CN114037075A - Diesel engine electronic speed regulation self-adaption method based on artificial intelligence algorithm - Google Patents
Diesel engine electronic speed regulation self-adaption method based on artificial intelligence algorithm Download PDFInfo
- Publication number
- CN114037075A CN114037075A CN202111340944.XA CN202111340944A CN114037075A CN 114037075 A CN114037075 A CN 114037075A CN 202111340944 A CN202111340944 A CN 202111340944A CN 114037075 A CN114037075 A CN 114037075A
- Authority
- CN
- China
- Prior art keywords
- neural network
- algorithm
- diesel engine
- self
- training
- 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.)
- Granted
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 86
- 238000000034 method Methods 0.000 title claims abstract description 71
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 28
- 238000013528 artificial neural network Methods 0.000 claims abstract description 62
- 238000012549 training Methods 0.000 claims abstract description 57
- 238000005457 optimization Methods 0.000 claims abstract description 19
- 239000011159 matrix material Substances 0.000 claims abstract description 11
- 238000012545 processing Methods 0.000 claims abstract description 11
- 239000002245 particle Substances 0.000 claims abstract description 7
- 230000006870 function Effects 0.000 claims description 30
- 230000003044 adaptive effect Effects 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 13
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 11
- 241001206881 Myrmeleon inconspicuus Species 0.000 claims description 8
- 230000002068 genetic effect Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000007405 data analysis Methods 0.000 claims description 4
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000003491 array Methods 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000007619 statistical method Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 6
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 241000258923 Neuroptera Species 0.000 description 7
- 230000008901 benefit Effects 0.000 description 7
- 230000004044 response Effects 0.000 description 7
- 238000007796 conventional method Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 238000002759 z-score normalization Methods 0.000 description 4
- 241000257303 Hymenoptera Species 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000013506 data mapping Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000011425 standardization method Methods 0.000 description 1
Images
Classifications
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D31/00—Use of speed-sensing governors to control combustion engines, not otherwise provided for
- F02D31/001—Electric control of rotation speed
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- Feedback Control In General (AREA)
Abstract
An electronic speed regulation self-adaptive method of a diesel engine based on an artificial intelligence algorithm comprises the steps of firstly carrying out standardization processing on current and rotating speed data of the diesel engine, dividing a data set, optimizing initial characteristic parameters by adopting a particle swarm algorithm before neural network training, then constructing a BP neural network based on an objective function and the optimized characteristic parameters, optimizing the network, reconstructing the characteristic parameters according to a weight matrix and an offset value output by the BP neural network, carrying out self-adaptive optimization by utilizing the intelligent algorithm, optimally outputting the characteristic parameters of each node, and finally carrying out self-adaptive control on the rotating speed of the diesel engine by utilizing the characteristic of low delay of the invention. The invention optimizes the initial parameters of the neural network and finally reconstructs the optimized output of the characteristic parameters through an intelligent algorithm, reduces the decision time to the millisecond level, improves the training speed and the precision of the neural network, further leads the fluctuation of the rotating speed of the diesel engine to be smaller through self-adaptive control, has better control effect and is beneficial to the practical engineering application.
Description
Technical Field
The invention belongs to the technical field of self-adaptive control of the rotating speed of a diesel engine, and particularly relates to a method for realizing self-adaptive control of the rotating speed of the diesel engine by combining a BP neural network and an artificial intelligence algorithm.
Background
The diesel engine is used as a traditional power mechanical device, and the output power of the diesel engine is required to be changed along with the change of the load of the traction motor along with the external environment. At present, the regulation of the rotating speed is usually controlled manually, the response speed is slow, and the regulation range is only within a certain limit, so that the optimal balance point of the rotating speed and the power cannot be reached. The traditional method for establishing a physical model or setting an observer has the prediction accuracy of 80-90% and the response speed of about 40 milliseconds, and cannot meet the requirement of low-delay self-adaptive control on the rotating speed of a diesel engine. The traditional method needs to fully understand the working mechanism of the diesel engine, and the same accurate prediction effect cannot be obtained on different types of engines due to the complexity of model building of the traditional method, so that the generalization is not strong, and the rapid and accurate regulation and control of the rotating speed of the diesel engine cannot be realized under the condition of highly complex working conditions. The traditional method combining machine learning and statistical learning improves the generalization ability of the statistical learning method by utilizing the characteristic extraction ability of the machine learning, and can not fundamentally solve the problem of rotating speed regulation of the diesel under complex working conditions. At present, with the progress of a diesel engine state real-time monitoring system and the accumulation of a large amount of engineering data, implicit characteristics in a sample can be fully excavated through a neural network, and the optimal solution of state parameters is realized on the basis, but the problem of adaptive control cannot be solved.
