Coal preparation plant energy consumption prediction method based on improved BP neural network

Through the improved BP neural network method, combined with hierarchical analysis method and genetic algorithm, the problem of insufficient accuracy of traditional energy consumption prediction methods in complex nonlinear performance consumption changes is solved, and high-precision prediction of energy consumption in coal preparation plants is achieved. Support the industry's energy conservation, emission reduction and clean production.

CN118917522AInactive Publication Date: 2024-11-08YHD

Patent Information

Application Number
CN202410928471.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional energy consumption prediction methods are insufficient in dealing with complex nonlinear performance consumption changes, and cannot build high-precision energy consumption models, and have high dependence on data quality and quantity, making it difficult to meet the accurate prediction needs of the coal coal preparation industry.

Method used

The improved BP neural network method is adopted to determine the combined weight of energy consumption factors through hierarchical analysis method, coefficient of variation method and genetic algorithm, establish and optimize the BP neural network model, and conduct comprehensive energy consumption prediction.

Benefits of technology

It improves the accuracy, stability and applicability of energy consumption prediction, can more accurately predict the main energy consumption factors of coal preparation plants, provide scientific management decision-making basis, and support energy conservation, emission reduction and clean production.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917522A_ABST
    Figure CN118917522A_ABST
Patent Text Reader

Abstract

The invention discloses a coal preparation plant energy consumption prediction method based on an improved BP neural network. The method comprises the following steps: determining coal preparation plant energy consumption factors; analyzing the energy consumption factors based on an analytic hierarchy process to obtain subjective weights; analyzing the energy consumption factors based on a variable coefficient method to obtain objective weights; calculating the combined weight of the energy consumption factors by using a weighted average method, and obtaining main energy consumption factor data; the method comprises the following steps: establishing a BP neural network comprehensive energy consumption basic model according to historical energy consumption factor data, training the model, optimizing the trained model by using a genetic algorithm, training to obtain an optimized BP neural network comprehensive energy consumption basic model, inputting main energy consumption factor data into the model, and predicting the comprehensive energy consumption. According to the method, the diversity of energy consumption of the coal preparation plant and the characteristic of difficulty in accurate measurement are comprehensively considered, and the prediction effectiveness and accuracy are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of coal mine energy consumption prediction, and in particular to an improved BP neural network method for predicting energy consumption in a coal preparation plant. Background Art

[0002] In coal mining and its downstream industries, coal preparation is an important component, which plays an important role in affecting the quality of coal and its further use. However, the energy consumption problem in the coal preparation process has always been one of the key factors affecting the development of the coal industry.

[0003] Traditional energy consumption prediction methods are mainly based on statistical learning and prediction based on historical data. However, the prediction accuracy of such methods is limited by the quality and quantity of data sets and cannot handle complex nonlinear problems. In addition, with the interaction of multiple factors in the process, energy consumption changes are increasingly showing complex, nonlinear and variable characteristics, which brings some difficulties to the prediction and construction of high-precision energy consumption models.

[0004] In this context, neural networks have been introduced into the field of energy consumption prediction. However, the prediction methods based on traditional neural networks often have a high dependence on the quality and quantity of data, and the performance of some learning tasks is not ideal. This requires a method that can more accurately predict the energy consumption of the coal preparation industry. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method for predicting the energy consumption of a coal preparation plant by improving the BP neural network, so as to more accurately and efficiently predict the energy consumption of the coal preparation plant, thereby providing strong support for the implementation of energy conservation, emission reduction and clean production, and providing a new solution for achieving accurate prediction. The purpose is to achieve the purpose of solving problems and improving the use value through the implementation of the present invention, provide a new, flexible and effective tool for researchers in this field, and promote scientific research and industrial development in related fields.

[0006] In order to solve the above technical problems, the present invention proposes a method for predicting energy consumption of a coal preparation plant using an improved BP neural network, comprising:

[0007] S1. Determine the energy consumption factors of the coal preparation plant.

[0008] S2. Analyze energy consumption factors based on the hierarchical analysis method and obtain subjective weights.

[0009] S3. Analyze energy consumption factors based on the coefficient of variation method and obtain objective weights.

[0010] S4. Use the weighted average method to calculate the combined weights of energy consumption factors and obtain the data of main energy consumption factors.

