CN113139692A - Waste slag dam deformation prediction method based on DPSO-ANFIS - Google Patents
Waste slag dam deformation prediction method based on DPSO-ANFIS Download PDFInfo
- Publication number
- CN113139692A CN113139692A CN202110485694.2A CN202110485694A CN113139692A CN 113139692 A CN113139692 A CN 113139692A CN 202110485694 A CN202110485694 A CN 202110485694A CN 113139692 A CN113139692 A CN 113139692A
- Authority
- CN
- China
- Prior art keywords
- neural network
- particle
- fuzzy neural
- self
- adaptive fuzzy
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 239000002893 slag Substances 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 title claims abstract description 38
- 239000002699 waste material Substances 0.000 title claims abstract description 25
- 238000013528 artificial neural network Methods 0.000 claims abstract description 70
- 230000003044 adaptive effect Effects 0.000 claims abstract description 33
- 238000012544 monitoring process Methods 0.000 claims abstract description 20
- 230000000694 effects Effects 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 239000002245 particle Substances 0.000 claims description 95
- 230000006870 function Effects 0.000 claims description 19
- 150000001875 compounds Chemical class 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 230000003203 everyday effect Effects 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims 2
- 238000005457 optimization Methods 0.000 description 10
- 238000010276 construction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000009412 basement excavation Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/043—Architecture, e.g. interconnection topology based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Tourism & Hospitality (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Bioinformatics & Computational Biology (AREA)
- Entrepreneurship & Innovation (AREA)
- Evolutionary Biology (AREA)
- Fuzzy Systems (AREA)
- Development Economics (AREA)
- Computational Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Automation & Control Theory (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Analysis (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
Abstract
The invention discloses a DPSO-ANFIS-based deformation prediction method for a waste slag dam, which comprises the following steps of: step 1, monitoring the original data of the abandoned slag dam, and preprocessing the original data; step 2, building a self-adaptive fuzzy neural network system, and assigning network parameters in the system; step 3, optimizing the parameters of the adaptive fuzzy neural network in the step 2; and 4, substituting the data of the abandoned slag dam preprocessed in the step 1 into the self-adaptive fuzzy neural network for training and prediction, wherein parameters in the self-adaptive fuzzy neural network are optimized through the step 3, calculating prediction precision values, and comparing prediction effects of measuring points at different positions. The prediction method has the advantages of high accuracy, good generalization and reliable stability, and excellent practical and comprehensive engineering performance.
Description
Technical Field
The invention belongs to the technical field of deformation prediction of waste slag dams, and particularly relates to a DPSO-ANFIS-based deformation prediction method of a waste slag dam.
Background
With the development of social economy, modern hydraulic engineering technology is continuously improved, and reservoir dam construction becomes the main body of the current hydraulic engineering content. The method for establishing the deformation prediction model by using the actual monitoring data of the dam is a means for effectively and safely monitoring the dam, and has important significance for monitoring the safe operation of the dam. Aiming at dam deformation monitoring data with complex nonlinear characteristics, an Adaptive fuzzy neural Network System (ANFIS) has the characteristics of structure and parameter identification, adaptively generates fuzzy rules and optimizes membership and output functions. According to input influence factors, system parameters can be automatically designed, self-learning of a fuzzy system is achieved, and the method has good applicability to a nonlinear system and good application prospect in dam safety monitoring. At present, the adaptive fuzzy neural network is applied more in the field of dam deformation prediction, but the problems of difficulty in optimizing adaptive fuzzy neural network fitness parameters, low prediction precision of a dam deformation prediction method and poor generalization performance still exist.
Disclosure of Invention
The invention aims to provide a DPSO-ANFIS-based deformation prediction method for a abandoned slag dam, and solves the problems of low accuracy and poor generalization performance of the conventional deformation prediction method for the abandoned slag dam.
The technical scheme adopted by the invention is that the waste slag dam deformation prediction method based on DPSO-ANFIS specifically comprises the following steps:
step 1, monitoring the original data of the abandoned slag dam, and preprocessing the original data;
step 2, building a self-adaptive fuzzy neural network system, and assigning network parameters in the system;
step 3, optimizing the parameters of the adaptive fuzzy neural network in the step 2;
and 4, substituting the data of the abandoned slag dam preprocessed in the step 1 into the self-adaptive fuzzy neural network for training and prediction, wherein parameters in the self-adaptive fuzzy neural network are optimized through the step 3, calculating prediction precision values, and comparing prediction effects of measuring points at different positions.
