Fuel consumption analysis method based on ship driving conditions

Abstract

The present invention discloses a fuel consumption analysis method based on ship driving conditions, which relates to the technical field of ship energy management, including: collecting ship driving sensor data, pre-processing the data, and selecting features related to fuel consumption as key features; inputting the key features into a fuel consumption analysis model established based on random forest rules for learning and training, and then performing parameter tuning through cross-validation; and then inputting the real-time collected ship driving fuel consumption features into the optimized trained fuel consumption analysis model for fuel consumption analysis, and outputting the analysis results. The present invention can analyze the fuel consumption of a ship in real time by inputting the real-time collected ship driving data into the optimized fuel consumption analysis model, thereby improving the accuracy and practicality of the fuel consumption analysis and greatly enhancing the efficiency and economy of ship fuel consumption management.

Description

Technical Field

[0001] The present invention relates to the technical field of ship energy management, and in particular to a fuel consumption analysis method based on ship driving conditions. Background Art

[0002] With the rapid development of the global shipping industry, the energy efficiency and fuel consumption management of ships have become the focus of the industry. In recent years, traditional fuel consumption analysis methods have mostly relied on empirical estimates or simple mechanical measurements. These methods often cannot reflect the changes in fuel consumption under complex sea conditions in real time and have limited accuracy. In addition, due to the lack of highly integrated data processing technology, these traditional methods cannot effectively process and analyze large amounts of sensor data, thus affecting the formulation and implementation of fuel consumption optimization strategies.

[0003] Although some automated tools and software have been introduced into existing fuel consumption analysis technologies to monitor the dynamic parameters of ships, such as speed, load and heading, these technologies still face some significant limitations. First, they usually cannot fully consider environmental factors such as ocean currents, wind speed and their combined impact on fuel consumption. Second, existing technologies are insufficient in data processing and feature extraction, and fail to fully utilize the potential of machine learning technology in pattern recognition and prediction.

[0004] In addition, most existing solutions lack effective real-time data processing capabilities and cannot provide ship managers with immediate fuel consumption adjustment recommendations. These shortcomings still make ship operators face many challenges in fuel consumption management and decision-making. Summary of the invention

[0005] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract of the specification and the title of the invention of this application to avoid blurring the purpose of this section, the abstract of the specification and the title of the invention, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the technical problem solved by the present invention is that the existing ship fuel consumption analysis technology is insufficient in terms of data processing capability, real-time performance and comprehensive consideration of environmental factors.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: collecting ship driving sensor data, performing data preprocessing on the data, and selecting features related to fuel consumption as key features;

[0009] The key features are input into a fuel consumption analysis model established based on random forest rules for learning and training, and then parameters are tuned through cross-validation;

[0010] The fuel consumption characteristics of the ship collected in real time are then input into the optimized and trained fuel consumption analysis model to perform fuel consumption analysis, and the analysis results are output.

[0011] As a preferred solution of the fuel consumption analysis method based on ship driving conditions described in the present invention, the collected ship driving sensor data includes but is not limited to the ship's speed, position, fuel consumption, load, engine power, engine speed, wind speed, wind direction, ocean current, temperature and humidity.

[0012] As a preferred solution of the fuel consumption analysis method based on ship driving conditions described in the present invention, numerical data in a unified format are formed by performing data cleaning and standardization processing on the collected ship driving sensor data.

[0013] As a preferred solution of the fuel consumption analysis method based on ship driving conditions described in the present invention, the correlation between data features and fuel consumption is analyzed by using the Pearson correlation coefficient, so that data features with a large correlation with the fuel consumption are selected as the key features.

[0014] As a preferred solution of the fuel consumption analysis method based on ship driving conditions described in the present invention, the correlation between the Pearson correlation coefficient analysis data characteristics and fuel consumption can be obtained by calculating the correlation coefficient, and its mathematical expression formula is as follows:

[0015]

[0016] Among them, R f represents the correlation coefficient between data feature f and fuel consumption c, v i represents the data characteristics of the i-th observation, and They represent the sample means of the feature and fuel consumption c respectively, n represents the sample size, and the sample is the preprocessed ship driving sensor data. α is the threshold and β is the slope, which is used to control the sensitivity of feature selection.

