Engineering machinery oil consumption prediction method based on deep learning
The construction of a fuel consumption prediction model for engineering machinery through deep learning methods has solved the problem of inaccurate fuel consumption prediction in the existing technology, achieved accurate prediction of fuel consumption, reduced R&D costs, improved prediction accuracy, and guided the work of engineering machinery.
Patent Information
- Application Number
- CN202510403844.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-01
- Publication Date
- 2025-07-04
AI Technical Summary
In the prior art, there is a lack of effective methods for predicting fuel consumption of construction machinery, which leads to consistency problems and increased R&D costs, and the prediction results are inaccurate, making it impossible to effectively guide the work of construction machinery.
Using a deep learning-based method, the excavator's fuel consumption-related feature data is collected, standardized processing and principal component analysis is performed, and the support vector machine is used to establish a functional relationship between feature data and fuel consumption, a fuel consumption prediction model is constructed, and the loss function is optimized through deep learning to improve prediction accuracy.
It realizes accurate prediction of fuel consumption of construction machinery, reduces R&D costs, improves the accuracy of fuel consumption prediction, and can better guide the work of construction machinery.
Smart Images
Figure CN120256938A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of excavators, and in particular to a fuel consumption prediction method for construction machinery based on deep learning. Background Art
[0002] The statements in this part only provide background technical information related to the present invention and do not necessarily constitute prior art.
[0003] Construction machinery such as excavators has relatively high requirements for fuel consumption. In the prior art for construction machinery, usually a prototype is used to match the fuel consumption of the whole vehicle. After the matching, the matching results are mass-produced. However, due to product consistency problems, there are often deviations in fuel consumption. When the deviation results are large, it is necessary to re-match the fuel consumption of the whole vehicle and re-develop and design, which undoubtedly increases the R & D cost. If the fuel consumption of construction machinery can be predicted during the working process, the R & D cost can be reduced. There are many factors affecting the fuel consumption of construction machinery. For example, factors such as the bucket of construction machinery, the current during the rise / fall of the boom, and the working time length are all relevant. There is no reasonable solution in the prior art to effectively predict the fuel consumption of construction machinery, or the accuracy of the prediction results cannot be guaranteed, so it is impossible to better guide the work of construction machinery such as excavators. Summary of the Invention
[0004] Aiming at the deficiencies of the prior art, the purpose of the present invention is to provide a fuel consumption prediction method for construction machinery based on deep learning, which reduces the test and development costs and better guides the work of construction machinery.
[0005] In order to achieve the above purpose, the present invention is implemented through the following technical solutions:
[0006] A fuel consumption prediction method for construction machinery based on deep learning includes the following:
[0007] Collect raw data and extract multiple feature data related to the fuel consumption of the excavator;
[0008] Preprocess the extracted multiple feature data to obtain standardized data;
[0009] Perform principal component analysis on the standardized data to determine that the number of principal components in the feature data is K and obtain training data;
[0010] Use a support vector machine to establish a functional relationship between K feature data and fuel consumption through the training data, and establish a fuel consumption prediction model through deep learning;
[0011] Evaluate the accuracy of the fuel consumption prediction model. After passing the evaluation, it indicates that the established fuel consumption prediction model is feasible.
[0012] A fuel consumption prediction method for construction machinery based on deep learning as described above standardizes multiple extracted feature data to obtain the standardized data, which is conducive to subsequent principal component analysis.
[0013] A fuel consumption prediction method for construction machinery based on deep learning as described above performs principal component analysis on the standardized data to replace the original relatively large number of variables with fewer variables while retaining as much information as possible from the original multiple variables, including the following:
[0014] Calculate the covariance matrix of the standardized data, and calculate the eigenvalues and eigenvectors of the covariance matrix;
[0015] Calculate the cumulative variance contribution rate of the feature data based on the eigenvalues of the covariance matrix. When the variance contribution rate of the first K feature data exceeds the set ratio, the first K feature data are used as the feature data set after principal component analysis.
[0016] A fuel consumption prediction method for construction machinery based on deep learning as described above obtains the training data through the matrix composed of the eigenvectors of the first K feature data and the standardized data to ensure the accuracy of the fuel consumption prediction result.