The present invention is also characterized in that,
the step 2 specifically comprises the following steps:
step 2.1, firstly, building a self-adaptive fuzzy neural network, inputting the waste slag dam data preprocessed in the step 1 into a first-layer input layer of the self-adaptive fuzzy neural network as a data sample, performing fuzzification processing, and outputting corresponding membership degrees;
in the formula (I), the compound is shown in the specification,the membership degree of the corresponding node;
wherein, muAi(x),μBj(y) are membership functions corresponding to the input variables respectively; i. j is respectively an adaptive node with a node function in the adaptive fuzzy neural network, a bell-shaped membership function is selected, and the formulas (3) and (4) are substituted into the formulas (1) and (2) respectively to obtain membership
Step 2.2, calculating the fitness of each rule according to the membership degrees obtained in the step 2.1, taking the membership degree obtained in the first layer as an input signal of a fuzzy rule layer of the second layer, and then accumulating and multiplying the input signal to obtain an output result of the second layer, namely the fitness value;
in the formula, Oi 2To correspond to the fitness value of the rule, ωiThe excitation strength of the ith rule;
and 2.3, selecting an ANFIS initial parameter, and assigning the parameter of the self-adaptive fuzzy neural network through a clustering method.
And 3, specifically, extracting the fitness value of a second fuzzy rule layer in the self-adaptive fuzzy neural network in the step 2 to serve as a particle swarm initialization parameter to be optimized, solving a space initialization standard particle swarm, and iteratively optimizing dynamically updated inertia weight and algorithm by adopting a dynamic weight particle swarm algorithm to generate an optimal fitness value of the fuzzy rule layer.
The particle group algorithm in the step 3 specifically comprises the following steps:
step 3.1, initializing standard particle swarm algorithm parameters, and assigning values to the size of a population, the initial position and speed of particles, the number of iterations, the error precision and the inertia weight;
step 3.2, calculating particle fitness values, calculating the fitness value corresponding to each particle at different positions according to a set objective function, wherein the set objective function takes the Mean Square Error (MSE) of the model prediction output value and the measured value, and the objective function form is as follows:
in the formula: n is the number of the particle swarm training samples, yiIn order to achieve the target value,for the output value, the smaller the adaptive value of the particle is, the optimal position of the current particle is;
step 3.3, outputting an optimal fuzzy rule layer fitness value by judging whether the updating iteration meets the precision requirement or reaches the maximum iteration times;
in an N-dimensional search space, a population is composed of m particles, Z represents a matrix, ZiThe position of the ith particle is shown, and the speed and the position of the ith particle are respectively:
Vi=(Vi1,Vi2,...,ViN)T (7)
Zi=(Zi1,Zi2,...,ZiN)T (8)
in the formula, ViNRepresenting the velocity of the ith particle in the nth dimension; ziNIndicating the position of the ith particle in the nth dimension; i is 1,2,. m represents the number of particles;
with the operation of each iteration, the particle updates the speed and position of the particle by the individual and global optimal values, and the updating formula is as follows:
in the formula (I), the compound is shown in the specification,respectively representing the speed and the position of the nth dimension in the (k + 1) th iteration; k is the kth iteration, and omega is the inertial weight; n is 1,2,. cndot.n; 1,2, 1, m, c1And c2For the learning factor, alpha is a constraint factor,respectively representing the position of an individual and a global extreme point of the whole population in the nth dimension;
after updating the global optimal values of the individual local part and all the particles, updating the inertia weight, changing the global and local search of the inertia weight, and improving the inertia weight omega in a self-adaptive weight mode, wherein the improved inertia weight calculation formula is as follows:
where ω is the inertial weight, ωmax,ωminRepresenting the maximum and minimum values of the inertial weight, f representing the current fitness value of the particle, fminRepresents the minimum fitness value among all particles; f. ofaverRepresenting the average fitness value in the population of particles.