[0017] As a preferred solution of the fuel consumption analysis method based on ship driving conditions of the present invention, the value range of the correlation coefficient is [-1, 1], where:

[0018] When the value is close to 1, it indicates a positive correlation, that is, the trend of change is the same, the stronger the correlation;

[0019] When the value is close to 0, it means there is no correlation;

[0020] When the value is close to -1, it indicates a negative correlation, that is, the trend of change is opposite and the correlation is weaker.

[0021] As a preferred solution of the fuel consumption analysis method based on ship driving conditions described in the present invention, parameter tuning is performed through cross-validation, including but not limited to the number of decision trees, the depth of the tree, and the minimum number of samples required for splitting nodes, so as to find the optimal model parameter configuration.

[0022] As a preferred solution of the fuel consumption analysis method based on ship driving conditions of the present invention, the analysis results include:

[0023] When the output value of the fuel consumption analysis model is close to 1, it indicates that the fuel consumption rate of the ship is high during the driving process;

[0024] When the output value of the fuel consumption analysis model is close to 0, it indicates that the fuel consumption rate of the ship is low during the driving process.

[0025] The beneficial effects of the present invention are as follows: the present invention collects ship driving sensor data, performs data preprocessing on the data, accurately selects key features directly related to fuel consumption, and realizes dynamic learning and real-time prediction of fuel consumption by applying a machine learning model based on random forest rules; in addition, the accuracy and robustness of the model under changeable sea conditions are further improved by using cross-validation technology to tune parameters, and the fuel consumption of the ship can be analyzed in real time by inputting the real-time collected ship driving data into the optimized fuel consumption analysis model, thereby improving the accuracy and practicability of the fuel consumption analysis and greatly enhancing the efficiency and economy of ship fuel consumption management. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0027] Figure 1 It is a flow chart of the fuel consumption analysis method based on the ship driving conditions shown in the present invention. DETAILED DESCRIPTION

[0028] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0029] Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without making any creative work should fall within the scope of protection of the present invention.

[0030] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0031] Example 1

[0032] According to an embodiment of the present invention, Figure 1 The flowchart shown in the figure is a fuel consumption analysis method based on the ship driving conditions, which specifically includes the following steps:

[0033] S1. Collect ship driving sensor data, perform data preprocessing on it, and select features related to fuel consumption as key features. It should be noted that in this step, the correlation between data features and fuel consumption is analyzed by the Pearson correlation coefficient, so that data features with a large correlation with fuel consumption are selected as key features;

[0034] Specifically, the correlation between the Pearson correlation coefficient analysis data characteristics and fuel consumption can be calculated by the correlation coefficient, and its mathematical expression formula is as follows:

[0035]

[0036] Among them, R f represents the correlation coefficient between data feature f and fuel consumption c, v i represents the data characteristics of the i-th observation, and They represent the sample means of the feature and fuel consumption c respectively, n represents the sample size, and the sample is the preprocessed ship driving sensor data. α is the threshold and β is the slope, which is used to control the sensitivity of feature selection.

[0037] Furthermore, the range of the correlation coefficient is [-1,1], where:

[0038] When the value is close to 1, it indicates a positive correlation, that is, the trend of change is the same, the stronger the correlation;

[0039] When the value is close to 0, it means there is no correlation;

[0040] When the value is close to -1, it indicates a negative correlation, that is, the trend of change is opposite and the correlation is weaker.

[0041] As an example, according to the calculated absolute value of the Pearson correlation coefficient, data features with a high correlation with fuel consumption are selected as key features. For example, if the absolute value of the correlation coefficient is greater than a preset threshold, the threshold can be set to 0.5.

[0042] It should also be noted that in this step, by normalizing the conditional function Not only the statistical correlation between the features and fuel consumption is calculated, but also the sensitivity and selection threshold of the correlation score are adjusted, that is, the correlation coefficient undergoes a nonlinear transformation, which enhances the influence of the intermediate value, making the selection of correlation more flexible and adaptable.