[0017] A fuel consumption prediction method for construction machinery based on deep learning as described above aims to fit the hyperplane of the data to the greatest extent and ensure that the error is less than the set threshold. The functional relationship is a regression functional relationship. A regression functional relationship between the K feature data and the fuel consumption is established through the weight vector, the training data, and the bias term.
[0018] A fuel consumption prediction method for construction machinery based on deep learning as described above, after establishing the regression functional relationship between the K feature data and the fuel consumption, establishes an insensitive loss function through the regression function, the actual fuel consumption value, the predicted fuel consumption value, and the tolerated error threshold.
[0019] A fuel consumption prediction method for construction machinery based on deep learning as described above, on the basis of obtaining the training data, in order to achieve deep learning, uses the training data and the corresponding actual fuel consumption value, inputs them into the insensitive loss function to optimize the insensitive loss function, obtains the optimal weight vector, aims to minimize the objective function, obtains the optimal bias term, and establishes an initial fuel consumption prediction model through the optimal weight vector, the optimal bias term, and the training data;
[0020] Divide the original data into an 80% training set and a 20% test set, and bring the test set into the initial fuel consumption prediction model to obtain the fuel consumption prediction model.
[0021] As described above, in the method for predicting fuel consumption of construction machinery based on deep learning, the fuel consumption prediction model is a relationship established by the fuel consumption prediction value and the weight vector, bias term and test set.
[0022] As described above, a method for predicting fuel consumption of construction machinery based on deep learning obtains predicted fuel consumption value through the fuel consumption prediction model, and the accuracy of the established fuel consumption prediction model is evaluated through mean square error, root mean square error and determination coefficient according to the average value of the actual fuel consumption value, the predicted fuel consumption value and the actual fuel consumption value.
[0023] A method for predicting fuel consumption of construction machinery based on deep learning as described above, the raw data includes but is not limited to the pump pressure of the first pump, the pump pressure of the second pump, the control current of the first pump, the control current of the second pump, the hydraulic oil temperature, the actual engine speed, the engine torque, the real-time fuel injection amount, the first boom raising solenoid valve current, the boom merging solenoid valve current, the first boom lowering solenoid valve current, the boom regeneration valve current, the boom quick oil return current, the first dipper arm digging current, the second dipper arm digging current, the first dipper arm unloading current, the second dipper arm unloading current, the first bucket digging current, the first bucket unloading current, the left swing current, the right swing current, the straight travel valve current, the boom raising pilot pressure, and the boom lowering pilot pressure.
[0024] The beneficial effects of the present invention are as follows:
[0025] 1) The present invention first extracts multiple feature data from the original data, eliminates some irrelevant original data, reduces the feature data set, performs principal component analysis on the standardized data after standardization, so as to obtain the number of principal component feature data, so as to further reduce the feature data, but the impact on the prefabricated results will not be too large, and obtains training data, uses the training data to establish a functional relationship between K feature data and fuel consumption, and establishes a fuel consumption prediction model through deep learning, so as to ensure the accuracy of the prediction results of the fuel consumption prediction model, which is closer to the actual value, and guide the action of the engineering machinery through more accurate prediction values.
[0026] 2) In the present invention, feature data is extracted from the original data, and the feature data is standardized to facilitate principal component analysis. The covariance matrix and cumulative variance contribution rate of the standardized data are used to select appropriate K feature data as principal components, thereby obtaining training data. Through principal component analysis, fewer variables are used to replace the original more variables, while retaining most of the information of the original multiple variables as much as possible.
[0027] 3) The functional relationship between the K characteristic data and the fuel consumption in the present invention is a regression functional relationship. The hyperplane that fits the data to the greatest extent is utilized by the support vector machine to ensure that the error is less than the set threshold, and the regression functional relationship between the K characteristic data and the fuel consumption is reasonably established through the weight vector, the training data, and the bias term.
[0028] 4) The present invention uses the training data and the actual fuel consumption values to optimize the loss function, obtain the optimal weight vector and bias term, divide the original data into an 80% training set and a 20% test set, bring the test set into the insensitive loss function, and obtain the fuel consumption prediction model. In this way, the fuel consumption prediction model is established through deep learning, effectively ensuring the accuracy and reliability of the fuel consumption prediction results. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] The accompanying drawings forming a part of this specification are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention.