Step 4.1, carrying out normalization calculation on the fitness of each rule optimized by the dynamic weight particle swarm algorithm in the step 3 to obtain a corresponding normalized fitness value, namely a third normalization layer of the self-adaptive fuzzy neural network;
wherein, Oi 3Normalizing the applicability value for the corresponding rule;the proportion of the applicability of the ith rule to the sum of the applicability of the used rules is shown;
4.2, calculating an output result of each rule of the adaptive fuzzy neural network, namely calculating an output layer of the fourth layer of the adaptive fuzzy neural network;
wherein, Oi 4De-blurring the output value of each node; f. ofiFuzzy rules of the Sugeno type;
4.3, calculating the total output of the whole fuzzy system, namely a fifth layer total output layer of the self-adaptive fuzzy neural network;
wherein, Oi 5The total output value of the whole adaptive fuzzy neural network is obtained; f. ofiFuzzy rules of the Sugeno type;
and 4.4, comparing the data of the abandoned slag dam preprocessed in the step 1 with the data output by the self-adaptive fuzzy neural network system to output data of different groups, so as to achieve the prediction effect of different days.
In step 2.3, the assignment includes the influence radius, the initial step number, the maximum iteration number, the target error, and the step change rate.
In step 3.3, to avoid blind particle search, particles are placed in the first placenOf dimensionThe position variation range is limited to [ -Z [)max,-Zmin]The speed variation range is limited to [ -V ]max,-Vmin]And (4) the following steps.
And 3.3, the value of omega in the particle group algorithm is between 0.4 and 0.9.
The step 1 specifically comprises the following steps: and sorting the original monitoring data according to a time sequence, eliminating measured values of other time periods every day, only keeping one time period measured value, drawing the sorted data sequence, eliminating obvious jump point gross errors, and obtaining the actual monitoring data arranged according to the time sequence.
The method has the advantages that the self-adaptive fuzzy neural network is introduced into the deformation prediction model of the waste slag dam, the dynamic weight particle swarm algorithm is utilized to carry out parameter optimization on the fitness value of the fuzzy rule layer in the self-adaptive fuzzy neural network, and the self-adaptive fuzzy neural network with the optimal fitness value is obtained.
Drawings
FIG. 1 is a schematic diagram of a DPSO-ANFIS-based deformation prediction method for a waste slag dam according to the invention;
FIG. 2 is a flow chart of a particle group algorithm in the DPSO-ANFIS-based waste slag dam deformation prediction method;
FIG. 3 is a design view of the slag discarding dam in this embodiment;
FIG. 4 is a comparison graph of the predicted results of different models of the EL1095 horse way measuring points in this embodiment;
FIG. 5 is a comparison graph of the predicted results of different models of the EL1070 pavement measuring points in the present embodiment;
FIG. 6 is a comparison chart of the predicted results of different models of EL1016 street measuring points in this embodiment.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention discloses a DPSO-ANFIS-based deformation prediction method for a waste slag dam, which comprises the following steps of:
step 1, monitoring the original data of the abandoned slag dam, and preprocessing the original data.
The step 1 specifically comprises the following steps: and sorting the original monitoring data according to a time sequence, eliminating measured values of other time periods every day, only keeping one time period measured value, drawing the sorted data sequence, eliminating obvious jump point gross errors, and obtaining the actual monitoring data arranged according to the time sequence.
And 2, building a self-adaptive fuzzy neural network system, and assigning network parameters in the system.
Step 2.1, firstly, building a self-adaptive fuzzy neural network, inputting the waste slag dam data preprocessed in the step 1 into a first-layer input layer of the self-adaptive fuzzy neural network as a data sample, performing fuzzification processing, and outputting corresponding membership degrees;
in the formula (I), the compound is shown in the specification,the membership degree of the corresponding node;
wherein, muAi(x),μBj(y) are membership functions corresponding to the input variables respectively; i. j is respectively an adaptive node with a node function in the adaptive fuzzy neural network, a bell-shaped membership function is selected, and the formulas (3) and (4) are substituted into the formulas (1) and (2) respectively to obtain membership
And 2.2, calculating the fitness of each rule according to the membership degrees obtained in the step 2.1, taking the membership degree obtained in the first layer as an input signal of a fuzzy rule layer of the second layer, and then accumulating and multiplying the input signal to obtain an output result of the second layer, namely the fitness value.