[0043] As an example, the collected ship driving sensor data includes but is not limited to the ship's speed, position, fuel consumption, load, engine power, engine speed, wind speed, wind direction, ocean current, temperature and humidity.

[0044] Data preprocessing is to clean and standardize the collected ship driving sensor data to form numerical data in a unified format.

[0045] In an optional embodiment, a data integrity check algorithm is used to verify the continuity and timestamp of each received data point, including detecting whether the data is within a reasonable operating range, such as the ship speed should not exceed the design maximum speed, and checking the integrity of the data, eliminating or marking missing or obviously erroneous data to confirm that the data has not been lost or erroneously repeated during transmission.

[0046] In an optional embodiment, the data from each sensor is converted into a unified measurement unit and numerical range, such as converting the speed from knots to meters per second and converting the temperature from Fahrenheit to Celsius, and then all numerical data are normalized so that their range is between 0 and 1 to eliminate the impact of dimension and facilitate subsequent analysis.

[0047] As an example, all sensor data can be synchronized according to timestamps to ensure that the data is recorded at the same or close time points, which facilitates timing analysis and feature association.

[0048] As an example, outliers can be identified and processed through statistical analysis methods, such as using the box plot method to identify and eliminate data points with values ​​outside the normal range, to ensure the quality and reliability of the data.

[0049] The data integrity check algorithm, normalization processing, statistical analysis method and timestamp synchronization method of the aforementioned data preprocessing can be performed using methods and means in the prior art and will not be described in detail in this example.

[0050] Preferably, this step ensures the quality and availability of the data by accurately collecting the ship's driving sensor data and preprocessing the data, and screens out key features directly related to fuel consumption, such as speed, load, and sea conditions, thereby improving the pertinence and efficiency of data processing.

[0051] Preferably, the processing of this step further ensures the accuracy and practicality of subsequent model training, so that the model can more accurately reflect the relationship between ship fuel consumption and operating parameters, and reduce the complexity of subsequent calculations, thereby improving the speed and response time of the overall analysis.

[0052] S2. Input the key features into the fuel consumption analysis model established based on the random forest rule for learning and training, and then perform parameter tuning through cross-validation.

[0053] In an optional implementation, this embodiment further enhances the generalization ability of the model and reduces overfitting by establishing multiple decision trees and introducing randomness in the training process for learning and training.

[0054] As an example, learning training includes but is not limited to generating different decision trees using different random seeds to ensure the stability and accuracy of the model.

[0055] In an optional implementation, this embodiment applies cross-validation technology to tune the parameters of the model, including but not limited to the number of decision trees, the depth of the tree, the minimum number of samples required to split a node, etc., to find the optimal model parameter configuration.

[0056] As an example, a K-fold cross-validation method can be used, where K is usually 5 or 10, to effectively evaluate the performance of the model on the predicted data and prevent the model from overfitting.

[0057] It is not difficult to understand that random forest, as an integrated learning method, conducts training and prediction by constructing multiple decision trees. Each tree is built independently. This structure itself greatly increases the model's ability to capture nonlinear relationships and complexities in the data. In the scenario of ship fuel consumption analysis, multiple factors such as ship speed, load, engine power, etc. have a direct or indirect impact on fuel consumption. Random forest can effectively handle the complex interactions between such features and performs better than traditional single decision trees or linear models.

[0058] Cross-validation can effectively prevent model overfitting and ensure that the model has good generalization ability. In the embodiment of the present invention, cross-validation is not only used to evaluate model performance, but also to optimize key parameters such as the number of decision trees and the depth of the tree, which ensures that the model can maintain high accuracy under various operating conditions.

[0059] By taking the average loss of cross-validation as the adjustment factor and using the Sigmoid function to smooth these loss values, not only the nonlinear characteristics of the model output are increased, but also the weight of each tree in the final prediction can be dynamically adjusted according to the results of cross-validation, thereby optimizing the prediction results; that is, the model can adjust its own prediction strategy according to the size of the loss function during the verification process, so that the output is closer to the fuel consumption in the actual operating environment.