Disclosure of Invention
The invention aims to improve the electric control performance of a diesel engine and improve the response speed and the accuracy of the diesel engine, and provides an artificial intelligence algorithm-based self-adaptive method for electronic speed regulation of the diesel engine.
The invention realizes the quick response of the diesel engine control unit and the adaptive tracking of the characteristic parameters by a simple, real-time, high-precision and high-efficiency adaptive electric control speed regulation method.
In order to achieve the purpose, the invention adopts the technical scheme that:
an artificial intelligence algorithm-based electronic speed regulation self-adaptive method for a diesel engine comprises the following steps:
step 1: carrying out standardization processing on diesel engine current and rotating speed data acquired by a sensor arranged on a diesel engine by adopting a Z-Score method to obtain statistical distribution of summarized unified samples;
step 2: dividing the data after the standardization processing in the step 1 into a training set and a testing set, and optimizing initial characteristic parameters of the neural network by adopting a particle swarm algorithm (PSO algorithm) before the neural network training;
and step 3: constructing a BP neural network based on an objective function and the initial characteristic parameters obtained by adopting the PSO algorithm in the step 2, carrying out network training by utilizing a training set, adding L2 regularization to a loss function during forward propagation, and carrying out reverse optimization by applying an Adam optimizer in the process of backward propagation of the BP neural network;
and 4, step 4: reconstructing a characteristic parameter value according to the weight matrix and the offset value output by the BP neural network in the step 3, and applying a ant lion-genetic algorithm, namely ALO-GA to carry out self-adaptive optimization on the characteristic parameters of each output node, and finally carrying out self-adaptive control on the rotating speed of the diesel engine by utilizing the characteristics of low delay and high precision of the invention;
further preferably, in step 2, the specific method for dividing the data after the normalization processing into the training set and the test set is that, according to 8: and 2, randomly dividing the ratio into a training set and a test set, taking the training set as the input of the next BP neural network training, and using the test set for finally verifying the accuracy.
Preferably, in step 2, the specific content and method for optimizing the initial characteristic parameters of the neural network by using the particle swarm optimization, i.e. the PSO algorithm, before the neural network training is performed, includes predicting by using a large amount of acquired engineering data, narrowing the training range of the BP neural network, accelerating the training speed of the BP neural network, and performing the PSO algorithm,The prediction precision is improved; the initial characteristic parameters of the optimization are as follows: diesel engine speed PkBalance current value IbacDiesel engine parameter KacPeriod TsReal time current I(k-t)。
Preferably, in step 3, the specific content and method for constructing the BP neural network based on the objective function and the initial characteristic parameter obtained by using the PSO algorithm in step 2 are that the initial characteristic parameter processed by the PSO algorithm, i.e. the balance current value IbacDiesel engine parameter KacThe feature array whose dimension is two-dimensional is composed according to the physical characteristics [ [ x1, x2 ]],[x3,x4],···[xN-1,xN]]And training the BP neural network as input data, and continuously correcting the weight matrix and the offset value through a back propagation process.
Further preferably, the BP neural network constructed based on the objective function and the initial characteristic parameters obtained by using the PSO algorithm in step 2 in step 3 is a BP neural network having a hidden layer and multiple inputs and multiple outputs; the structure and function of the BP neural network comprise the following steps:
an input layer: the processed data, i.e. the equilibrium current value IbacDiesel engine parameter KacConverting into 2 two-dimensional arrays as input of a hidden layer;
a hidden layer; inputting the processed data into a hidden layer of a BP neural network for calculation, completing forward propagation at the same time, and adjusting a weight matrix W and a bias b;
an output layer: activating the output of the BP neural network hidden layer through a tanh function to obtain the weight calculated this time, and calculating a weight matrix derivative value required by back propagation;
adjusting network parameters: setting internal initial parameters of the BP neural network, optimizing a training process through an Adam algorithm in a network training process, comparing prediction and real data at each moment, calculating a loss function value in a training set, and gradually reducing the loss function value through back propagation so as to realize convergence of characteristic parameters.