In the formula, Oi 2To correspond to the fitness value of the rule, ωiThe excitation strength of the ith rule;
and 2.3, selecting an ANFIS initial parameter, and assigning the parameter of the self-adaptive fuzzy neural network through a clustering method, wherein the assignment comprises an influence radius, an initial step number, a maximum iteration number, a target error and a step length change rate.
Step 3, optimizing the parameters of the adaptive fuzzy neural network in the step 2;
and 3, specifically, extracting the fitness value of a second fuzzy rule layer in the self-adaptive fuzzy neural network in the step 2 to serve as a particle swarm initialization parameter to be optimized, solving a space initialization standard particle swarm, and iteratively optimizing dynamically updated inertia weight and algorithm by adopting a dynamic weight particle swarm algorithm to generate an optimal fitness value of the fuzzy rule layer.
The particle swarm optimization and DPSO have high optimization efficiency, and the optimized DPSO does not have the phenomenon of local optimization, so that the optimization capability of the algorithm is continuously maintained, and the efficiency of the algorithm is improved.
The specific steps of the particle group algorithm in step 3 are as shown in fig. 2:
step 3.1, initializing standard particle swarm algorithm parameters, and assigning values to the size of a population, the initial position and speed of particles, the number of iterations, the error precision and the inertia weight;
step 3.2, calculating particle fitness values, calculating the fitness value corresponding to each particle at different positions according to a set objective function, wherein the set objective function takes the Mean Square Error (MSE) of the model prediction output value and the measured value, and the objective function form is as follows:
in the formula: n is the number of the particle swarm training samples, yiIn order to achieve the target value,for the output value, the smaller the adaptive value of the particle is, the optimal position of the current particle is.
Step 3.3, outputting an optimal fuzzy rule layer fitness value by judging whether the updating iteration meets the precision requirement or reaches the maximum iteration times;
in an N-dimensional search space, a population is composed of m particles, Z represents a matrix, ZiThe position of the ith particle is shown, and the speed and the position of the ith particle are respectively:
Vi=(Vi1,Vi2,...,ViN)T (7)
Zi=(Zi1,Zi2,...,ZiN)T (8)
in the formula, ViNRepresenting the velocity of the ith particle in the nth dimension; ziNIndicating the position of the ith particle in the nth dimension; i is 1,2,. m represents the number of particles;
with the operation of each iteration, the particle updates the speed and position of the particle by the individual and global optimal values, and the updating formula is as follows:
in the formula (I), the compound is shown in the specification,respectively representing the speed and the position of the nth dimension in the (k + 1) th iteration; k is the kth iteration, and omega is the inertial weight; n is 1,2,. cndot.n; 1,2, 1, m, c1And c2For the learning factor, alpha is a constraint factor,respectively representing the position of an individual and a global extreme point of the whole population in the nth dimension;
to avoid blind searching of particles, particles are put onnThe position variation range of dimension is limited to [ -Z ]max,-Zmin]The speed variation range is limited to [ -V ]max,-Vmin]Internal; omega in the particle swarm algorithm is usually selected to be 0.4-0.9, and due to the adoption of linear decrement, omega is reduced along with the increase of iteration times, and premature and oscillation phenomena are easily generated near the global optimal solution.
After updating the global optimal values of the individual local part and all the particles, updating the inertia weight, changing the global and local search of the inertia weight, and improving the inertia weight omega in a self-adaptive weight mode, wherein the improved inertia weight calculation formula is as follows:
where ω is the inertial weight, ωmax,ωminRepresenting the maximum and minimum values of the inertial weight, f representing the current fitness value of the particle, fminRepresents the minimum fitness value among all particles; f. ofaverRepresenting the average fitness value in the population of particles.
And 4, substituting the data of the abandoned slag dam preprocessed in the step 1 into the self-adaptive fuzzy neural network for training and prediction, wherein parameters in the self-adaptive fuzzy neural network are optimized through the step 3, calculating prediction precision values, and comparing prediction effects of measuring points at different positions.