[0011] S5. According to the historical energy consumption factor data, a BP neural network comprehensive energy consumption basic model is established and the model is trained.

[0012] S6. Utilize a genetic algorithm to optimize the model trained in step S5 and perform training to obtain an optimized BP neural network comprehensive energy consumption basic model, input the main energy consumption factor data into the model, and predict the comprehensive energy consumption.

[0013] Furthermore, in step S1, the energy consumption factors include electricity consumption, dielectric consumption, water consumption and chemical consumption.

[0014] Furthermore, in step S2, the subjective weights obtained include the following:

[0015] S201. Establish a hierarchical analysis structure diagram of the comprehensive energy consumption of the coal preparation plant. The top part of the diagram is the target layer, i.e. the comprehensive energy consumption of the coal preparation plant; the middle part is the indicator layer, i.e. all energy consumption factors; the bottom part is divided into the object layer, i.e. the specific energy consumption data; the comprehensive energy consumption is the sum of all energy consumption factors.

[0016] S202, using the analytic hierarchy process to compare the energy consumption factors at the same level, and establish a judgment matrix. The expression of the judgment matrix is:

[0017]

[0018] Among them, A represents the judgment matrix, a ij It represents the comparison result between the ith energy consumption factor and the jth energy consumption factor, and n represents the total number of energy consumption factors.

[0019] The elements in the judgment matrix are normalized, and the eigenvalue method is used to obtain the eigenvector. The specific expression is:

[0020] W = {ω 1 ,ω 2 ,ω 3 ,…,ω i ,…,ω n}

[0021] Where W represents the eigenvector; ω i represents the weight of the i-th energy consumption factor, and

[0022] Normalize the eigenvector to obtain the subjective weight of the energy consumption factor.

[0023] S203, perform consistency check on the judgment matrix and calculate the maximum eigenvalue. The specific formula is:

[0024]

[0025] Among them, λ max represents the maximum eigenvalue, W i Represents the eigenvector of the i-th energy consumption factor.

[0026] The calculation formulas for consistency index and consistency ratio are:

[0027]

[0028]

[0029] Among them, CI represents the consistency index, l represents the order of the judgment matrix, CR represents the consistency ratio, and RI represents the random consistency index.

[0030] S204, when λ max =n and CI = 0, it indicates that the consistency test of the judgment matrix fully meets the standard; when CR < 0.1, it indicates that the consistency of the judgment matrix is ​​valid, otherwise the consistency of the judgment matrix is ​​invalid.

[0031] When the consistency check does not fully meet the standard or the consistency is invalid, steps S202-S203 are repeated until CR < 0.1, and the final subjective weight is obtained.

[0032] Furthermore, in step S3, obtaining the objective weight includes the following contents:

[0033] The data of energy consumption factors are standardized, and the specific formula is:

[0034] The formula for positive indicator processing is:

[0035]

[0036] The formula for negative indicator processing is:

[0037]

[0038] Among them, Y ik represents the standardized value of the kth data of the i-th energy consumption factor, x ik represents the original value of the kth data of the i-th energy consumption factor, represents the minimum original value of the kth data of the i-th energy consumption factor, Represents the maximum original value of the kth data of the ith energy consumption factor; the larger the positive index, the better, and the larger the negative index, the worse.

[0039] According to the standardized value, the average value, standard deviation and coefficient of variation of the index are calculated using the coefficient of variation method, and then the objective weight of the energy consumption factor is obtained. The specific formula is:

[0040]

[0041] Among them, Y i represents the average value of the i-th energy consumption factor, S i represents the standard deviation of the i-th energy consumption factor, P i represents the coefficient of variation of the i-th energy consumption factor, V i Represents the objective weight of the i-th energy consumption factor.

[0042] Furthermore, in step S4, the combined weights include the following:

[0043] In order to balance the subjective and objective weights, the weighted average method is used to combine these two weights to determine the combined weight of energy consumption factors. The specific calculation formula is:

[0044]

[0045] Among them, β i represents the combined weight of the i-th energy consumption factor, U i Represents the subjective weight of the i-th energy consumption factor.