Further preferably, in step 3, the loss function in the forward propagation of the BP neural network is normalized by adding L2 to prevent the over-fitting phenomenon, the backward propagation process adopts an Adam algorithm to replace a traditional gradient descent method, a method for adaptively changing the learning rate is used, and the decision time is gradually reduced to a millisecond level through deep self-learning.
Preferably, in step 4, the specific content and method for reconstructing and outputting the characteristic parameters through the optimized data obtained by the BP neural network in step 3 is to perform adaptive tracking of the characteristic parameters by applying a multi-characteristic adaptive learning method based on an artificial intelligence algorithm, i.e. an ALO-GA algorithm; the characteristic parameter searching range is narrowed through an ant lion algorithm, namely an ALO algorithm, then an adaptive genetic algorithm, namely a GA algorithm is adopted, and the self-adaptive function of the cross rate and the variation rate according to the data convergence characteristics is added, so that the convergence speed is increased, and the prediction precision is improved.
Preferably, in step 4, a statistical analysis method is used for carrying out data analysis on the prediction result output by the artificial intelligence algorithm, the test set data is used for comparing the characteristic parameter predicted value with the true value, an absolute error value is calculated, and finally the characteristics of low delay and high precision are used for carrying out the self-adaptive control on the rotating speed of the diesel engine.
The invention has the advantages and beneficial effects that:
(1) the invention adopts the particle swarm algorithm to generate the initial parameters of the neural network to be trained, prepares for the next training of the neural network, and greatly improves the training speed of the neural network and the accuracy of the final result.
(2) The method of combining network initial parameter setting and Adam back propagation optimization is adopted, decision time is gradually reduced to 3-7 milliseconds through deep self-learning, training speed of the neural network is remarkably improved compared with a traditional method, and the requirement of low delay is met.
(3) The invention provides a multi-feature self-adaptive learning method based on an artificial intelligence algorithm. Meanwhile, in order to improve the convergence rate of the algorithm, a self-adaptive range search function is added into the algorithm, and the target value search range is quickly reduced through machine learning, so that the self-adaptive method achieves better balance in the search width and depth, the prediction error is reduced to be within 3%, and the accuracy is improved by over 10% compared with the traditional method.
Drawings
FIG. 1 is a general flow chart of an electronic speed regulation adaptive method of a diesel engine based on an artificial intelligence algorithm,
FIG. 2 is a detailed flow chart of the diesel engine electronic speed regulation self-adapting method based on the artificial intelligence algorithm,
FIG. 3 is a graph of error for the Kac conventional method,
figure 4 is a graph of the error of the Kac adaptation method,
figure 5 is an error graph of the Ibac conventional method,
figure 6 is an error diagram of the Ibac adaptation method,
figure 7 is a graph of the error of the conventional method,
figure 8 is a graph of the error for the t-adaptation method,
FIG. 9 is a schematic diagram of an intelligent algorithm of a deep learning training network.
Detailed Description
So that the manner in which the above recited objects, features and advantages of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to the embodiments thereof which are illustrated in the appended drawings. It should be noted that the present invention may be embodied in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
For the convenience of understanding of the embodiments of the present invention, the following description will be further explained by taking specific embodiments as examples with reference to the drawings, and the drawings are not to be construed as limiting the embodiments of the present invention.
The embodiment provides an electronic speed regulation self-adaptive method of a diesel engine based on an artificial intelligence algorithm aiming at the problem of response delay between a current value (I) and an engine rotating speed value (P) in a diesel engine system. The rotating speed of the diesel engine can be tracked and controlled in a self-adaptive manner by the element through an artificial intelligence algorithm, and the electric control performance of the diesel engine is improved. In the embodiment, the relationship between the current value (I) and the diesel engine rotation speed value (P) is expressed by a simple first-order system:
Pk-Pk-1＝(I(k-t)-Ibac)·Ts·Kac (1)
wherein I is current, P is diesel engine speed, Ts is period (0.00625s), t is delay period, Ibac is balance current, and Kac is constant. The embodiment requires that:
(1) obtaining constant values of three parameters t, Ibac and Kac in the formula (1);
(2) the method can adaptively track the actual state of the diesel engine (namely dynamic parameters t _ k, Ibac _ k and Kac _ k);
as shown in fig. 1 and 2, an artificial intelligence algorithm-based electronic speed regulation adaptive method for a diesel engine, which takes a diesel engine of a certain model as an application object, comprises the following specific steps:
1. the acquired current and rotation speed data are preprocessed, and raw input data are normalized through a Z-Score normalization method, wherein the Z-Score normalization is data normalization based on the mean value and the standard deviation of the raw data. The transfer function is as follows:
where μ is the mean and σ is the standard deviation. Z-score normalization results in all data clustered around 0 with a variance of 1.
