CN116108995A - Fuel consumption prediction method, device and electronic equipment for ships in tidal reach - Google Patents

Abstract

The application discloses a tidal range ship oil consumption prediction method, a tidal range ship oil consumption prediction device and electronic equipment, wherein the method comprises the following steps: obtaining a training dataset comprising: on-line monitoring data of the ship, static data of the ship and real-time power of the ship; creating an initial convolutional neural network model, inputting the training data set into the initial convolutional neural network model for iterative training, and obtaining a convolutional neural network model with complete training by taking predicted real-time power as output; and acquiring real-time operation data of the ship, obtaining predicted instantaneous power of the ship at each moment of sailing according to the real-time operation data of the ship and the convolutional neural network with complete training, and obtaining predicted oil consumption and predicted total oil consumption of each period according to the predicted instantaneous power through conversion factors. According to the invention, the training data is combined and spliced by taking time as a main key so as to capture the trend change of the sailing state, so that the effective mining prediction of time sequence data is completed, and the oil consumption of the ship is accurately predicted.

Description

Method, device and electronic equipment for predicting fuel consumption of ships in tidal river sections

Technical Field

The present invention relates to the field of ship navigation, and in particular to a method, device, electronic equipment and computer-readable storage medium for predicting fuel consumption of ships in tidal river sections.

Background Art

In the current fierce environment of global economic competition, the shipping industry is facing a fierce competitive environment. With the increase in fuel prices, port fees and maintenance costs, the operating costs of ships have risen sharply. Therefore, strengthening the control of ship operating costs and reducing costs have become the key to whether shipping companies can gain a competitive advantage in the market competition.

In order to reduce the operating costs of the shipping industry, many scholars, universities and research institutions at home and abroad are actively applying big data technology to mine and study ship fuel consumption data, and to build fuel consumption models for ship main engines under different working conditions, to achieve fuel consumption optimization under different navigation environments, and have achieved good results.

However, there are still some problems in the current field of ship fuel consumption optimization: At present, the main research directions for ship energy efficiency optimization are speed optimization and route optimization to predict ship fuel consumption, among which route optimization is the most studied; but for tidal river sections, tidal changes at different sailing times have a great impact on the fuel consumption of ship navigation. The current energy efficiency big data used for data mining mostly only stays at the mining of static energy efficiency data points. However, energy efficiency data is a typical time series data. Relying solely on static data cannot effectively capture the trend changes of navigation status and realize effective identification of navigation status.

Summary of the invention

In view of this, it is necessary to provide a method, device, electronic device and computer storable medium for predicting ship fuel consumption in tidal river sections to solve the technical problem in the prior art that the dynamic energy efficiency data in tidal river sections is difficult to effectively capture and mine, resulting in inaccurate ship fuel consumption prediction.

In order to solve the above problems, the present invention provides a method for predicting fuel consumption of ships in tidal river sections, comprising:

Acquire a training data set, wherein the training data set includes: ship online monitoring data, ship static data and ship real-time power;

Creating an initial convolutional neural network model, and inputting the training data set into the initial convolutional neural network model for iterative training, so as to predict the real-time power as output, and obtain a fully trained convolutional neural network model;

Real-time ship operation data is obtained, and the predicted instantaneous power of the ship at each time of navigation is obtained according to the real-time ship operation data and the well-trained convolutional neural network. The predicted fuel consumption for each time period and the predicted total fuel consumption are obtained through conversion factors according to the predicted instantaneous power.

Furthermore, the obtaining of the training data set includes:

Collect ship online monitoring data, ship static data and ship real-time power based on different sensors, and merge and splice the ship online monitoring data, ship static data and ship real-time power with time as the primary key to obtain spliced data;

Based on the Laida criterion, abnormal data in the spliced data are eliminated, and the vacant data are filled using the mean method to obtain post-processed data;

Standardizing each parameter in the post-processing data so that each parameter is at the same order of magnitude;

The standardized data is subjected to noise reduction processing to obtain the training data set.

Furthermore, the step of performing noise reduction on the standardized data to obtain the training data set includes:

The correlation coefficient between parameters is calculated based on the Pearson correlation coefficient principle and according to the following formula:

Among them, μ _x , μ _y represent the means of two different parameters, σ _x , σ _y represent the standard deviations of the means of two different parameters, and x, y represent two different characteristic parameters;

Parameters with low correlation coefficients are removed according to the correlation coefficients between the parameters.

