Coal seam gas emission prediction method based on SSA-CIRCLE-ELM model
Through the SSA-CIRCLE-ELM model, the limit learning machine is optimized by using CIRCLE chaotic mapping and sparrow search algorithm, and the accuracy of coal mine gas influx is solved and more efficient prediction effect is achieved.
Patent Information
- Application Number
- CN202210281912.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-22
- Publication Date
- 2025-04-01
- Estimated Expiration
- 2042-03-22
AI Technical Summary
The prior art is difficult to accurately predict coal mine gas outflow, especially when multidimensional data matrix and single-layer feedforward neural network models are not fully utilized.
The SSA-CIRCLE-ELM model is used to initialize the sparrow search algorithm through CIRCLE chaotic mapping, optimize the SLFN node of the extreme learning machine, build a multi-dimensional state matrix, and enable the extreme learning machine to feature map and predict the gas data.
It improves the accuracy and efficiency of gas influx prediction, reduces prediction errors, and provides an effective model for coal mine gas monitoring systems.
Smart Images
Figure CN114611406B_ABST
Abstract
Description
Technical Field
[0001] The invention discloses a coal seam gas emission prediction method based on an SSA-CIRCLE-ELM model, belonging to the technical field of gas prediction. Background Art
[0002] In the coal industry, further research is needed on regional gas systems. In addition to rock deformation, stress concentration, stress concentration, and stress superposition, a multi-level and multi-directional integrated gas emission prediction model should be studied. Coal mine gas emission has the characteristics of fuzziness, unknowability, and randomness. Such characteristics make it impossible for traditional prediction methods to accurately predict gas emission. Based on the suddenness and complexity of gas accidents, how to accurately grasp the law of gas emission and carry out accurate and rapid prediction of gas is the key to controlling large-scale gas emission in advance. Based on the storage conditions of gas, gas emission has certain chaotic characteristics, and the complexity of information makes the geological system of coal and gas a nonlinear system where the superposition principle no longer holds. The gas system is analyzed from multiple factors such as original gas content, burial depth, coal thickness, mining height, recovery rate, gas content in adjacent layers, and mining intensity. On the basis of coal mine gas accidents, some domestic scholars have analyzed the essential causes of gas accidents and coal mine emergency models with other objective factors, and conducted detailed analysis according to each link of the system. The examples were analyzed and simulated, preventive measures were proposed, and a coal mine gas emission prediction system was established. A real-time gas monitoring system was constructed for underground coal mines. A coal mine gas emission prediction model was constructed. Many scholars have made certain contributions to the prediction of coal mine gas emission. There is little research on the multidimensional data matrix in the prediction of gas emission, and the gas single-layer feedforward neural network model has not been trained. Therefore, the present invention will analyze and establish a model from the perspective of the multidimensional state of gas emission and the single-layer feedforward neural network and make predictions. Summary of the invention
[0003] The technical problem to be solved by the present invention is: the present invention provides a coal seam gas emission prediction method of the SSA-CIRCLE-ELM model, constructs a gas emission prediction model under a multi-dimensional data matrix state and performs prediction.
[0004] The technical solution of the present invention is: a method for predicting coal seam gas emission based on the SSA-CIRCLE-ELM model, and the specific steps of the method are as follows:
[0005] Step 1: Use CIRCLE chaotic mapping to initialize the search space of the sparrow search algorithm, evenly distribute the search space, and obtain the optimized sparrow search algorithm;
[0006] Step 2: Use the sparrow search algorithm to optimize the SLFN nodes of the extreme learning machine so that its parameters generate a continuous probability distribution;
[0007] Step 3: construct a multidimensional state matrix, use an extreme learning machine to perform feature mapping on the multidimensional state matrix and minimize its error function;
[0008] Step 4: Based on the original gas data, use the optimized extreme learning machine to predict the gas data.
[0009] Furthermore, the specific steps of step 1 are:
[0010] Step 1.1: Construct a sparrow search algorithm, set the population size n, the number of iterations t, and the number of individuals z k , k = 1, 2, 3…l position information;
[0011] Step 1.2: Introduce CIRCLE chaos mapping:
[0012] Step 1.3, chimera CIRCLE chaotic mapping, strengthen the search space of sparrows, expand the foraging search range of sparrow populations for iteration;
[0013] At each iteration, the discoverer's position is updated as follows:
[0014]
[0015] Among them. max is a constant, indicating the maximum number of iterations. α∈(0,1] is a random number. R 2 (R 2 ∈[0,1]) and ST(ST∈[0.5,1]) represent the warning value and safety value respectively. Q is a random number that follows a normal distribution. L represents a 1×d matrix, in which all elements in the matrix are 1. 2 When <ST, the sparrow performs a search operation. 2 When >ST, the sparrow discovers the predator and performs evasive maneuvers. The position updates of other sparrows during evasive maneuvers are described as follows:
[0016]
[0017] Among them, Z worst It indicates the current global worst position, Z best It indicates the current global optimal position. A represents a 1×d matrix, in which each element is randomly assigned a value of 1 or -1. When i>n / 2, it means that the i-th participant with a lower fitness value has not obtained food and is in a very hungry state. At this time, it needs to fly to other places to forage for food.