The raw data will be uniformly scaled to a statistical probability distribution between 0-1 by the Z-Score normalization process, so that the mean of the input signals for all samples is close to 0 or small compared to their mean squared error. This avoids the following situations: when the input signals of all samples are positive values, the weights connected to the first hidden layer neurons can only be increased or decreased simultaneously, resulting in a slow learning rate. In addition, singular sample data often exists in data, and the network training time is increased due to the existence of the singular sample data, and the network can not be converged possibly.
Aiming at the input original acquisition data, the Z-score standardization method is adopted to carry out normalization processing on the original data so as to obtain statistical distribution of generalized uniform samples, and preparation is made for carrying out BP neural network training in the next step.
2. And forming a training set and a testing set by the preprocessed data, and according to the ratio of 8: 2, the training set is used as the input of the BP neural network, the test set is used for evaluating the final model prediction precision, and a PSO algorithm (particle swarm optimization) is adopted to generate initial parameters of the BP neural network before the BP neural network is trained, the parameter generation is mainly performed to avoid experimental errors caused by manual selection of the initial parameters, so that the time of experimenters is saved, and meanwhile, the accuracy of the algorithm is improved to the greatest extent, so that a more excellent model is obtained in the next BP network training.
3. Constructing a BP neural network based on the target function and the initial parameters obtained by adopting the PSO algorithm in the step 2, adjusting the parameters of the BP neural network and training, wherein the structure and the function of the BP neural network comprise the following steps:
step (3.1) input layer: data I after being processedbac、KacConverting into 2 two-dimensional arrays as input of a hidden layer;
step (3.2) hiding the layer; inputting the processed data into a hidden layer of a BP neural network for calculation, completing forward propagation at the same time, and adjusting a weight matrix W and a bias b;
step (3.3) output layer: activating the output of the BP neural network hidden layer through a tanh function to obtain the weight calculated this time, and calculating a weight matrix derivative value required by back propagation;
and (3.4) adjusting network parameters: setting internal initial parameters of the BP neural network, optimizing a training process through an Adam algorithm in a network training process, comparing prediction and real data at each moment, calculating a loss function value in a training set, and gradually reducing the loss function value through back propagation so as to realize convergence of characteristic parameters.
In the back propagation stage of the BP neural network, an Adam updating method is applied to replace the traditional gradient descent method, which is essentially RMSProp with momentum terms, and the learning rate of each parameter is dynamically adjusted by utilizing the first moment estimation and the second moment estimation of the gradient, so that the network learning is accelerated by optimizing the learning rate. Adam has the advantages that after bias correction, the learning rate of each iteration has a certain range, so that the parameters are stable, and the formula is as follows:
mt＝μ*mt-1+(1-μ)*gt (3)
wherein m ist、ntRespectively, first and second moment estimates of the gradient, it can be seen that for the expected E | gt|,
(ii) an estimate of (d);
is to mt、ntSo that it can be approximated as an unbiased estimate of the expectation. It can be seen that the moment estimates directly on the gradient can be dynamically adjusted according to the gradient, while (7) forms a dynamic constraint on the learning rate and has a well-defined range. Adam combines the advantages of Adagrad being good at processing sparse gradients and RMSprop being good at processing non-stationary targets, calculates different adaptive learning rates for different parameters, and has a large and special data set in a diesel engineThe convergence effect is good under the condition of higher eigenvalue dimensionality.
In order to prevent overfitting, the weight regularization level is controlled, L2 regularization is adopted for optimization after comparison, influence of noisy features on final prediction accuracy is prevented, and loss functions added after L2 regularization are as follows:
where m is the number of samples, hw(xi) Is the predicted value of the ith sample, yiIs the true value of the ith sample, λ is the penalty coefficient, Wj 2Is the square of the weight value. The regularization is beneficial to eliminating the great influence of a single or a small number of characteristics on the final prediction result, and has obvious advantages of preventing overfitting and improving the prediction precision.