Furthermore, the ship online monitoring data includes wind speed, navigation speed, flow velocity and ship draft; the ship static data includes tide level and load capacity.

Furthermore, the creating an initial convolutional neural network model includes:

Create a multilayer perceptron network consisting of an input layer, a hidden layer, and an output layer, where each neuron in the input layer, hidden layer, and output layer is connected to all neurons in the adjacent layer, but there is no connection between neurons in the same layer, and establish the following mapping relationship and activation function:

g(x)＝

Among them, a _i , b _j , c _t represent the units of the input layer, hidden layer, and output layer respectively, w _ij represents the connection weight from the i-th input layer unit to the j-th hidden layer unit; v _jt represents the connection weight from the j-th hidden layer unit to the t-th output layer unit, θ _j represents the bias of the j-th hidden layer unit; γ _t represents the bias of the t-th output layer unit; f represents the activation function of the hidden layer unit, g represents the activation function of the output layer unit, and x represents the input amount of the hidden layer unit activation function f or the output layer activation function g.

Furthermore, the training data set is input into the initial convolutional neural network model for iterative training to predict the real-time power as output, thereby obtaining a fully trained convolutional neural network model, including:

The parameter data in the training data set are grouped according to the time primary key, and the corresponding parameters at similar time points are taken as the same group of parameters. The mean and variance of each parameter in each group of data are calculated as feature quantities and input into the initial convolutional neural network model. The initial convolutional neural network is iteratively trained using the global error function as the loss function until the loss no longer decreases, thereby obtaining a fully trained convolutional neural network model.

Furthermore, the real-time operation data of the ship is obtained, and the predicted instantaneous power of the ship at each time of navigation is obtained according to the real-time operation data of the ship and the well-trained convolutional neural network, and the predicted fuel consumption of each time period and the predicted total fuel consumption are obtained by conversion factors according to the predicted instantaneous power, including:

The real-time operation data of the ship corresponding to different departure time windows of the voyage are obtained, and the predicted instantaneous power at each moment is obtained according to the real-time operation data of the ship and the well-trained convolutional neural network. The predicted fuel consumption of each time period is obtained by conversion factor according to the predicted instantaneous power, and the predicted fuel consumption of the entire voyage is accumulated to obtain the predicted total fuel consumption of the current time window. The predicted total fuel consumption of different departure time windows is compared, and the time window with the lowest predicted total fuel consumption is selected as the recommended time window.

The present invention also provides a device for predicting fuel consumption of ships in tidal river sections, comprising:

A data acquisition module is used to obtain a training data set, wherein the training data set includes: ship online monitoring data, ship static data and ship real-time power;

A model training module, used to create an initial convolutional neural network, and input the training data set into the initial convolutional neural network model for iterative training, so as to predict the real-time power as output, and obtain a fully trained convolutional neural network model;

The fuel consumption prediction module is used to obtain real-time ship operation data, obtain the predicted instantaneous power of the ship at each moment of navigation according to the real-time ship operation data and the well-trained convolutional neural network, and obtain the predicted fuel consumption at each moment and the predicted total fuel consumption through a conversion factor according to the predicted instantaneous power.

The present invention also provides an electronic device, comprising a memory and a processor, wherein:

The memory is used to store programs;

The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in any one of the above-mentioned methods for predicting fuel consumption of ships in tidal river sections.

The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for predicting fuel consumption of ships in a tidal river section as described in any one of the above is implemented.

Compared with the prior art, the beneficial effects of the present invention include: the present invention first obtains a training data set, the training data set includes: ship online monitoring data, ship static data and ship real-time power; then creates an initial convolutional neural network, and inputs the training data set into the initial convolutional neural network model for iterative training, with the predicted real-time power as the output, to obtain a fully trained convolutional neural network model; finally obtains the ship's real-time operation data, and obtains the predicted instantaneous power of the ship at each moment of navigation according to the ship's real-time operation data and the fully trained convolutional neural network, and obtains the predicted fuel consumption at each moment and the predicted total fuel consumption according to the predicted instantaneous power through the conversion factor. The ship's online data is detected in real time through different sensors, and the ship's online monitoring data, ship static data and ship real-time power are merged and spliced with time as the primary key to capture the trend change of the navigation state, complete the effective mining of time series data, and realize the fuel consumption of the ship's navigation from the perspective of the time window, so as to achieve the purpose of reducing costs.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of an embodiment of a method for predicting fuel consumption of ships in tidal river sections provided by the present invention;