[0030] Figure 1 is a flowchart of a construction machinery fuel consumption prediction method based on deep learning according to one or more embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0031] It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.
[0032] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the present invention otherwise clearly indicates, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof;
[0033] As introduced in the background art, there is a problem that there is no suitable method for predicting the fuel consumption of construction machinery in the prior art. To solve the above technical problems, the present invention proposes a construction machinery fuel consumption prediction method based on deep learning.
[0034] In a typical embodiment of the present invention, referring to Figure 1 as shown, a construction machinery fuel consumption prediction method based on deep learning includes the following contents:
[0035] 1) Collect the original data and extract and mine multiple characteristic data related to the fuel consumption of the excavator;
[0036] There are 141 items of original data for the excavator, and 51 items of characteristic data related to the fuel consumption of the excavator are extracted from them. Specifically, they include the pump pressure X1 of the first pump, the pump pressure X2 of the second pump, the control current X3 of the first pump, the control current X4 of the second pump, the hydraulic oil temperature X5, the actual engine speed X6, the engine torque X7, the real-time fuel injection volume X8, the current of the first boom lift solenoid valve X9, the current of the boom confluence solenoid valve X10, the current of the first boom lower solenoid valve X11, the current of the boom regeneration valve X12, the current of the boom quick return oil X13, the current of the first stick excavation X14, the current of the second stick excavation X15, the current of the first stick unloading X16, the current of the second stick unloading X17, the current of the first bucket excavation X18, the current of the first bucket discharging X19, the left rotation current X20, the right rotation current X21, the current of the straight travel valve X22, the pilot pressure of boom lift X23, the pilot pressure of boom lower X24, the pressure of the large chamber of the boom X25, the pressure of the small chamber of the boom X26, the pilot pressure of stick excavation X27, the pilot pressure of stick discharging X28, the pilot pressure of bucket excavation X29, the pilot pressure of bucket discharging X30, the pilot pressure of rotation X31, the pilot pressure of left travel X32, the pilot pressure of right travel X33, the working gear X34, the working time X35, the working mode feedback X36, the cumulative fuel consumption X37, the cooling water temperature X38, the oil pressure X39, the atmospheric pressure X40, the supercharging pressure X41, the supercharging temperature X42, the intake air temperature X43, the common rail pressure X44, the power set value X45, the actual power value X46, the working time of the day X47, the traveling time of the day X48, the air conditioner on / off state X49, the set temperature of the air conditioner X50, and the set air volume of the air conditioner X51.
[0037] It is easily understandable that in other examples, fewer or more characteristic data can also be selected.
[0038] 2) Standardize the extracted multiple characteristic data to obtain the standardized data, which is conducive to subsequent principal component analysis:
[0039] Form the original matrix X = [X1, X2, X3,..., X 51 with 51 characteristic data. By standardizing each characteristic data, the standardized data X P is obtained: where μ is the mean of multiple characteristic data, and σ is the standard deviation of multiple characteristic data. Ensure that the mean of each characteristic is 0 and the standard deviation is 1;
[0040] 3) Perform principal component analysis on the standardized data to replace the original relatively large number of variables with a smaller number of variables, while retaining as much information as possible from the original multiple variables, and determine the number of principal components. Here, determining the number of principal components refers to determining the number of principal components in the characteristic data:
[0041] Perform principal component analysis (PCA) on the standardized data X P to calculate the covariance matrix C of the standardized data:
[0042]
[0043] Calculate the eigenvalues λ i (i = 1, 2, 3....., 51) and eigenvectors. Each element of the eigenvector is the eigenvector coefficient. The eigenvector defines the principal component direction of the characteristic data, and the eigenvalue represents the variance size represented by each principal component;
[0044] Calculate the cumulative variance contribution rate of the characteristic data based on the eigenvalues of the covariance matrix:
[0045] Cumulative variance contribution rate ((j = 1, 2, 3....., 51));
[0046] Through calculation, it is known that the cumulative variance contribution rate of the first K characteristic data exceeds a set ratio such as 85%. Select the first K characteristic data as the reduced-dimensional characteristic data set. To ensure the accuracy of the fuel consumption prediction result, obtain the training data X through the matrix composed of the eigenvectors of the first K characteristic data and the standardized data pca :
[0047] X pca = X p V K
[0048] where V k is the matrix composed of the eigenvectors of the first K characteristic data. Through principal component analysis, the original 51 characteristic data are converted into K, and K is less than 51;
[0049] Determine the eigenvalue X of the test set through the characteristic data and the cumulative variance contribution rate of the characteristic data 测试集 :
[0050] X 测试集 = a i1 y1 + a i2 y2 + a i3 y3 + …… + a i51 y 51 (i = 1, 2, 3…, k);
[0051] where aij (j = 1, 2, 3....., 51) are the eigenvector coefficients (also called loadings or weights), which refer to the contribution degrees of the original variables in each principal component in principal component analysis (PCA). Specifically, these coefficients are the elements in the eigenvectors of the principal components, from which the importance of each original variable in the principal components can be known.