The artificial intelligence network training deep technology comprises the following main steps:
the first step is as follows: the data normalization processing is carried out at the input end of the training network, and an array vector is applied to the calibration and the differentiation of the steady-state data and the dynamic data in the initial stage of data training, but the training is not participated in. When steady-state data participate in training, no time delay exists; when the dynamic data participates in training, the input ends are respectively the sampling values of two moments.
The second step is that: and (3) carrying out structural research and correlation analysis on the data mathematical model. The project diesel engine is regarded as a 'data black box', mass data are used as samples, a deep learning network model of the diesel engine is built, and the problems existing in the traditional modeling method are solved. Firstly, analyzing parameter correlation of a data model, and determining a mathematical structure rough frame of a training parameter to be used as a training initial venation; and subsequently, automatically adjusting the frame structure according to the training result feedback.
The third step: and (3) carrying out deep learning modeling and training on data mapping relations of the expected reconstruction characteristic parameters, the training data, the position parameters and the current parameters. In consideration of the complex working condition and the strongly nonlinear data of the diesel engine in reality, the network layer of the project is selected from 3-10, the network error is expressed as a loss function, and parameters of each layer are adjusted layer by layer through reverse transfer.
The input layer and the output layer include: position parameter and current parameter
And outputting a result: reconstructing characteristic parameters
And (3) learning results: an electrically controlled speed regulation model of a diesel engine is shown in figure 9.
The fourth step: and analyzing the relation between the model structure and the back physical parameters. The trained parameter model structure is used for carrying out physical mechanism explanation on the physical meanings of the parameters and checking whether the physical meanings accord with the basic physical rule and the diesel engine operation rule or not; and verifying the model according to the standard conforming to the deep learning training result and the physical mechanism interpretation result.
4. And (3) reconstructing and outputting the characteristic parameters through the data obtained in the step (3), and carrying out adaptive tracking on the characteristic parameters by applying a multi-characteristic adaptive learning method based on an artificial intelligence algorithm (ALO-GA algorithm). On one hand, the algorithm improves the local optimization capability of the algorithm through cross operation, and on the other hand, the global optimization capability of the algorithm is improved through mutation operation, so that the search is well balanced in the breadth and depth. The method comprises the steps of utilizing the excellent search range reduction performance of the ant lion algorithm, reducing the characteristic value parameter range to be input into the genetic algorithm to be within 10% of error, and then inputting the optimized parameter into the genetic algorithm, so that the genetic algorithm can calculate the optimal fitness value within a smaller range, and the time of population iteration is reduced. The ant lion algorithm has the following specific formula:
in the formula CtFor the minimum of all variables at the t-th iteration, dtThe maximum value of all variables in the t iteration;
the position of the j-th ant lion selected at the t-th iteration.
The ant lion algorithm firstly determines the number and variable dimension of ants and ant lions, randomly initializes the positions of the ants and the ant lions in a feasible region, calculates corresponding fitness values, selects the ant lions with the best fitness in the initialized ant lions population as elite ant lions, selects one ant lion for each ant through roulette, recalculates the fitness values of the ants and the ant lions after each iteration, updates the ant lion positions according to the ant positions and the fitness, and sets the position with the best fitness as the position of a new elite ant lions. And finally, judging whether the maximum iteration times are reached or whether the search range is reduced to the required range, and then finishing the iteration, wherein the search range of the characteristic parameters is reduced to be within 10% of the error value, so as to perform preliminary data optimization for the subsequent genetic algorithm.
On the basis, the genetic algorithm also increases the self-adapting function of the cross rate and the variation rate, and the formula is as follows:
wherein P iscAs the probability of population crossing, PmIs the probability of population variation, fcFor the individuals with higher fitness value in the two crossed (or variant) selected individuals, fmaxAnd fminThe maximum value and the minimum value of the fitness value in the population are respectively, the value of k is selected by a user, and k is selected in the embodiment1＝0.9，k2＝0.9，k3＝0.7，k4＝0.005，k5＝0.01，k6The k value is verified to be good in selecting effect in multiple experiments, the prediction accuracy error can be controlled to be below 3%, and the response speed is controlled to be within 3 milliseconds. The algorithm can improve the local optimization capability of the algorithm by changing the cross probability and the variation probability under different fitness values in the early stage of operation and improve the global optimization capability of the algorithm by changing the variation probability in the later stage, and the addition of self-adaption obviously improves the overall optimization capability of the algorithmThe calculation efficiency is improved, the convergence rate of the algorithm is accelerated to a certain extent, and the prediction precision is correspondingly improved.