[0046] The influence of each energy consumption factor on the comprehensive energy consumption is obtained by combining the weights and ranking them. The energy consumption factors with an influence of more than 90% are taken as the main energy consumption factor data. The calculation formula for the influence is:

[0047]

[0048] Furthermore, in step S5, establishing a BP neural network comprehensive energy consumption basic model includes the following contents:

[0049] The basic model of comprehensive energy consumption of BP neural network includes input layer, hidden layer and output layer.

[0050] By adopting reverse processing, negative indicators in historical energy consumption factor data are converted into positive indicators, that is, the smaller the better indicators are converted into the larger the better indicators; positive indicators are standardized to eliminate the dimensional differences between different indicators and make them comparable.

[0051] Determine the number of input layers, hidden layers, and output layers of the BP neural network comprehensive energy consumption basic model, and determine the number of input layer nodes, output layer nodes, hidden layer nodes, and sample quantity:

[0052] Number of input layer nodes: equal to the number of energy consumption factors;

[0053] The number of nodes in the output layer is determined by the number of output results;

[0054] The number of hidden layer nodes is determined empirically or through methods such as cross-validation to achieve the best model performance;

[0055] Sample size: determined based on the actual amount of data collected.

[0056] Model training: Randomly initialize the weights and biases of the BP neural network comprehensive energy consumption basic model, input the standardized positive indicators into the input layer of the BP neural network comprehensive energy consumption basic model, and obtain the predicted value of comprehensive energy consumption through the processing of the weights and activation functions in the hidden layer, and output it through the output layer; the specific formula is:

[0057]

[0058] in, Represents the predicted value, Q represents the weight matrix, X represents the input sample, b represents the bias, and f represents the activation function.

[0059] Calculate the error between the predicted value and the actual value. The specific formula is:

[0060]

[0061] Among them, E represents the error, M represents the total number of samples, Represents the predicted value of the mth sample, y m Represents the actual value of the mth sample.

[0062] The weights are adjusted inversely according to the error between the predicted value and the actual value.

[0063] The model training process is repeated until the error converges or the preset number of iterations is reached, and the trained BP neural network comprehensive energy consumption basic model is obtained.

[0064] Furthermore, in step S6, the optimized BP neural network comprehensive energy consumption basic model includes the following contents:

[0065] Gene encoding: Encode the weights and bias parameters in the trained BP neural network comprehensive energy consumption basic model to form chromosomes.

[0066] Population initialization: Randomly generate a certain number of chromosomes to form an initial population.

[0067] Calculate the fitness value of the chromosome to evaluate the quality of each chromosome. The specific calculation formula is:

[0068]

[0069] Among them, F represents the fitness value; E represents the error; ∈ is a small positive number used to prevent the denominator from being zero.

[0070] The fitness value calculates the performance of the neural network model corresponding to each chromosome.

[0071] Through the iterative operation of selection, crossover and mutation operators in the genetic algorithm, the chromosome with the highest fitness value is obtained as the optimal individual.

[0072] Model training: Input the optimal individual into the trained BP neural network comprehensive energy consumption basic model, decode the weight and bias of the optimal individual, and obtain the predicted value of comprehensive energy consumption through the processing of the weight and activation function in the hidden layer, and output it through the output layer; the specific formula is:

[0073]

[0074] in, Represents the predicted value, Q represents the weight matrix, X represents the input sample, b represents the bias, and f represents the activation function.

[0075] Calculate the error between the predicted value and the actual value. The specific formula is:

[0076]

[0077] Among them, E represents the error, M represents the total number of samples, Represents the predicted value of the mth sample, y m Represents the actual value of the mth sample.

[0078] The weights are adjusted inversely according to the error between the predicted value and the actual value.

[0079] The model training process is repeated until the error converges or the preset number of iterations is reached to obtain the optimized BP neural network comprehensive energy consumption basic model.

[0080] Furthermore, in step S6, the comprehensive energy consumption prediction includes the following contents:

[0081] The data of main energy consumption factors are preprocessed, including missing value filling, outlier processing and data standardization.

[0082] The preprocessed data is input into the optimized BP neural network comprehensive energy consumption basic model, and the data is passed to the hidden layer through the input layer. After being processed by the weights and activation functions in the hidden layer, the predicted value of the comprehensive energy consumption is obtained and output through the output layer as the basis for energy consumption management and optimization of the coal preparation plant.