[0018] Furthermore, the specific steps of step 2 are:
[0019] Step 2.1, use the sparrow search algorithm to optimize the SLNF nodes of the extreme learning machine and adjust the feature mapping from the input layer to the hidden layer in the extreme learning machine;
[0020] Set N arbitrary different samples (x i, t i ), where x i =(x i1 ,x i2 ,…x in ) T , t i =(t i1 ,t i2 ,…t in ) T , then the ELM output is:
[0021] Where: w i =(w i1 ,w i2 ,…,w in ) T is the input weight between the input layer neurons and the hidden layer neurons; β i =(β i1 ,β i2 ,…,β im ) T is the output weight between the hidden layer neurons and the output layer neurons; b i is the bias of the hidden layer neurons; g(.) is the activation function of the hidden layer neurons.
[0022] Step 2.2: Rely on radial basis functions to encapsulate calculation nodes and generate continuous probability distribution.
[0023] Furthermore, the specific steps of step 3 are:
[0024] Step 3.1, construct a multidimensional state matrix based on gas data;
[0025] Step 3.2: According to the multidimensional state matrix and node parameters, an error function is constructed for the SLFN node of the extreme learning machine, and the error value is minimized;
[0026] Step 4: Use the predicted data with higher overlap with the original data, and finally optimize the output limit to learn the best initial weights and thresholds, obtain the assigned SSA-CIRCLE-ELM prediction model, and make predictions.
[0027] The beneficial effects of the present invention are:
[0028] According to the prediction mode and conditions of coal mine gas emission, the present invention models the gas emission process, establishes the SSA-CIRCLE-ELM gas emission prediction model and method, provides an effective monitoring model for the gas monitoring system, and provides key technical support for building a digital mine gas monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 The SSA-CIRCLE-ELM prediction model algorithm flow chart of the embodiment of the present invention is shown by way of example;
[0030] Figure 2 The relative error comparison curves of different models in the embodiment of the present invention are shown as examples;
[0031] Figure 3 The comparison result between the actual value and the predicted value in the embodiment of the present invention is shown by way of example. DETAILED DESCRIPTION
[0032] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the technical scheme in the embodiments of the present invention will be clearly and completely described in combination with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. It should be noted that the embodiments in this application and the features in the embodiments can be combined with each other arbitrarily without conflict.
[0033] The coal seam gas emission prediction method based on the SSA-CIRCLE-ELM model proposed in the present invention is implemented according to the following steps.
[0034] A coal seam gas emission prediction method based on the SSA-CIRCLE-ELM model, the specific steps of the method are as follows:
[0035] Step 1: Use CIRCLE chaotic mapping to initialize the search space of the sparrow search algorithm, evenly distribute the search space, and obtain the optimized sparrow search algorithm;
[0036] Step 2: Use the sparrow search algorithm to optimize the SLFN nodes of the extreme learning machine so that its parameters generate a continuous probability distribution;
[0037] Step 3: construct a multidimensional state matrix, use an extreme learning machine to perform feature mapping on the multidimensional state matrix and minimize its error function;
[0038] Step 4: Based on the original gas data, use the optimized extreme learning machine to predict the gas data.
[0039] Furthermore, the specific steps of step 1 are:
[0040] Step 1.1: Construct a sparrow search algorithm, set the population size n, the number of iterations t, and the number of individuals z k , k = 1, 2, 3…l position information;
[0041] Step 1.2: Introduce CIRCLE chaos mapping:
[0042] Step 1.3, chimera CIRCLE chaotic mapping, strengthen the search space of sparrows, expand the foraging search range of sparrow populations for iteration;
[0043] At each iteration, the discoverer's position is updated as follows:
[0044]
[0045] Among them. max is a constant, indicating the maximum number of iterations. α∈(0,1] is a random number. R 2 (R 2 ∈[0,1]) and ST(ST∈[0.5,1]) represent the warning value and safety value respectively. Q is a random number that follows a normal distribution. L represents a 1×d matrix, in which all elements in the matrix are 1. 2 When <ST, the sparrow performs a search operation. 2 When >ST, the sparrow discovers the predator and performs evasive maneuvers. The position updates of other sparrows during evasive maneuvers are described as follows:
[0046]
[0047] Among them, Z worst It indicates the current global worst position, Z best It indicates the current global optimal position. A represents a 1×d matrix, in which each element is randomly assigned a value of 1 or -1. When i>n / 2, it means that the i-th participant with a lower fitness value has not obtained food and is in a very hungry state. At this time, it needs to fly to other places to forage for food.
