Fuzzy analysis is then used to refine the influencing factors into three indices: water abundance index, tectonic index, and pressure index. Through multiple rounds of refinement, dimensionality reduction, and weighting, the degree of influence of each influencing factor and index is continuously differentiated through quantitative weighting, aiming to achieve a more realistic prediction process and improve the final prediction accuracy. In addition, PSO-SVR can seek the optimal parameters through continuous iteration to determine the final prediction model, which has great advantages in solving multi-objective optimization problems. It has good versatility, is suitable for handling various types of objective functions and constraints, and is easy to combine with traditional optimization methods to improve its own limitations. Therefore, it has higher accuracy than traditional formula methods.

[0211] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0212] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for predicting the height of a pressure-bearing original conductor belt, characterized in that, include: Acquire sample data, which includes the heights of multiple original guide zones and multiple sequences of related factor values; each sequence of related factor values ​​includes the values ​​of the same related factors for all original guide zones; the related factors include coal seam mining depth, unit water inflow from the pressure-bearing floor, aquifer thickness, floor permeability coefficient, working face slope length, advance speed, mining height, coal seam floor damage variables, fault strength index, fault fractal dimension, floor pressure, liquid surface tension coefficient, fracture coefficient, and density of water in the floor aquifer; The sample data is divided into training samples and test samples; For any original guide height band height and any related factor value sequence in the training sample, the grey relational analysis method is used to calculate the correlation degree between the original guide height band height and the related factor value sequence. The relevant factors are filtered based on the correlation between the height of each original guide band and the value sequence of each relevant factor to obtain the filtered relevant factors; The weights of each relevant factor after screening are calculated using the entropy weight method; The fuzzy comprehensive evaluation method was used to perform dimensionless processing on the relevant factor value sequences corresponding to each screened relevant factor to obtain the standardized screened relevant factor value sequences. The index system is determined based on the screened relevant factors, and the values ​​of each index in the index system corresponding to the height of each original guide band in the training sample are calculated based on the standardized value sequence of the screened relevant factors and the weight of each screened relevant factor. Based on PCA-PSO-SVR, the values ​​of each indicator in the indicator system corresponding to the height of each original guide strip in the training samples, and the test samples, an original guide strip height prediction model is obtained. The original guide strip height prediction model is used to predict the original guide strip height.

2. The method for predicting the height of the original pressure-bearing conductor strip according to claim 1, characterized in that, For any original guide height band height and any related factor value sequence in the training sample, the grey relational analysis method is used to calculate the correlation degree between the original guide height band height and the related factor value sequence, specifically including: For any original guide height band height and any related factor value sequence in the training sample, the fuzzy comprehensive evaluation method is used to perform dimensionless processing on the original guide height band height and the related factor value sequence to obtain the standardized related factor value sequence and the standardized original guide height band height. Calculate the absolute difference between the standardized sequence of relevant factor values ​​and the standardized original height of the guide band; The correlation coefficient between the height of the original guide zone and the sequence of related factor values ​​is obtained based on the absolute difference. The correlation coefficient is used to calculate the correlation degree between the height of the original guide zone and the sequence of related factor values.

3. The method for predicting the height of the original pressure-bearing conductor strip according to claim 2, characterized in that, For any original guide height band height and any related factor value sequence in the training samples, the fuzzy comprehensive evaluation method is used to perform dimensionless processing on the original guide height band height and the related factor value sequence to obtain standardized related factor value sequences and standardized original guide height band height, specifically including: The average value of the relevant factor value sequence in the training sample and the average value of the original guide height strip in the training sample are calculated based on the value sequence of each relevant factor in the training sample and the height of each original guide height strip in the training sample. Based on the average value of the relevant factor value sequence, the average height of the original guide band, the value sequence of each relevant factor in the training sample, and the height of each original guide band in the training sample, the standard deviation of the value sequence of the relevant factor in the training sample and the standard deviation of the height of the original guide band in the training sample are obtained. For any original guide height band height and any related factor value sequence in the training sample, the related factor value sequence and the original guide height band height are standardized according to the average value of the related factor value sequence in the training sample, the average value of the original guide height band height in the training sample, the standard deviation of the related factor value sequence in the training sample, and the standard deviation of the original guide height band height in the training sample, to obtain the standardized related factor value sequence and the standardized original guide height band height.

4. The method for predicting the height of the original pressure-bearing conductor strip according to claim 1, characterized in that, The original guideline height prediction model, based on PCA-PSO-SVR, the values ​​of each indicator in the indicator system corresponding to the height of each original guideline band in the training samples, and the test samples, specifically includes: Principal component analysis is used to calculate the weight of each index based on the value of each index in the index system corresponding to the height of each original guide height band in the training sample. For any original guide height in the training sample, the values ​​of each indicator in the indicator system corresponding to the height of the original guide height are weighted according to the weight of each indicator to obtain the weighted indicator. Using PSO-SVR, the original guide zone height prediction model is obtained based on the weighted indicators and the test samples.

5. The method for predicting the height of the original pressure-bearing conductor strip according to claim 4, characterized in that, The method employing PSO-SVR, based on the weighted indicators and the test samples, yields the original guide zone height prediction model, specifically including: Initialize the particle swarm; the particle swarm includes multiple sets of parameters for the SVR model; each set of parameters includes a penalty factor coefficient and a kernel function; Substitute the weighted index into the SVR model corresponding to each group of parameters to obtain the fitness of the SVR model corresponding to each group of parameters. Based on the test samples, a first judgment result is obtained by determining whether the target model is the optimal target model; the target model is the SVR model corresponding to the parameter with the highest fitness. If the first judgment result is yes, then the target model is determined to be the original guide zone height prediction model; If the first judgment result is negative, then update the particle swarm and return to the step "substitute the weighted index into the SVR model corresponding to each group of parameters to obtain the fitness of the SVR model corresponding to each group of parameters".