[0052] 4) Use the support vector machine to establish the functional relationship between K feature data and fuel consumption through training data, and establish a fuel consumption prediction model by means of deep learning:
[0053] Use the support vector machine (SVR) to construct the functional relationship between K feature data and fuel consumption. The goal of using the support vector machine is to find a hyperplane that can best fit the data and ensure that the error is less than the set threshold, and establish the regression function relationship between K feature data and fuel consumption through the weight vector, training data and bias term. For the training data X pca , the regression function of the support vector machine is:
[0054] f(X pca ) = <ω, X pca > + b
[0055] where ω is the weight vector, representing the parameters of the model, X pca is the training data, and b is the bias term.
[0056] It should be explained that the goal of using the support vector machine is to minimize the loss function and make the model as simple as possible. An insensitive loss function is established through the regression function, the actual fuel consumption value, the predicted fuel consumption value and the tolerated error threshold. The support vector machine adopts the insensitive loss function, and its form is:
[0057]
[0058] where y is the actual fuel consumption, f(X pca ) is the predicted fuel consumption value, and ε is the tolerated error threshold.
[0059] Furthermore, a regularization function is introduced to prevent the occurrence of overfitting, and the following objective function is minimized:
[0060]
[0061] where ||ω|| 2 is a measure of the complexity of the model, ξ i is the slack variable, representing the allowed error, C is the regularization function, which adjusts the balance between complexity and error.
[0062] In this embodiment, on the basis of obtaining the training data, to achieve deep learning, through the training data Xpca and the corresponding actual fuel consumption value y are input into the insensitive loss function to optimize the insensitive loss function, obtain the optimal weight vector ω, aim to minimize the objective function, obtain the optimal bias term b, and input the obtained optimal weight vector ω and optimal bias term b into the regression function to obtain the initial fuel consumption prediction model.
[0063] In some examples, the original data is divided into an 80% training set and a 20% test set, and the test set is brought into the initial fuel consumption prediction model to obtain the fuel consumption prediction value as the fuel consumption prediction model:
[0064]
[0065] 5) Evaluate the accuracy of the fuel consumption prediction model. After passing the evaluation, it indicates that the established fuel consumption prediction model is feasible:
[0066] In this embodiment, according to the true fuel consumption value, the predicted fuel consumption value, and the average value of the true fuel consumption value, the mean square error (MSE), root mean square error (RMSE), and coefficient of determination R 2 are used to evaluate the model accuracy, and the evaluation is carried out through the following formula:
[0067]
[0068] where y i is the true fuel consumption value, is the predicted fuel consumption value, is the average value of the true fuel consumption value;
[0069] Specifically, the true fuel consumption value, the predicted fuel consumption value, and the average value of the true fuel consumption value are input into the above formula. The true fuel consumption value and the average value of the true fuel consumption value are both obtained through the construction machinery control center to determine the mean square error (MSE), root mean square error (RMSE), and coefficient of determination R 2 to check whether the requirements are met. If the requirements are met, the established fuel consumption prediction model is feasible. By inputting relevant data through the fuel consumption prediction model, the fuel consumption prediction result can be obtained, and accordingly, the fuel consumption prediction curve can be drawn to achieve fuel consumption prediction.
[0070] The fuel consumption prediction method provided in this embodiment first extracts multiple feature data from the original data, eliminates some irrelevant original data to narrow the feature data set, performs principal component analysis on the standardized data after standardization to obtain the number of principal component feature data, further narrow the feature data, but the impact on the prefabricated result will not be too large, and obtain training data. Use the training data to establish the functional relationship between K feature data and fuel consumption, and establish a fuel consumption prediction model through deep learning, so as to ensure the accuracy of the prediction result of the fuel consumption prediction model, be closer to the actual value, and guide the actions of construction machinery through relatively accurate prediction values.