After the self-adaptive artificial intelligence algorithm is used for carrying out multi-characteristic parameter self-adaptive optimization and optimizing and outputting the characteristic parameters of each node, the prediction result output by the artificial intelligence algorithm is subjected to data analysis by a statistical analysis method, the predicted value of the characteristic parameters is compared with the true value by using test set data, and the traditional method is compared with the prediction I of the inventionbac、KacThe t result is shown in fig. 3, fig. 4, fig. 5, fig. 6, fig. 7 and fig. 8, and the result of comparison with the prediction result of the conventional method is shown in table 1, the algorithm can be greatly improved in response speed through statistical data analysis, the requirement of low delay is realized, the prediction error is controlled within 3% -4% in precision, and the experimental effect proves that the method is obviously improved in convergence speed and prediction precision compared with the conventional method. Therefore, the method has the remarkable advantages of low delay and high precision.
TABLE 1
Although the embodiments of the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to the above-described embodiments and application fields, and the above-described embodiments are illustrative, instructive, and not restrictive. Those skilled in the art, having the benefit of this disclosure, may effect numerous modifications thereto without departing from the scope of the invention as defined by the appended claims.
Claims (8)
1. An artificial intelligence algorithm-based electronic speed regulation self-adaptive method for a diesel engine is characterized by comprising the following steps:
(1) carrying out standardization processing on diesel engine current and rotating speed data acquired by a sensor arranged on a diesel engine by adopting a Z-Score method to obtain statistical distribution of summarized unified samples;
(2) dividing the data after the standardization processing into a training set and a testing set, and optimizing initial characteristic parameters of the neural network by adopting a particle swarm algorithm (PSO algorithm) before the neural network training;
(3) constructing a BP neural network based on the target function and the initial characteristic parameters obtained by adopting the PSO algorithm in the step (2), carrying out network training by utilizing a training set, adding L2 regularization to the loss function during forward propagation, and carrying out backward propagation optimization by applying an Adam optimizer in the backward propagation process of the BP neural network;
(4) and (4) reconstructing a characteristic parameter value according to the weight matrix and the offset value output by the BP neural network in the step (3), and applying a ant lion-genetic algorithm, namely ALO-GA, to perform self-adaptive optimization on the characteristic parameters of each output node, and finally performing self-adaptive control on the rotating speed of the diesel engine by using the characteristics of low delay and high precision.
2. The diesel engine electronic speed regulation self-adaption method based on the artificial intelligence algorithm as claimed in claim 1, wherein in the step (2), the specific method for dividing the data after the standardization processing into the training set and the test set is that according to 8: and 2, randomly dividing the ratio into a training set and a test set, taking the training set as the input of the next BP neural network training, and using the test set for finally verifying the accuracy.
3. The diesel engine electronic speed regulation self-adaptive method based on the artificial intelligence algorithm according to claim 1, characterized in that in the step (2), the specific content and method for optimizing the initial characteristic parameters of the neural network by adopting the Particle Swarm Optimization (PSO) algorithm before the neural network training are that a large amount of acquired engineering data is adopted for prediction, the training range of the BP neural network is narrowed, the training speed of the BP neural network is accelerated, and the prediction precision is improved; the initial characteristic parameters of the optimization are as follows: diesel engine speed PkBalance current value IbacDiesel engine parameter KacPeriod TsReal time current I(k-t)。
4. The adaptive method for electronic speed regulation of a diesel engine based on an artificial intelligence algorithm as claimed in claim 1, wherein in step (3), the specific content and method for constructing the BP neural network based on the objective function and the initial characteristic parameters obtained by adopting the PSO algorithm in step (2) are that the initial characteristic parameters processed by the PSO algorithm, namely the balance current value IbacDiesel engine parameter KacThe feature array whose dimension is two-dimensional is composed according to the physical characteristics [ [ x1, x2 ]]，[x3，x4]，…[xN-1，xN]]And training the BP neural network as input data, and continuously correcting the weight matrix and the offset value through a back propagation process.