[0083] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0084] The present invention combines genetic algorithm with BP neural network to make more accurate predictions of the main energy consumption factors of coal preparation plants, improving the accuracy, stability and wide applicability of the prediction results. In addition, for management, only the corresponding parameters and data need to be input to understand the overall energy consumption of the plant under the current environment and its possible development trend. Combined with this information, management can take appropriate strategies and actions to effectively control the comprehensive energy consumption of coal preparation plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 It is an overall implementation flow chart of the present invention.

[0086] Figure 2 It is a structural diagram of the hierarchical analysis method of the present invention. DETAILED DESCRIPTION

[0087] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0088] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0089] It should be noted that, in the description of the present invention, unless otherwise specified, "plurality" means two or more than two; the terms "upper", "lower", "left", "right", "inner", "outer", "front end", "rear end", "head", "tail" and the like indicate the orientation or position relationship based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0090] Furthermore, the terms “first,” “second,” “third,” etc. are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.

[0091] At the same time, in the description of the present invention, unless otherwise clearly specified and limited, the terms "connected" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0092] To achieve the above object, the present invention proposes a method for predicting energy consumption of a coal preparation plant using an improved BP neural network, such as Figure 1 As shown, the specific steps are as follows:

[0093] S1. Determine the energy consumption factors of the coal preparation plant, including electricity consumption, medium consumption, water consumption and chemical consumption.

[0094] Electricity consumption: mainly consumed by the mechanical operation of coal preparation equipment, including factors such as equipment idling time and equipment arrangement.

[0095] Medium loss: mainly occurs in the process of coal preparation using heavy media. Process design and operation control are the main factors affecting medium loss.

[0096] Water consumption: The dehydration process in the coal preparation process and the control of product moisture are the key factors in determining the water consumption.

[0097] Chemical consumption: includes the use of flotation reagents and flocculants. The type of reagents, coal slime properties, pretreatment equipment, etc. affect the flotation chemical consumption, while the selection, preparation and dosing method of flocculants affect the flocculant consumption.

[0098] S2. Analyze energy consumption factors based on the AHP method to obtain subjective weights. The AHP method solves the challenges of evaluating multi-objective and multi-criteria problems with its unique simplicity, hierarchical structure, system perspective and mathematical modeling. This method is particularly applicable when it is difficult to make accurate calculations using data. Figure 2 The specific contents are as follows:

[0099] S201. Establish a hierarchical analysis structure diagram of the comprehensive energy consumption of the coal preparation plant. The top part of the diagram is the target layer, i.e. the comprehensive energy consumption of the coal preparation plant, which refers to the core decision-making issues that need to be discussed and analyzed; the middle part is the indicator layer, i.e. all energy consumption factors, which are various key influencing elements to help achieve the decision-making goals; the bottom part is divided into the object layer, i.e. specific energy consumption data, to complete the implementation of the decision-making path; the comprehensive energy consumption is the sum of all energy consumption factors.

[0100] S202, using the analytic hierarchy process to compare the energy consumption factors at the same level, and establish a judgment matrix. The expression of the judgment matrix is:

[0101]

[0102] Among them, A represents the judgment matrix; a ij It represents the comparison result between the ith energy consumption factor and the jth energy consumption factor. The comparison result is a numerical value given by experts to the energy consumption factors at the same level based on their experience and professional knowledge. The numerical value reflects the relative importance of the two. n represents the total number of energy consumption factors.

[0103] The elements in the judgment matrix are normalized, and the eigenvalue method is used to obtain the eigenvector. The specific expression is:

[0104] W = {ω 1 ,ω 2 ,ω 3 ,…,ω i ,…,ω n}

[0105] Where W represents the eigenvector; ω i represents the weight of the i-th energy consumption factor, and

[0106] Normalize the eigenvector to obtain the subjective weight of the energy consumption factor.

[0107] S203, perform consistency check on the judgment matrix and calculate the maximum eigenvalue. The specific formula is:

[0108]

[0109] Among them, λ max represents the maximum eigenvalue, W i Represents the eigenvector of the i-th energy consumption factor.

[0110] The calculation formulas for consistency index and consistency ratio are:

[0111]

[0112] Among them, CI represents the consistency index, l represents the order of the judgment matrix, CR represents the consistency ratio, and RI represents the random consistency index.