Step 4.1, carrying out normalization calculation on the fitness of each rule optimized by the dynamic weight particle swarm algorithm in the step 3 to obtain a corresponding normalized fitness value, namely a third normalization layer of the self-adaptive fuzzy neural network;
wherein, Oi 3Normalizing the applicability value for the corresponding rule;the proportion of the applicability of the ith rule to the sum of the applicability of the used rules is shown;
and 4.2, calculating an output result of each rule of the adaptive fuzzy neural network, namely calculating an output layer by the fourth layer of the adaptive fuzzy neural network.
Wherein, Oi 4De-blurring the output value of each node; f. ofiIs a fuzzy rule of the Sugeno type.
And 4.3, calculating the total output of the whole fuzzy system, namely a fifth layer total output layer of the self-adaptive fuzzy neural network.
Wherein, Oi 5The total output value of the whole adaptive fuzzy neural network is obtained; f. ofiFuzzy rules of the Sugeno type;
and 4.4, comparing the data of the abandoned slag dam preprocessed in the step 1 with the data output by the self-adaptive fuzzy neural network system to output data of different groups, so as to achieve the prediction effect of different days.
Examples
In the embodiment, a certain waste slag dam is positioned in a trench of Zhen' an county and West Mill in Shang Luo city, and the waste slag dam is mainly used for connecting roads between an upper warehouse and a lower warehouse, piling waste slag in excavation of an external traffic passage and a hydraulic tunnel and serving as a later-stage transfer yard. The occupied area of the waste slag dam is 8.12 ten thousand meters 2, and the total amount of the waste slag is 475.41 ten thousand meters3Belongs to a super 1-grade super-large slag yard. The slag yard is built into two platforms, the elevation of each platform is 1005m and 949m, and the design drawing of the slag abandoning dam is shown in figure 3.
Step 1, analyzing dam crest horizontal displacement monitoring data, and preprocessing original data;
monitoring data of monitoring points distributed on the left side parts of abandoned slag dam 1# construction platforms EL1095, EL1070 and EL1016 streets along the gully direction are selected as data samples, and the monitoring frequency is once a day. This time series contains 470 groups of data totaling 2016 from 2.6.2017 to 13.9.2017.
Step 2, building a self-adaptive fuzzy neural network system, and assigning network parameters in the system;
in the embodiment, when the ANFIS initial parameter is selected, the parameters of the adaptive fuzzy neural network are assigned by a subtractive clustering method, where the assignment includes an influence radius, an initial step number, a maximum iteration number, a target error, and a step change rate, and the subtractive clustering parameter specifically takes a value as shown in table 1.
TABLE 1 Subtraction clustering parameter selection Table
Step 3, optimizing the parameters of the adaptive fuzzy neural network in the step 2; and (3) initializing parameters to be optimized by taking the fitness value of a second fuzzy rule layer in the self-adaptive fuzzy neural network in the step (2) as a particle swarm, solving a space initialization standard particle swarm, and iteratively optimizing dynamically updated inertial weight and the algorithm by adopting a dynamic weight particle swarm algorithm to generate an optimal fitness value of the fuzzy rule layer. The particle swarm algorithm parameters take specific values, as shown in table 2.
TABLE 2 particle swarm algorithm parameter value-taking table
And 4, substituting the data of the abandoned slag dam preprocessed in the step 1 into the self-adaptive fuzzy neural network for training and prediction, wherein parameters in the self-adaptive fuzzy neural network are optimized through the step 3, calculating prediction precision values, and comparing prediction effects of measuring points at different positions. The previous 380 sets of data are fuzzified and trained, here, the results of the prediction 30 sets are taken as an example, the prediction results of different models of different measuring points are collated, as shown in fig. 4, fig. 5 and fig. 6, in this embodiment, the difference between the actual value and the prediction value of the present invention is calculated, and the correlation coefficient R, Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) are used to illustrate the high or low of the prediction accuracy.
In order to accurately evaluate the model precision and embody the applicability of the self-adaptive fuzzy neural network in the aspect of the abandoned slag dam deformation prediction method, the superiority of the dynamic weight particle swarm optimization method in the aspects of prediction precision and generalization is explained by comparing the traditional BP neural network and the standard particle swarm optimization method. Comparing the ANFIS model, the PSO-ANFIS model and the BP neural network model, the performance evaluation index of each measuring point is shown in Table 3.