6. A system for predicting the height of a pressure-bearing original conductor belt, characterized in that, include: The acquisition module is used to acquire sample data, which includes the heights of multiple original guide height zones and multiple sequences of related factor values. Each sequence of related factor values ​​includes the values ​​of the same related factors for all original guide height zones. The related factors include coal seam mining depth, unit water inflow from the pressure floor, aquifer thickness, floor permeability coefficient, working face slope length, advance speed, mining height, coal seam floor damage variables, fault strength index, fault fractal dimension, floor pressure, liquid surface tension coefficient, fracture coefficient, and density of water in the floor aquifer. A training sample and test sample generation module is used to divide the sample data into training samples and test samples; The correlation calculation module is used to calculate the correlation between the height of any original guide band and the sequence of related factor values ​​in the training sample using grey relational analysis. The relevant factor filtering module is used to filter relevant factors based on the correlation between the height of each original guide band and the value sequence of each relevant factor to obtain the filtered relevant factors; The weight calculation module is used to calculate the weights of each relevant factor after screening using the entropy weight method. The standardization module is used to perform dimensionless processing on the sequence of relevant factor values ​​corresponding to each screened relevant factor using the fuzzy comprehensive evaluation method to obtain the standardized sequence of screened relevant factor values. The indicator calculation module is used to determine the indicator system based on the screened relevant factors, and to calculate the value of each indicator in the indicator system corresponding to the height of each original guide height band in the training sample based on the standardized value sequence of the screened relevant factors and the weight of each screened relevant factor. The original guide zone height prediction model determination module is used to obtain the original guide zone height prediction model based on PCA-PSO-SVR, the values ​​of each indicator in the indicator system corresponding to the height of each original guide zone in the training samples, and the test samples. The original guide zone height prediction model is used to predict the height of the original guide zone.

7. The system for predicting the height of the original pressure-bearing conductor belt according to claim 6, characterized in that, The correlation calculation module specifically includes: The standardization unit is used to perform dimensionless processing on the height of any original guide height band and the sequence of related factor values ​​in the training sample using the fuzzy comprehensive evaluation method, so as to obtain the standardized sequence of related factor values ​​and the standardized height of the original guide height band. An absolute difference calculation unit is used to calculate the absolute difference between the standardized sequence of relevant factor values ​​and the height of the standardized original guide strip. The correlation coefficient calculation unit is used to obtain the correlation coefficient between the height of the original guide band and the sequence of related factor values ​​based on the absolute difference. The correlation calculation unit is used to calculate the correlation between the height of the original guide band and the sequence of related factor values ​​based on the correlation coefficient.

8. The system for predicting the height of the original pressure-bearing conductor belt according to claim 7, characterized in that, The standardized unit specifically includes: The average value calculation subunit is used to calculate the average value of the relevant factor value sequence in the training sample and the average value of the original guide height strip in the training sample based on the value sequence of each relevant factor in the training sample and the height of each original guide height strip in the training sample; The standard deviation calculation subunit is used to obtain the standard deviation of the relevant factor value sequence in the training sample and the standard deviation of the original guide height band in the training sample based on the average value of the relevant factor value sequence, the average height of the original guide height band, the relevant factor value sequence in the training sample, and the height of the original guide height band in the training sample. The standardization subunit is used to standardize the height of any original guide height band and any related factor value sequence in the training sample based on the average value of the related factor value sequence in the training sample, the average height of the original guide height band in the training sample, the standard deviation of the related factor value sequence in the training sample, and the standard deviation of the height of the original guide height band in the training sample, to obtain the standardized related factor value sequence and the standardized height of the original guide height band.

9. The system for predicting the height of the original pressure-bearing conductor belt according to claim 6, characterized in that, The original guide strip height prediction model determination module specifically includes: The weight calculation unit is used to calculate the weight of each indicator based on the value of each indicator in the indicator system corresponding to the height of each original guide height band in the training sample using the principal component analysis method. The weighting unit is used to assign weights to the values ​​of each indicator in the indicator system corresponding to the height of any original guide height in the training sample according to the weights of each indicator, so as to obtain the weighted indicator. The original guide zone height prediction model determination unit is used to obtain the original guide zone height prediction model based on the weighted index and the test sample using PSO-SVR.

10. The system for predicting the height of the original pressure-bearing conductor belt according to claim 9, characterized in that, The original guide strip height prediction model determination unit specifically includes: An initialization subunit is used to initialize the particle swarm; the particle swarm includes multiple sets of parameters for the SVR model; each set of parameters includes a penalty factor coefficient and a kernel function; The fitness calculation subunit is used to substitute the weighted index into the SVR model corresponding to each group of parameters to obtain the fitness of the SVR model corresponding to each group of parameters. The judgment subunit is used to determine whether the target model is the optimal target model based on the test sample to obtain a first judgment result; the target model is the SVR model corresponding to the parameter with the highest fitness. The first result determination subunit is used to determine that the target model is the original guide zone height prediction model if the first judgment result is yes. The second result determination subunit is used to update the particle swarm if the first judgment result is negative, and return to the step "substitute the weighted index into the SVR model corresponding to each group of parameters to obtain the fitness of the SVR model corresponding to each group of parameters".

Images