[0113] S204, when λ max =n and CI = 0, it indicates that the consistency test of the judgment matrix fully meets the standard; when CR < 0.1, it indicates that the consistency of the judgment matrix is ​​valid, otherwise the consistency of the judgment matrix is ​​invalid.

[0114] When the consistency check is not fully up to standard or the consistency is invalid, experts are invited to re-evaluate the energy consumption factors at the same level to obtain a comparison result, or the number of experts participating in the evaluation is increased, and the comparison result is obtained by summarizing the opinions of multiple experts, and steps S202-S203 are repeated until CR < 0.1 to obtain the final subjective weight.

[0115] S3. Analyze energy consumption factors based on the coefficient of variation method and obtain objective weights; the specific contents are as follows:

[0116] The data of energy consumption factors are standardized, and the specific formula is:

[0117] The formula for positive indicator processing is:

[0118]

[0119] The formula for negative indicator processing is:

[0120]

[0121] Among them, Y ik represents the standardized value of the kth data of the i-th energy consumption factor, x ik represents the original value of the kth data of the i-th energy consumption factor, represents the minimum original value of the kth data of the i-th energy consumption factor, Represents the maximum original value of the kth data of the ith energy consumption factor; the larger the positive index, the better, and the larger the negative index, the worse.

[0122] According to the standardized value, the average value, standard deviation and coefficient of variation of the index are calculated using the coefficient of variation method, and then the objective weight of the energy consumption factor is obtained. The specific formula is:

[0123]

[0124] Among them, Y i represents the average value of the i-th energy consumption factor, S i represents the standard deviation of the i-th energy consumption factor, P i Represents the coefficient of variation of the i-th energy consumption factor.

[0125] S4. Use the weighted average method to calculate the combined weights of energy consumption factors and obtain the data of main energy consumption factors; the specific contents are:

[0126] In order to balance the subjective and objective weights, the weighted average method is used to combine these two weights to determine the combined weight of energy consumption factors. The specific calculation formula is:

[0127]

[0128] Among them, β i represents the combined weight of the i-th energy consumption factor, U i Represents the subjective weight of the i-th energy consumption factor.

[0129] By combining the weights, we can get the influence of each energy consumption factor on the comprehensive energy consumption and sort them. The energy consumption factors with an influence of more than 90% are taken as the main energy consumption factor data to explain the changes in the comprehensive energy consumption, which is also conducive to simplifying the complexity of the subsequent neural network prediction model. The calculation formula for the influence degree is:

[0130]

[0131] S5. Based on the historical energy consumption factor data, establish a BP neural network comprehensive energy consumption basic model and train the model; the specific contents are as follows:

[0132] The basic model of comprehensive energy consumption of BP neural network includes input layer, hidden layer and output layer.

[0133] By adopting reverse processing, negative indicators in historical energy consumption factor data are converted into positive indicators, that is, the smaller the better indicators are converted into the larger the better indicators; positive indicators are standardized to eliminate the dimensional differences between different indicators and make them comparable.

[0134] Determine the number of input layers, hidden layers, and output layers of the BP neural network comprehensive energy consumption basic model, and determine the number of input layer nodes, output layer nodes, hidden layer layers, and sample quantity:

[0135] Number of input layer nodes: equal to the number of energy consumption factors, i.e. set to 4;

[0136] Number of nodes in the output layer: usually 1, indicating the predicted value of comprehensive energy consumption;

[0137] The number of hidden layer nodes can be determined based on experience or through methods such as cross-validation to achieve the best model performance; H represents the number of nodes in the hidden layer, e represents the number of nodes in the input layer, c represents the number of nodes in the output layer, and d represents the empirical constant. If d = 3, then

[0138] Sample size: A representative sample size of 5 times the total weight of network connections is selected to ensure the accuracy of the network. Assuming the total weight of network connections is 50, the sample size is selected as 250.

[0139] Model training: Randomly initialize the weights and biases of the BP neural network comprehensive energy consumption basic model, input the standardized positive indicators into the input layer of the BP neural network comprehensive energy consumption basic model, and obtain the predicted value of comprehensive energy consumption through the processing of the weights and activation functions in the hidden layer, and output it through the output layer; the specific formula is:

[0140]

[0141] in, Represents the predicted value, Q represents the weight matrix, X represents the input sample, b represents the bias, and f represents the activation function.