TABLE 3 Performance evaluation chart for each measuring point under different models
As can be seen from the predicted trends in fig. 4, fig. 5 and fig. 6 and the numerical values in table 3, the correlation coefficients of the ANFIS model are all above 0.88 in terms of the prediction accuracy, indicating that the degree of collinearity is better, while the correlation coefficient of the BP model is worst at 0.80, and the degree of collinearity is general. In terms of root mean square error and average absolute error, the DPSO-ANFIS prediction value is smaller than that of other models, and the prediction precision of the method is high. The DPSO-ANFIS has good prediction effect at different position measuring points, the prediction trend accords with the actual data trend, the precision is superior to other models, and good generalization is shown. In the aspect of predicting time series, along with the increase of the number of prediction days, in an EL1095 measuring point, the DPSO-ANFIS model is good in prediction effect and stable in overall prediction performance, and the prediction precision of the other three models is improved. In both measurement points EL1070, EL1016, all methods show a tendency that the prediction accuracy decreases as the number of days of prediction increases. The DPSO-ANFIS model has the best performance in the prediction effect of 3 measuring points, and the prediction precision is in a reasonable range and is superior to other methods.
The adaptive fuzzy neural network is optimized by adopting a dynamic weight particle swarm algorithm, the adaptability value of a fuzzy layer in the adaptive fuzzy neural network is formed, the adaptive fuzzy neural network for searching the optimal adaptability value is formed, the adaptive fuzzy neural network is introduced into the abandoned slag dam deformation prediction method by combining with actual engineering monitoring data, the adaptive fuzzy neural network is optimized by adopting the dynamic weight particle swarm algorithm, the prediction precision is improved, and the adaptive fuzzy neural network has good generalization and reliable stability; compared with the traditional BP neural network and ANFIS methods, the self-adaptive fuzzy neural network model optimized by the dynamic weight particle swarm optimization overcomes the defects of insufficient optimization degree and low prediction precision of the standard particle swarm optimization, and meanwhile, the method has better performance in measuring points at different positions and predicting time sequences, is superior to other methods in overall stability and prediction effect, and is suitable for deformation prediction of abandoned dams.
Claims (9)
1. The waste slag dam deformation prediction method based on DPSO-ANFIS is characterized by comprising the following steps:
step 1, monitoring the original data of the abandoned slag dam, and preprocessing the original data;
step 2, building a self-adaptive fuzzy neural network system, and assigning network parameters in the system;
step 3, optimizing the parameters of the adaptive fuzzy neural network in the step 2;
and 4, substituting the data of the abandoned slag dam preprocessed in the step 1 into the self-adaptive fuzzy neural network for training and prediction, wherein parameters in the self-adaptive fuzzy neural network are optimized through the step 3, calculating prediction precision values, and comparing prediction effects of measuring points at different positions.
2. The DPSO-ANFIS-based waste slag dam deformation prediction method as claimed in claim 1, wherein the step 2 is specifically as follows:
step 2.1, firstly, building a self-adaptive fuzzy neural network, inputting the waste slag dam data preprocessed in the step 1 into a first-layer input layer of the self-adaptive fuzzy neural network as a data sample, performing fuzzification processing, and outputting corresponding membership degrees;
in the formula (I), the compound is shown in the specification,the membership degree of the corresponding node;
wherein, muAi(x),μBj(y) are membership functions corresponding to the input variables respectively; i. j is respectively an adaptive node with a node function in the adaptive fuzzy neural network, a bell-shaped membership function is selected, and the formulas (3) and (4) are substituted into the formulas (1) and (2) respectively to obtain membership
Step 2.2, calculating the fitness of each rule according to the membership degrees obtained in the step 2.1, taking the membership degree obtained in the first layer as an input signal of a fuzzy rule layer of the second layer, and then accumulating and multiplying the input signal to obtain an output result of the second layer, namely the fitness value;
in the formula (I), the compound is shown in the specification,to correspond to the fitness value of the rule, ωiThe excitation strength of the ith rule;
and 2.3, selecting an ANFIS initial parameter, and assigning the parameter of the self-adaptive fuzzy neural network through a clustering method.
3. The DPSO-ANFIS-based waste slag dam deformation prediction method is characterized in that the concrete step of the step 3 is to initialize parameters to be optimized by extracting the fitness value of a second fuzzy rule layer in the self-adaptive fuzzy neural network in the step 2 as a particle swarm, solve a space initialization standard particle swarm, adopt a dynamic weight particle swarm algorithm to iteratively optimize dynamically updated inertial weight and the algorithm, and generate the optimal fitness value of the fuzzy rule layer.