[0060] Preferably, this step utilizes the high flexibility and powerful nonlinear learning ability of random forests, so that the model can achieve highly accurate predictive analysis in complex data environments. The random forest model improves the stability and accuracy of predictions by establishing multiple decision trees, and further enhances the generalization ability of the model by performing parameter tuning through cross-validation, ensuring that the performance of the model on unknown data is consistent with that on the training set, thereby achieving more reliable results in practical applications.

[0061] The learning, training and cross-validation methods of the aforementioned model can be performed using methods and means in the prior art and will not be described in detail in this example.

[0062] S3, input the real-time collected ship fuel consumption characteristics into the optimized and trained fuel consumption analysis model to perform fuel consumption analysis, and output the analysis results. It should be noted again in this step that the fuel consumption analysis model P can be expressed by the following mathematical expression formula:

[0063]

[0064] Among them, Y is the target variable, i.e., the fuel consumption of the ship, X is the matrix of input variables, including key features such as the ship’s speed, position, and fuel consumption, and T i is the i-th decision tree in the random forest, and its parameters are determined by , which are generated by a random process. N is the total number of decision trees. σ is the Sigmoid function, which is used to convert the results of cross-validation into values ​​in the interval [0,1], increase the nonlinear characteristics of the model output and adjust the smoothness of the analysis results. CV j is the mean square error of the j-th cross validation, is the analysis result in the jth cross-validation, and K is the number of cross-validation folds, such as 5 or 10.

[0065] Specifically, the analysis results include:

[0066] When the output value of the fuel consumption analysis model is close to 1, it means that the fuel consumption rate of the ship is high during the voyage;

[0067] When the output value of the fuel consumption analysis model is close to 0, it means that the fuel consumption rate of the ship is low during navigation.

[0068] It should be noted that the above formula integrates the outputs of multiple decision trees in the random forest, adjusts the weight of the model prediction through the average loss of cross-validation, and then introduces the average loss of cross-validation as an adjustment factor to adjust the contribution of each tree, thereby optimizing the model's ability to analyze data, thereby improving the accuracy of data analysis and the generalization ability of the model and preventing the occurrence of overfitting.

[0069] Among them, the Sigmoid function in the formula, as a smooth and nonlinear conversion tool, can effectively map the value of the loss function to an appropriate range, and then serve as the output of the weight adjustment model.

[0070] The above mapping of the value of the loss function into a suitable range is not limited to the range in the embodiment of the present invention and can be customized according to the actual application scenario.

[0071] It should also be noted that the ship driving fuel consumption characteristics collected in real time in this step are obtained by preprocessing and extracting the ship driving sensor data collected in real time by the sensor through step S1, which will not be described in detail in this example.

[0072] Preferably, this step inputs the real-time collected ship fuel consumption characteristics into the optimized and trained fuel consumption analysis model, performs fuel consumption analysis and outputs the analysis results, which can realize real-time fuel consumption monitoring and prediction, help ship operators understand the fuel consumption status in real time, and make timely adjustments based on the data provided by the model, such as adjusting the speed or changing the route, to optimize fuel consumption and ensure navigation safety.

[0073] Furthermore, the embodiment of the present invention selected five different types of commercial ships, including container ships, bulk carriers, tankers, ferries and cruise ships, covering the main operating types in the shipping industry. These ships are equipped with modern sensor systems that can provide real-time driving data, including key information such as speed, position, and fuel consumption. Sensor brands include ABB and Siemens. These devices can collect data at a high frequency (at least once per second) to ensure the accuracy and real-time nature of the data.

[0074] The experimental environment was set up on a regular commercial route to ensure the authenticity and repeatability of the data. During the experiment, each ship sailed according to its normal operating plan to ensure the consistency of the natural conditions of the experiment and commercial operations.

[0075] The data acquisition system transmits data in real time via a satellite link back to the experiment control center. It includes a GPS positioning system to ensure the accuracy of the position data, and an interface with the engine management system (EMS) to obtain data such as engine power and speed.

[0076] Before the experiment began, all relevant sensors and data transmission equipment were rigorously tested and calibrated to ensure accurate data entry. The sensor data of each ship was encoded and uploaded to the central database, followed by preliminary data cleaning and preprocessing using dedicated software, including removal of obvious outliers, data normalization, and time synchronization.