[0142] Calculate the error between the predicted value and the actual value. The specific formula is:

[0143]

[0144] Among them, E represents the error, M represents the total number of samples, Represents the predicted value of the mth sample, y m Represents the actual value of the mth sample.

[0145] The weights are adjusted inversely according to the mean square error (MSE) between the predicted value and the actual value.

[0146] The model training process is repeated until the error converges or the preset number of iterations is reached, and the trained BP neural network comprehensive energy consumption basic model is obtained.

[0147] S6, using genetic algorithm to optimize the model trained in step S5 and train it to obtain the optimized BP neural network comprehensive energy consumption basic model, input the main energy consumption factor data into the model, and predict the comprehensive energy consumption; the specific content is:

[0148] Gene coding: The weights and bias parameters in the trained BP neural network comprehensive energy consumption basic model are encoded to form chromosomes; each chromosome represents a possible solution, and these gene codes are derived from the parameter representation of the BP neural network.

[0149] Population initialization: A certain number of chromosomes are randomly generated to form an initial population; the size of the population can be determined based on the scale of the actual problem and computing resources, and the source of the population is the randomly generated initial parameter set.

[0150] Calculate the fitness value of the initial population to evaluate the quality of each chromosome. The specific calculation formula is:

[0151]

[0152] Among them, F represents the fitness value; E represents the error of the BP neural network; ∈ is a small positive number used to prevent the denominator from being zero. The fitness value calculates the performance of the neural network model corresponding to each chromosome.

[0153] Through the iterative operation of selection, crossover and mutation operators in the genetic algorithm, the chromosome with the highest fitness value is obtained as the optimal individual:

[0154] Selection: Chromosomes are selected based on their fitness values. Chromosomes with higher fitness values ​​are more likely to be selected. Commonly used methods include roulette wheel selection and tournament selection.

[0155] Crossover: Perform a crossover operation on the selected chromosome to generate a new chromosome. The crossover operation can be performed using single-point crossover or multi-point crossover methods.

[0156] Mutation: Perform mutation operations on the chromosomes after crossover to increase the diversity of the population. The mutation operation can randomly change certain genes in the chromosomes.

[0157] Iterative operation: Repeat the selection, crossover and mutation operations, and after multiple generations of iterations, gradually optimize the chromosomes in the population until the chromosome with the highest fitness value is found as the optimal individual.

[0158] Model training: Input the optimal individual into the trained BP neural network comprehensive energy consumption basic model, decode the weight and bias of the optimal individual, and obtain the predicted value of comprehensive energy consumption through the processing of the weight and activation function in the hidden layer, and output it through the output layer; the specific formula is:

[0159]

[0160] in, Represents the predicted value, Q represents the weight matrix, X represents the input sample, b represents the bias, and f represents the activation function.

[0161] Calculate the error between the predicted value and the actual value. The specific formula is:

[0162]

[0163] Among them, E represents the error, M represents the total number of samples, Represents the predicted value of the mth sample, y m Represents the actual value of the mth sample.

[0164] The weights are adjusted inversely according to the error between the predicted value and the actual value.

[0165] The model training process is repeated until the error converges or the preset number of iterations is reached to obtain the optimized BP neural network comprehensive energy consumption basic model.

[0166] The data of main energy consumption factors are preprocessed, including missing value filling, outlier processing and data standardization.

[0167] The preprocessed data is input into the optimized BP neural network comprehensive energy consumption basic model, and the data is passed to the hidden layer through the input layer. After being processed by the weights and activation functions in the hidden layer, the predicted value of the comprehensive energy consumption is obtained and output through the output layer as the basis for energy consumption management and optimization of the coal preparation plant.

[0168] The main energy consumption factors of the coal preparation plant are predicted by the improved BP neural network of genetic algorithm, and the prediction accuracy, stability and generalization are improved. The managers only need to input the corresponding parameters and data to obtain the comprehensive energy consumption and its changing trend in the plant under the current situation. According to the comprehensive energy consumption status and changing trend obtained, the managers can take corresponding countermeasures and measures to control the comprehensive energy consumption of the coal preparation plant.