4. The DPSO-ANFIS-based waste slag dam deformation prediction method as claimed in claim 3, wherein the particle group algorithm in the step 3 comprises the following specific steps:
step 3.1, initializing standard particle swarm algorithm parameters, and assigning values to the size of a population, the initial position and speed of particles, the number of iterations, the error precision and the inertia weight;
step 3.2, calculating particle fitness values, calculating the fitness value corresponding to each particle at different positions according to a set objective function, wherein the set objective function takes the mean square error between the model prediction output value and the measured value, and the form of the objective function is as follows:
in the formula: n is the number of the particle swarm training samples, yiIn order to achieve the target value,for the output value, the smaller the adaptive value of the particle is, the optimal position of the current particle is;
step 3.3, outputting an optimal fuzzy rule layer fitness value by judging whether the updating iteration meets the precision requirement or reaches the maximum iteration times;
in an N-dimensional search space, a population is composed of m particles, Z represents a matrix, ZiThe position of the ith particle is shown, and the speed and the position of the ith particle are respectively:
Vi=(Vi1,Vi2,...,ViN)T (7)
Zi=(Zi1,Zi2,...,ZiN)T (8)
in the formula, ViNRepresenting the velocity of the ith particle in the nth dimension; ziNIndicating the position of the ith particle in the nth dimension; i is 1,2,. m represents the number of particles;
with the operation of each iteration, the particle updates the speed and position of the particle by the individual and global optimal values, and the updating formula is as follows:
in the formula (I), the compound is shown in the specification,respectively representing the speed and the position of the nth dimension in the (k + 1) th iteration; k is the kth iteration, and omega is the inertial weight; n is 1,2,. cndot.n; 1,2, 1, m, c1And c2For the learning factor, alpha is a constraint factor,respectively representing the position of an individual and a global extreme point of the whole population in the nth dimension;
after updating the global optimal values of the individual local part and all the particles, updating the inertia weight, changing the global and local search of the inertia weight, and improving the inertia weight omega in a self-adaptive weight mode, wherein the improved inertia weight calculation formula is as follows:
where ω is the inertial weight, ωmax,ωminRepresenting the maximum and minimum values of the inertial weight, f representing the current fitness value of the particle, fminRepresents the minimum fitness value among all particles; f. ofaverRepresenting the average fitness value in the population of particles.
5. The DPSO-ANFIS-based waste slag dam deformation prediction method as claimed in claim 4, wherein the step 4 is as follows: step 4.1, carrying out normalization calculation on the fitness of each rule optimized by the dynamic weight particle swarm algorithm in the step 3 to obtain a corresponding normalized fitness value, namely a third normalization layer of the self-adaptive fuzzy neural network;
wherein the content of the first and second substances,normalizing the applicability value for the corresponding rule;the proportion of the applicability of the ith rule to the sum of the applicability of the used rules is shown;
4.2, calculating an output result of each rule of the adaptive fuzzy neural network, namely calculating an output layer of the fourth layer of the adaptive fuzzy neural network;
wherein the content of the first and second substances,de-blurring the output value of each node; f. ofiFuzzy rules of the Sugeno type;
4.3, calculating the total output of the whole fuzzy system, namely a fifth layer total output layer of the self-adaptive fuzzy neural network;
wherein, Oi 5The total output value of the whole adaptive fuzzy neural network is obtained; f. ofiFuzzy rules of the Sugeno type;
and 4.4, comparing the data of the abandoned slag dam preprocessed in the step 1 with the data output by the self-adaptive fuzzy neural network system to output data of different groups, so as to achieve the prediction effect of different days.
6. The DPSO-ANFIS-based deformation prediction method for the abandoned slag dam according to claim 2, wherein in the step 2.3, the assignment comprises an influence radius, an initial step number, a maximum iteration number, a target error and a step size change rate.
7. The DPSO-ANFIS-based deformation prediction method for slag abandoning dam as claimed in claim 3, wherein in the step 3.3, in order to avoid blind particle search, the particles are arranged at the second positionnThe position variation range of dimension is limited to [ -Z ]max,-Zmin]The speed variation range is limited to [ -V ]max,-Vmin]And (4) the following steps.