[0077] This experiment optimizes the generalization ability of the model by using a 5-fold cross-validation method. In each round of cross-validation, the dataset is divided into five parts, each part is used as a test set in turn, and the rest are used as training sets.

[0078] The random forest model was conducted on a high-performance computer using Python's Scikit-learn library. Before model training, the data set was randomly divided into a training set and a test set, 80% of which was used for training and 20% for testing. The results are shown in the following table:

[0079] Table 1. Fuel consumption analysis score table

[0080]

[0081] Referring to Table 1, it can be intuitively seen that the fuel consumption analysis model proposed in the embodiment of the present invention accurately predicts fuel consumption under different environmental and operating conditions. For example, when the wind speed is high or the ship load is large, the model can effectively predict higher fuel consumption, which proves that the model can understand and process a variety of factors affecting fuel consumption. Through real-time monitoring and analysis, ship operators can adjust the speed and load according to the model prediction, optimize fuel efficiency, and achieve the dual goals of cost savings and environmental protection.

[0082] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A fuel consumption analysis method based on ship driving conditions, characterized in that: include: Collect ship driving sensor data, perform data preprocessing, and select features related to fuel consumption as key features; The correlation between the data feature and the fuel consumption is analyzed by the Pearson correlation coefficient, so that the data feature with a large correlation with the fuel consumption is selected as the key feature; the correlation between the data feature and the fuel consumption analyzed by the Pearson correlation coefficient is calculated by the correlation coefficient, and its mathematical expression formula is as follows: Among them, R f represents the correlation coefficient between data feature f and fuel consumption c, represents the data characteristics of the i-th observation, and They represent the sample means of the characteristics and fuel consumption c, respectively, and n represents the sample size. The sample is the preprocessed ship driving sensor data. is the threshold value, is the slope, which is used to control the sensitivity of feature selection; The key features are input into a fuel consumption analysis model established based on random forest rules for learning and training, and then parameters are tuned through cross-validation; The parameter tuning includes but is not limited to the number of decision trees, the depth of the tree, the minimum number of samples required to split a node, and converting the cross-validation loss value into a weight output through a Sigmoid function; Then, the real-time collected ship fuel consumption characteristics are input into the optimized and trained fuel consumption analysis model to perform fuel consumption analysis, and the analysis results are output; the fuel consumption analysis model is represented by the following formula: Among them, Y is the target variable, i.e., the fuel consumption of the ship, X is the matrix of input variables, including the key features of the ship’s speed, position, and fuel consumption, T i is the i-th decision tree in the random forest, whose parameters are determined by , which are generated by a random process, N is the total number of decision trees, It is the Sigmoid function, which is used to convert the cross-validation results into values ​​in the interval [0,1], increase the nonlinear characteristics of the model output and adjust the smoothness of the analysis results. j is the mean square error of the j-th cross validation, is the analysis result in the jth cross validation, and K is the number of cross validation folds; The analysis results include: When the output value of the fuel consumption analysis model is 1, it indicates that the fuel consumption rate of the ship is high during the driving process; When the output value of the fuel consumption analysis model is 0, it indicates that the fuel consumption rate of the ship is low during the driving process.

2. The fuel consumption analysis method based on ship driving conditions according to claim 1 is characterized in that: The collected ship driving sensor data includes but is not limited to the ship's speed, position, fuel consumption, load, engine power, engine speed, wind speed, wind direction, ocean current, temperature and humidity.

3. The fuel consumption analysis method based on ship driving conditions according to claim 1 or 2, characterized in that: The collected ship driving sensor data is cleaned and standardized to form numerical data in a unified format.

4. The fuel consumption analysis method based on ship driving conditions according to claim 1 is characterized in that: The range of the correlation coefficient is [-1,1], where: When the value is 1, it indicates positive correlation, that is, the change trend is the same and the correlation is strong; When the value is 0, it means there is no correlation; When the value is -1, it indicates a negative correlation, that is, the change trend is opposite and the correlation is weak.

Images