[0071] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A fuel consumption prediction method for construction machinery based on deep learning, characterized in that, It includes the following contents: Collect the original data and extract multiple feature data related to the fuel consumption of the excavator; Preprocess the extracted multiple feature data to obtain the standardized data; Perform principal component analysis on the standardized data to determine that the number of principal components in the feature data is K, and obtain the training data; Use the support vector machine to establish the functional relationship between K feature data and the fuel consumption through the training data, and establish the fuel consumption prediction model through the deep learning method; Evaluate the accuracy of the fuel consumption prediction model. After passing the evaluation, it indicates that the established fuel consumption prediction model is feasible.
2. The method for predicting fuel consumption of construction machinery based on deep learning according to claim 1, wherein Perform standardization processing on the extracted multiple feature data to obtain the above-mentioned standardized data.
3. The fuel consumption prediction method for construction machinery based on deep learning according to claim 1, characterized in that The principal component analysis of the standardized data includes the following contents: Calculate the covariance matrix of the standardized data, and calculate the eigenvalues and eigenvectors of the covariance matrix; Calculate the cumulative variance contribution rate of the feature data according to the eigenvalues of the covariance matrix. When the variance contribution rate of the first K feature data exceeds the set ratio, use the first K feature data as the feature data set after principal component analysis.
4. The method for predicting fuel consumption of construction machinery based on deep learning according to claim 3, characterized in that, Obtain the above-mentioned training data through the matrix composed of the eigenvectors of the first K feature data and the standardized data.
5. A fuel consumption prediction method for construction machinery based on deep learning according to claim 1, characterized in that The functional relationship is a regression functional relationship, and establish the regression functional relationship between K feature data and the fuel consumption through the weight vector, the above-mentioned training data and the bias term.
6. The method for predicting fuel consumption of construction machinery based on deep learning according to claim 5, wherein, After establishing the regression function between K feature data and the fuel consumption, establish an insensitive loss function through the regression function, the actual fuel consumption value, the predicted fuel consumption value and the tolerable error threshold.
7. A method for predicting fuel consumption of construction machinery based on deep learning according to claim 6, characterized in that, Use the above-mentioned training data and the corresponding actual fuel consumption value, input them into the insensitive loss function to optimize the insensitive loss function, obtain the optimal weight vector, aim to minimize the objective function, obtain the optimal bias term, and establish the initial fuel consumption prediction model through the optimal weight vector, the optimal bias term and the above-mentioned training data; Divide the above-mentioned original data into an 80% training set and a 20% test set, bring the test set into the initial fuel consumption prediction model, and obtain the above-mentioned fuel consumption prediction model.
8. A method for predicting fuel consumption of construction machinery based on deep learning according to claim 7, characterized in that, The fuel consumption prediction model is the relational expression established through the fuel consumption prediction value, the weight vector, the bias term and the test set.
9. A fuel consumption prediction method for construction machinery based on deep learning according to claim 1, characterized in that, Obtain the predicted fuel consumption value through the fuel consumption prediction model, and evaluate the accuracy of the established fuel consumption prediction model through the mean square error, the root mean square error and the coefficient of determination according to the true fuel consumption value, the predicted fuel consumption value and the average value of the true fuel consumption value.
10. A method for predicting the fuel consumption of construction machinery based on deep learning according to claim 1, characterized in that, The original data includes but is not limited to the pump pressure of the first pump, the pump pressure of the second pump, the control current of the first pump, the control current of the second pump, the hydraulic oil temperature, the actual engine speed, the engine torque, the real-time fuel injection volume, the current of the first boom rising solenoid valve, the current of the boom confluence solenoid valve, the current of the first boom lowering solenoid valve, the current of the boom regeneration valve, the current of the boom quick return oil, the current of the first stick digging, the current of the second stick digging, the current of the first stick unloading, the current of the second stick unloading, the current of the first bucket digging, the current of the first bucket discharging, the left swing current, the right swing current, the straight travel valve current, the pilot pressure of the boom rising, the pilot pressure of the boom lowering.