[0169] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting energy consumption in a coal preparation plant based on an improved BP neural network, characterized in that: include: S1. Determine the energy consumption factors of the coal preparation plant; S2, analyze the energy consumption factors based on the hierarchical analysis method and obtain the subjective weights; S3, analyze the energy consumption factors based on the coefficient of variation method and obtain objective weights; S4. Calculate the combined weights of energy consumption factors using the weighted average method and obtain the data of main energy consumption factors; S5. According to the historical energy consumption factor data, a BP neural network comprehensive energy consumption basic model is established and the model is trained; S6. Utilize the genetic algorithm to optimize the model trained in step S5 to obtain an optimized BP neural network comprehensive energy consumption basic model, input the main energy consumption factor data into the model, and predict the comprehensive energy consumption.

2. The method for predicting energy consumption of a coal preparation plant using an improved BP neural network according to claim 1 is characterized in that: In step S1, the energy consumption factors include electricity consumption, dielectric consumption, water consumption and chemical consumption.

3. The method for predicting energy consumption of a coal preparation plant using an improved BP neural network according to claim 1, characterized in that: In step S2, the subjective weight is obtained including the following contents: S201. Establish a hierarchical analysis structure diagram of comprehensive energy consumption of coal preparation plant, the top part of which is the target layer, the middle part is the indicator layer, and the bottom part is the object layer; the comprehensive energy consumption is the sum of all energy consumption factors; S202, using the analytic hierarchy process to compare the energy consumption factors at the same level, and establish a judgment matrix. The expression of the judgment matrix is: Among them, A represents the judgment matrix, a ij represents the comparison result between the ith energy consumption factor and the jth energy consumption factor, and n represents the total number of energy consumption factors; The elements in the judgment matrix are normalized, and the eigenvalue method is used to obtain the eigenvector. The specific expression is: W={ω1,ω2,ω3,…,ω i ,…,oh n } Where W represents the eigenvector; ω i represents the weight of the i-th energy consumption factor, and Normalize the eigenvector to obtain the subjective weight of the energy consumption factor; S203, calculate the maximum eigenvalue of the judgment matrix, the specific formula is: Among them, λ max represents the maximum eigenvalue, W i Represents the eigenvector of the i-th energy consumption factor; The judgment matrix is ​​tested for consistency, and the calculation formulas for consistency index and consistency ratio are: Among them, CI represents the consistency index, l represents the order of the judgment matrix, CR represents the consistency ratio, and RI represents the random consistency index; S204, when λ max =n and CI = 0, it indicates that the consistency test of the judgment matrix fully meets the standard; when CR < 0.1, it indicates that the consistency of the judgment matrix is ​​valid, otherwise the consistency of the judgment matrix is ​​invalid; When the consistency check does not fully meet the standard or the consistency is invalid, steps S202-S203 are repeated until CR < 0.1, and the final subjective weight is obtained.

4. The method for predicting energy consumption of a coal preparation plant using an improved BP neural network according to claim 1, characterized in that: In step S3, obtaining the objective weight includes the following contents: The data of energy consumption factors are standardized, and the specific formula is: The formula for positive indicator processing is: The formula for negative indicator processing is: Among them, Y ik represents the standardized value of the kth data of the i-th energy consumption factor, x ik represents the original value of the kth data of the i-th energy consumption factor, represents the minimum original value of the kth data of the i-th energy consumption factor, Indicates the maximum original value of the kth data of the i-th energy consumption factor; According to the standardized value, the average value, standard deviation and coefficient of variation of the index are calculated using the coefficient of variation method, and then the objective weight of the energy consumption factor is obtained. The specific formula is: Among them, Y i represents the average value of the i-th energy consumption factor, S i represents the standard deviation of the i-th energy consumption factor, P i represents the coefficient of variation of the i-th energy consumption factor, V i represents the objective weight of the i-th energy consumption factor, and n represents the total number of energy consumption factors.