[0048] Furthermore, the specific steps of step 2 are:
[0049] Step 2.1, use the sparrow search algorithm to optimize the SLNF nodes of the extreme learning machine and adjust the feature mapping from the input layer to the hidden layer in the extreme learning machine;
[0050] Set N arbitrary different samples (x i, t i), where x i =(x i1 ,x i2 ,…x in ) T , t i =(t i1 ,t i2 ,…t in ) T , then the ELM output is:
[0051] Where: w i =(w i1 ,w i2 ,…,w in ) T is the input weight between the input layer neurons and the hidden layer neurons; β i =(β i1 ,β i2 ,…,β im ) T is the output weight between the hidden layer neurons and the output layer neurons; b i is the bias of the hidden layer neurons; g(.) is the activation function of the hidden layer neurons.
[0052] Step 2.2: Rely on radial basis functions to encapsulate calculation nodes and generate continuous probability distribution.
[0053] Furthermore, the specific steps of step 3 are:
[0054] Step 3.1, construct a multidimensional state matrix based on gas data;
[0055] Step 3.2: According to the multidimensional state matrix and node parameters, an error function is constructed for the SLFN node of the extreme learning machine, and the error value is minimized;
[0056] Step 4: Use the predicted data with higher overlap with the original data, and finally optimize the output limit to learn the best initial weights and thresholds, obtain the assigned SSA-CIRCLE-ELM prediction model, and make predictions.
[0057] First, the first embodiment of the present invention is described in conjunction with the accompanying drawings.
[0058] Due to the geostress and rheology of coal seam gas, rock deformation, stress concentration, stress superposition, vibration waves, etc., the coal mine excavation process is close to high tectonic strain energy gas-containing rock and coal seams. The special effects cause gas overflow and continuous gas outflow; and the rheological behavior of gas-containing coal is spatially complex, and it often forms a complex nonlinear system in the three major areas of relaxation, stress concentration, and original stress.
[0059] The selected coal seam gas data are shown in Table 1, where the factors include original gas content, burial depth, coal thickness, mining height, recovery rate, adjacent layer gas content, mining intensity and gas emission.
[0060] Table 1 Coal mine gas data
[0061]
[0062] The actual measured sample data of gas is selected and imported into the extreme learning machine model for training. The selected models are ELM gas prediction model, PSO-ELM gas prediction model, SSA-ELM gas prediction model, and SSA-CIRCLE-ELM gas prediction model.
[0063] The data in Table 1 are selected as the training set and test set to construct four coal seam gas prediction models.
[0064] The test data of coal mine gas content and its related factors are imported into the ASS-CIRCLE-ELM prediction model. The prediction results after stable distribution are obtained according to the chaotic CIRCLE mapping as shown in Table 2.
[0065] Table 2 Comparison of prediction errors of four extreme learning machines
[0066]
[0067] As can be seen from Table 2, the prediction results of the extreme learning machine have large errors, and there are abnormal data due to the random generation of weights, with a maximum relative error of 8.8% and an average relative error of 3.5%. The maximum relative error rate of the gas prediction data obtained by the PSO-ELM prediction model is 9.5%, and the average relative error rate is 5%. The maximum relative error rate of the gas prediction data obtained by the SSA-ELM prediction model is 7.2%, and the average relative error rate is 5.3%. The maximum relative error rate of the gas prediction data obtained by the SSA-CIRCLE-ELM prediction model is 1.47%, and the average relative error rate is 0.859%.
[0068] Therefore, it can be proved that the accuracy of the extreme learning machine optimized by sparrow search and chaotic mapping coupling algorithm is greatly improved, because chaotic mapping makes the convergence speed faster, and the uniform distribution avoids falling into the local optimum. According to MATALB 2016A software, the prediction results of PSO-ELM, ELM, SSA-ELM, and SSA-CIRCLE-ELM are as follows: Figure 2 shown.
[0069] Plot the extreme learning machine model under particle swarm optimization and the extreme learning machine to compare the gas emission prediction model and the absolute error comparison between the predicted value and the actual value, as well as the extreme learning machine model optimized by the sparrow search algorithm and the extreme learning machine under the improved chaos mapping to compare the gas emission prediction model and the absolute error comparison between the predicted value and the actual value Figure 3 As can be seen from the figure, the prediction results of gas inrush by the SSA-CIRCLE-ELM prediction model are most coupled with the actual gas inrush. The PSO-ELM model is second, and the extreme learning machine model is the farthest, and there are data outliers, which means that the SSA-CIRCLE-ELM prediction model is closest to the actual value.