5. The diesel engine electronic speed regulation self-adaption method based on the artificial intelligence algorithm as claimed in claim 1, wherein in the step (3), the BP neural network constructed based on the objective function and the initial characteristic parameters obtained by adopting the PSO algorithm in the step (2) is a BP neural network with a hidden layer and multiple inputs and multiple outputs; the structure and function of the BP neural network comprise the following steps:
an input layer: the processed data, i.e. the equilibrium current value IbacDiesel engine parameter KacConverting into 2 two-dimensional arrays as input of a hidden layer;
a hidden layer; inputting the processed data into a hidden layer of a BP neural network for calculation, completing forward propagation at the same time, and adjusting a weight matrix W and a bias b;
an output layer: activating the output of the BP neural network hidden layer through a tanh function to obtain the weight calculated this time, and calculating a weight matrix derivative value required by back propagation;
adjusting network parameters: setting internal initial parameters of the BP neural network, optimizing a training process through an Adam algorithm in a network training process, comparing prediction and real data at each moment, calculating a loss function value in a training set, and gradually reducing the loss function value through back propagation so as to realize convergence of characteristic parameters.
6. The diesel engine electronic speed regulation self-adaption method based on the artificial intelligence algorithm as claimed in claim 1, wherein in the step (3), the loss function in the forward propagation of the BP neural network is added with L2 regularization to prevent the over-fitting phenomenon from being generated, the traditional gradient descent method is replaced by an Adam algorithm in the backward propagation process, and the decision time is gradually reduced to the millisecond level through the deep self-learning by using a method of self-adaption for changing the learning rate.
7. The diesel engine electronic speed regulation self-adaption method based on the artificial intelligence algorithm as claimed in claim 1, wherein in the step (4), the specific content and method for reconstructing and outputting the characteristic parameters through the optimized data obtained by the BP neural network in the step (3) is that a multi-characteristic self-adaption learning method based on the artificial intelligence algorithm (ALO-GA algorithm) is used for carrying out self-adaption tracking on the characteristic parameters; the characteristic parameter searching range is narrowed through an ant lion algorithm, namely an ALO algorithm, then an adaptive genetic algorithm, namely a GA algorithm is adopted, and the self-adaptive function of the cross rate and the variation rate according to the data convergence characteristics is added, so that the convergence speed is increased, and the prediction precision is improved.
8. The diesel engine electronic speed regulation self-adaption method based on the artificial intelligence algorithm as claimed in claim 1, characterized in that in the step (4), a statistical analysis method is utilized to carry out data analysis on the prediction result output by the artificial intelligence algorithm, the test set data is utilized to compare the predicted value of the characteristic parameter with the true value, an absolute error value is calculated, and finally the characteristics of low delay and high precision are utilized to carry out the diesel engine rotating speed self-adaption control.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111340944.XA CN114037075B (en) | 2021-11-12 | 2021-11-12 | Diesel engine electronic speed regulation self-adaption method based on artificial intelligence algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111340944.XA CN114037075B (en) | 2021-11-12 | 2021-11-12 | Diesel engine electronic speed regulation self-adaption method based on artificial intelligence algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114037075A true CN114037075A (en) | 2022-02-11 |
| CN114037075B CN114037075B (en) | 2024-10-18 |
Family
ID=80137440
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111340944.XA Active CN114037075B (en) | 2021-11-12 | 2021-11-12 | Diesel engine electronic speed regulation self-adaption method based on artificial intelligence algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114037075B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115081625A (en) * | 2022-07-21 | 2022-09-20 | 常安集团有限公司 | Intelligent control method and system for miniature circuit breaker |
| CN116630398A (en) * | 2023-07-21 | 2023-08-22 | 浙江大学海南研究院 | Optimizer momentum coefficient regulation and control method based on data set concave-convex characteristic |
| CN116663396A (en) * | 2023-04-29 | 2023-08-29 | 中国人民解放军63963部队 | Power generation compatible type diesel engine displacement artificial intelligence decision-making method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107919813A (en) * | 2017-11-20 | 2018-04-17 | 无锡开放大学 | Ultrasound electric machine rotating speed control based on fuzzy neural network |