5. The method for predicting energy consumption of a coal preparation plant using an improved BP neural network according to claim 1, characterized in that: In step S4, the combined weights are obtained including the following: The weighted average method is used to combine subjective weights and objective weights to obtain the combined weights of energy consumption factors. The specific calculation formula is: Among them, β i represents the combined weight of the i-th energy consumption factor, U i represents the subjective weight of the i-th energy consumption factor, V i represents the objective weight of the i-th energy consumption factor, and n represents the total number of energy consumption factors; The influence of each energy consumption factor on the comprehensive energy consumption is obtained by combining the weights and ranking them. The energy consumption factors with an influence of more than 90% are taken as the main energy consumption factor data. The calculation formula for the influence is:

6. The method for predicting energy consumption of a coal preparation plant using an improved BP neural network according to claim 4 is characterized in that: In step S5, establishing a BP neural network comprehensive energy consumption basic model includes the following contents: The basic model of comprehensive energy consumption of BP neural network includes input layer, hidden layer and output layer; Determine the number of input layers, hidden layers and output layers of the BP neural network comprehensive energy consumption basic model, and determine the number of input layer nodes, output layer nodes, hidden layer nodes and sample quantity; Reverse processing is used to convert negative indicators in historical energy consumption factor data into positive indicators, and the positive indicators are standardized; Model training: Randomly initialize the weights and biases of the BP neural network comprehensive energy consumption basic model, input the standardized positive indicators into the input layer of the BP neural network comprehensive energy consumption basic model, and obtain the predicted value of comprehensive energy consumption through the processing of the weights and activation functions in the hidden layer, and output it through the output layer; the specific formula is: in, represents the predicted value, Q represents the weight matrix, X represents the input sample, b represents the bias, and f represents the activation function; Calculate the error between the predicted value and the actual value. The specific formula is: Among them, E represents the error, M represents the total number of samples, Represents the predicted value of the mth sample, y m Represents the actual value of the mth sample; According to the error between the predicted value and the actual value, the weight is adjusted inversely; The model training process is repeated until the preset number of iterations is reached, and the trained BP neural network comprehensive energy consumption basic model is obtained.

7. The method for predicting energy consumption of a coal preparation plant using an improved BP neural network according to claim 1, characterized in that: In step S6, the optimized BP neural network comprehensive energy consumption basic model includes the following contents: Gene encoding: Encode the weights and bias parameters in the trained BP neural network comprehensive energy consumption basic model to form chromosomes; Population initialization: randomly generate a certain number of chromosomes to form an initial population; Calculate the fitness value of the initial population. The specific calculation formula is: Among them, F represents the fitness value, E represents the error, and ∈ represents a positive number; Through the iterative operation of selection, crossover and mutation operators in the genetic algorithm, the chromosome with the highest fitness value is obtained as the optimal individual; Model training: Input the optimal individual into the trained BP neural network comprehensive energy consumption basic model, decode the weight and bias of the optimal individual, and obtain the predicted value of comprehensive energy consumption through the processing of the weight and activation function in the hidden layer, and output it through the output layer; the specific formula is: in, represents the predicted value, Q represents the weight matrix, X represents the input sample, b represents the bias, and f represents the activation function; Calculate the error between the predicted value and the actual value. The specific formula is: Among them, E represents the error, M represents the total number of samples, Represents the predicted value of the mth sample, y m Represents the actual value of the mth sample; According to the error between the predicted value and the actual value, the weight is adjusted inversely; The model training process is repeated until the preset number of iterations is reached, and the optimized BP neural network comprehensive energy consumption basic model is obtained.

8. The method for predicting energy consumption of a coal preparation plant using an improved BP neural network according to claim 1, characterized in that: In step S6, the comprehensive energy consumption prediction includes the following contents: Preprocess the data of main energy consumption factors, including missing value filling, outlier processing and data standardization; The preprocessed data is input into the optimized BP neural network comprehensive energy consumption basic model, and the data is passed to the hidden layer through the input layer. After being processed by the weights and activation functions in the hidden layer, the predicted value of the comprehensive energy consumption is obtained and output through the output layer as the basis for energy consumption management and optimization of the coal preparation plant.

Citation Information

Patent Citations

  • Electricity consumption prediction method

    CN108805331A

  • Full-deformation prediction method and system for shield construction tunnel based on BP neural network

    CN110442979A

  • Electricity consumption prediction method and electricity consumption adjustment method

    CN115796386A

  • Machine-learning techniques for factor-level monotonic neural networks

    US20230297847A1

Cited By

  • Management method and system for coal dressing quality intelligent control

    CN120143774A

  • Coal mine production energy prediction method based on physical-information fusion neural network

    CN122334603A