[0070] Comparison chart of actual coal mine gas emission values and predictions by various extreme learning machine algorithms, such as Figure 2 , 3 As shown in the figure. Under the same number of iterations and training parameters, the predicted value of the SSA-CIRCLE-ELM prediction model is most consistent with the actual gas emission value, and the prediction accuracy of SSA-CIRCLE-ELM is significantly improved compared with the prediction accuracy of the three extreme learning machines ELM-PSO, SSA-ELM, and ELM.
[0071] Among the above four gas prediction models, the absolute error curve changes as follows: the SSA-ELM gas prediction model has the largest change, and the PSO-ELM gas prediction model is second to SSA-ELM. The reason is that the local optimal defects in the above two extreme learning machine algorithms have not been solved. Because the local optimal value exists, there are large errors in the training process of the neural network; the ELM model is unstable, although the overall error is smaller than PSO-ELM and SSA-ELM, this is because the ordinary extreme learning machine cannot guarantee that the random weights are valid during the training process. Due to the characteristics of random assignment, it cannot effectively play its characteristics in some occasions. CIRCLE is used to search for individual uniform walks in a search space for SSA. The chaos algorithm not only improves the convergence speed and accuracy of the model, but also the absolute error line graph of the SSA-CIRCLE-ELM prediction model is more stable as a whole, and the absolute error rate is smaller.
[0072] The specific embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.
Claims
1. A coal seam gas emission prediction method based on the SSA-CIRCLE-ELM model is characterized by: The specific steps of the method are as follows: Step 1: Use CIRCLE chaotic mapping to initialize the search space of the sparrow search algorithm, evenly distribute the search space, and obtain the optimized sparrow search algorithm; Step 2: Use the sparrow search algorithm to optimize the SLFN nodes of the extreme learning machine so that its parameters generate a continuous probability distribution; Step 3: construct a multidimensional state matrix, use an extreme learning machine to perform feature mapping on the multidimensional state matrix and minimize its error function; Step 4: Based on the original gas data, the optimized extreme learning machine is used to predict the gas data; The specific steps of step 1 are: Step 1.1: Construct a sparrow search algorithm, set the population size n, the number of iterations t, and the number of individuals z k , k = 1, 2, 3…l position information; Step 1.2: Introduce CIRCLE chaos mapping: Step 1.3, chimera CIRCLE chaotic mapping, strengthen the search space of sparrows, expand the foraging search range of sparrow populations for iteration; At each iteration, the discoverer's position is updated as follows: Among them, item max is a constant, indicating the maximum number of iterations, α∈(0,1] is a random number, R2(R2∈[0,1]) and ST(ST∈[0.5,1]) represent the warning value and safety value respectively, Q is a random number that obeys the normal distribution, L represents a 1×d matrix, in which all elements in the matrix are 1, when R2<ST, the sparrow performs a search operation, when R2>ST, the sparrow finds the predator and performs an avoidance operation, and the position update of other sparrows during avoidance is described as follows: Among them, Z worst It indicates the current global worst position, Z best It represents the current global optimal position, A represents a 1×d matrix, in which each element is randomly assigned a value of 1 or -1. When i>n / 2, it indicates that the i-th participant with a low fitness value has not obtained food and is in a very hungry state. At this time, it needs to fly to other places to find food.
2. The method for predicting coal seam gas emission based on the SSA-CIRCLE-ELM model according to claim 1, characterized in that: The specific steps of step 2 are: Step 2.1, use the sparrow search algorithm to optimize the SLNF nodes of the extreme learning machine and adjust the feature mapping from the input layer to the hidden layer in the extreme learning machine; Step 2.2: Generate a continuous probability distribution based on the feature map.
3. The method for predicting coal seam gas emission based on the SSA-CIRCLE-ELM model according to claim 1, characterized in that: The specific steps of step 3 are: Step 3.1, construct a multidimensional state matrix based on gas data; Step 3.2: According to the multidimensional state matrix and node parameters, construct the error function for the SLFN node of the extreme learning machine and minimize the error value.
4. The method for predicting coal seam gas emission based on the SSA-CIRCLE-ELM model according to claim 1, characterized in that: The specific steps of step 4 are: using the predicted data with a higher degree of overlap with the original data, the final optimized output limit is used to learn the best initial weights and thresholds, obtain the assigned SSA-CIRCLE-ELM prediction model, and perform prediction.