8. The DPSO-ANFIS-based waste slag dam deformation prediction method as claimed in claim 3, wherein in the step 3.3, the value of omega in the particle group algorithm is between 0.4 and 0.9.
9. The DPSO-ANFIS-based waste slag dam deformation prediction method as claimed in claim 1, wherein the step 1 is specifically as follows: and sorting the original monitoring data according to a time sequence, eliminating measured values of other time periods every day, only keeping one time period measured value, drawing the sorted data sequence, eliminating obvious jump point gross errors, and obtaining the actual monitoring data arranged according to the time sequence.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110485694.2A CN113139692A (en) | 2021-04-30 | 2021-04-30 | Waste slag dam deformation prediction method based on DPSO-ANFIS |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110485694.2A CN113139692A (en) | 2021-04-30 | 2021-04-30 | Waste slag dam deformation prediction method based on DPSO-ANFIS |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113139692A true CN113139692A (en) | 2021-07-20 |
Family
ID=76816832
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110485694.2A Withdrawn CN113139692A (en) | 2021-04-30 | 2021-04-30 | Waste slag dam deformation prediction method based on DPSO-ANFIS |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113139692A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113283183A (en) * | 2021-07-08 | 2021-08-20 | 国网新源控股有限公司 | Waste slag dam deformation prediction method based on DPSO-ANFIS |
-
2021
- 2021-04-30 CN CN202110485694.2A patent/CN113139692A/en not_active Withdrawn
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113283183A (en) * | 2021-07-08 | 2021-08-20 | 国网新源控股有限公司 | Waste slag dam deformation prediction method based on DPSO-ANFIS |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107967803A (en) | Traffic congestion Forecasting Methodology based on multi-source data and variable-weight combined forecasting model | |
CN103226741B (en) | Public supply mains tube explosion prediction method | |
CN111882869B (en) | Deep learning traffic flow prediction method considering adverse weather | |
CN1760912A (en) | Modeling method of uncertain hydraulics model for urban seweage and drainage system | |
CN111461457A (en) | Foundation pit displacement prediction method based on particle swarm optimization BP neural network | |
CN112801342A (en) | Adaptive runoff forecasting method based on rainfall runoff similarity | |
CN114792071B (en) | Drainage pump station optimal scheduling method based on machine learning technology | |
CN112784140B (en) | Search method of high-energy-efficiency neural network architecture | |
CN112100927A (en) | Prediction method for slope deformation and soft soil foundation settlement based on GA-BP neural network | |
CN113190945B (en) | Urban drainage pipe network optimization method based on online agent model assisted evolution algorithm | |
CN112765902B (en) | Soft measurement modeling method for COD concentration in rural domestic sewage treatment process based on TentFWA-GD RBF neural network | |
CN113221439B (en) | BP neural network-based drainage system real-time calibration and dynamic prediction method | |
CN115374995A (en) | Distributed photovoltaic and small wind power station power prediction method | |
CN111582541A (en) | Firefly algorithm-based inland inundation model prediction method | |
CN113554466A (en) | Short-term power consumption prediction model construction method, prediction method and device | |
CN113807545A (en) | River and lake ecological flow forecasting and early warning method based on deep learning and physical model | |
CN113283183A (en) | Waste slag dam deformation prediction method based on DPSO-ANFIS | |
CN116050628A (en) | Reservoir flood control optimization scheduling method considering forecast uncertainty based on digital twinning | |
CN116933946A (en) | Rail transit OD passenger flow prediction method and system based on passenger flow destination structure | |
CN113139692A (en) | Waste slag dam deformation prediction method based on DPSO-ANFIS | |
Wang et al. | Discussion on the prediction of engineering cost based on improved BP neural network algorithm | |
CN109214610A (en) | A kind of saturation Methods of electric load forecasting based on shot and long term Memory Neural Networks | |
CN112819087B (en) | Method for detecting abnormality of BOD sensor of outlet water based on modularized neural network | |
CN113869606A (en) | Method for garbage land deployment modeling based on urban population mobility | |
CN113297805A (en) | Wind power climbing event indirect prediction method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210720 |