| EA201650033A1 (en) * | 2016-10-21 | 2018-04-30 | Акционерное общество "Ижевский радиозавод" | CONTROL STATION FOR ROD OF DEPTH PUMP |
| CN111351668A (en) * | 2020-01-14 | 2020-06-30 | 江苏科技大学 | Diesel engine fault diagnosis method based on optimized particle swarm algorithm and neural network |
| US20210201155A1 (en) * | 2019-12-30 | 2021-07-01 | Dalian University Of Technology | Intelligent control method for dynamic neural network-based variable cycle engine |
-
2021
- 2021-11-12 CN CN202111340944.XA patent/CN114037075B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EA201650033A1 (en) * | 2016-10-21 | 2018-04-30 | Акционерное общество "Ижевский радиозавод" | CONTROL STATION FOR ROD OF DEPTH PUMP |
| CN107919813A (en) * | 2017-11-20 | 2018-04-17 | 无锡开放大学 | Ultrasound electric machine rotating speed control based on fuzzy neural network |
| US20210201155A1 (en) * | 2019-12-30 | 2021-07-01 | Dalian University Of Technology | Intelligent control method for dynamic neural network-based variable cycle engine |
| CN111351668A (en) * | 2020-01-14 | 2020-06-30 | 江苏科技大学 | Diesel engine fault diagnosis method based on optimized particle swarm algorithm and neural network |
Non-Patent Citations (1)
| Title |
|---|
| 郭继通;郑方志;徐成宇;朱永伟;: "基于遗传算法和神经网络的软脆工件研磨加工工艺智能决策系统", 表面技术, no. 04, 20 April 2020 (2020-04-20) * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115081625A (en) * | 2022-07-21 | 2022-09-20 | 常安集团有限公司 | Intelligent control method and system for miniature circuit breaker |
| CN116663396A (en) * | 2023-04-29 | 2023-08-29 | 中国人民解放军63963部队 | Power generation compatible type diesel engine displacement artificial intelligence decision-making method and system |
| CN116663396B (en) * | 2023-04-29 | 2023-11-03 | 中国人民解放军63963部队 | Power generation compatible type diesel engine displacement artificial intelligence decision-making method and system |
| CN116630398A (en) * | 2023-07-21 | 2023-08-22 | 浙江大学海南研究院 | Optimizer momentum coefficient regulation and control method based on data set concave-convex characteristic |
| CN116630398B (en) * | 2023-07-21 | 2024-05-10 | 浙江大学海南研究院 | Optimizer momentum coefficient regulation and control method based on data set concave-convex characteristic |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114037075B (en) | 2024-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114037075B (en) | Diesel engine electronic speed regulation self-adaption method based on artificial intelligence algorithm | |
| CN111222677A (en) | Wind speed prediction method and system based on long-short term memory time neural network | |
| CN106600059B (en) | Intelligent power grid short-term load prediction method based on improved RBF neural network | |
| Song et al. | Coordinated optimization on energy capture and torque fluctuation of wind turbines via variable weight NMPC with fuzzy regulator | |
| CN111144552B (en) | Multi-index grain quality prediction method and device | |
| CN111260124A (en) | Chaos time sequence prediction method based on attention mechanism deep learning | |
| CN113972667A (en) | Power distribution network reactive power optimization method based on one-dimensional convolutional neural network | |
| CN114895206B (en) | Lithium ion battery SOH estimation method based on RBF neural network of improved gray wolf optimization algorithm | |
| CN117439069A (en) | Electric quantity prediction method based on neural network | |
| CN112257847A (en) | Method for predicting geomagnetic Kp index based on CNN and LSTM | |
| CN116646929A (en) | PSO-CNN-BILSTM-based short-term wind power prediction method | |
| CN109408896B (en) | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production | |
| Jiang et al. | Super short-term wind speed prediction based on CEEMD decomposition and BILSTM-Transformer model | |
| CN116542882B (en) | Photovoltaic power generation smoothing method, system and storage medium | |
| CN111027612B (en) | Energy metering data feature reduction method and device based on weighted entropy FCM | |
| Ghofrani et al. | K-means clustering with a new initialization approach for wind power forecasting | |
| CN116108975A (en) | Method for establishing short-term load prediction model of power distribution network based on BR-SOM clustering algorithm | |
| CN115759343A (en) | E-LSTM-based user electric quantity prediction method and device | |
| CN112508239A (en) | Energy storage output prediction method based on VAE-CGAN | |
| CN114139442A (en) | Method for optimizing steam turbine valve flow characteristic function based on K-means-LSTM | |
| CN113935457A (en) | Pulse neural network input signal coding method based on normal distribution | |
| Gao | New evolutionary neural networks | |
| CN114219069B (en) | Brain effect connection network learning method based on automatic variation self-encoder | |
| CN109829237B (en) | Wireless channel attenuation model fitting method based on quantum sea squirt group | |
| CN116954086B (en) | Intelligent prediction control method and device for adjusting system of pumped storage unit |
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 | ||
| GR01 | Patent grant |