FIG2 is a flow chart of an embodiment of step S101 of the present invention;

FIG3 is a schematic structural diagram of an embodiment of a device for predicting fuel consumption of ships in tidal river sections provided by the present invention;

FIG. 4 is a schematic structural diagram of an embodiment of an electronic device provided by the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

Embodiments of the present invention provide a method, device, electronic device and computer storable medium for predicting fuel consumption of ships in a tidal river section, which are described below respectively.

FIG1 is a method for predicting fuel consumption in a tidal river section provided by the present invention. As shown in FIG1 , the method for predicting fuel consumption of ships in a tidal river section includes:

S101, obtaining a training data set, wherein the training data set includes: ship online monitoring data, ship static data and ship real-time power;

S102, creating an initial convolutional neural network model, and inputting the training data set into the initial convolutional neural network model for iterative training, so as to predict the real-time power as output, and obtain a fully trained convolutional neural network model;

S103, acquiring real-time operation data of the ship, obtaining predicted instantaneous power of the ship at each time of navigation according to the real-time operation data of the ship and the well-trained convolutional neural network, and obtaining predicted fuel consumption for each time period and predicted total fuel consumption through conversion factors according to the predicted instantaneous power.

It should be noted that the training data set may also include real-time weather data, navigation sway amplitude and other data.

In a specific embodiment, the ship's online monitoring data may be wind speed, navigation speed, flow rate, ship draft, wave level, propeller speed and engine temperature, etc.; the ship's static data may be tide level, load capacity, sailing time and anchoring time, etc.; some parameters may be added or reduced based on the actual situation of the route.

Compared with the prior art, the present invention first obtains a training data set, which includes: ship online monitoring data, ship static data and ship real-time power; then creates an initial convolutional neural network, and inputs the training data set into the initial convolutional neural network model for iterative training, with the predicted real-time power as the output, to obtain a fully trained convolutional neural network model; finally, obtains the ship's real-time operation data, and obtains the predicted instantaneous power of the ship at each moment of navigation according to the ship's real-time operation data and the fully trained convolutional neural network, and obtains the predicted fuel consumption at each moment and the predicted total fuel consumption according to the predicted instantaneous power through the conversion factor. The embodiment of the present invention performs real-time detection of ship online data through different sensors, and merges and splices the ship's online monitoring data, ship static data and ship real-time power with time as the primary key to capture the trend change of navigation status, complete the effective mining of time series data, and realize the fuel consumption of ship navigation from the perspective of time window, so as to achieve the purpose of reducing costs.

In a specific embodiment of the present invention, step S101 obtains a training data set, please refer to FIG. 2 , which specifically includes:

S201, collecting ship online monitoring data, ship static data and ship real-time power based on different sensors, and merging and splicing the ship online monitoring data, ship static data and ship real-time power with time as the primary key to obtain spliced data;

S202, eliminating abnormal data in the spliced data based on the Laida criterion, and filling in missing data using the mean method to obtain post-processed data;

S203, performing standardization processing on each parameter in the post-processing data so that each parameter is at the same order of magnitude;

S204: Perform noise reduction processing on the standardized data to obtain the training data set.

Specifically, in obtaining the training data set, the ship's online monitoring data is collected through different sensors, and the ship's static data and real-time power are obtained through manual input. Since the data are collected by different sensors during the data collection process, the collection frequency of each parameter is different and prone to fluctuations. Therefore, when merging different parameters, time is used as the primary key to splice the parameters to ensure that the frequency of each parameter is the same.

It should be noted that when performing parameter splicing, for data with different frequencies, a new frequency is set with time as the primary key to integrate the data. According to the time primary key, for each piece of data to be spliced, if the corresponding parameter has a collection parameter value at that time point, the collection parameter value will be used for data splicing; if the corresponding parameter does not have a parameter value at that time point due to different frequencies, the parameter value at the most recent time point will be used as the parameter value of the time primary key for data splicing.

At the same time, during the data collection process, due to the influence of the actual collection environment and communication failures during the transmission process, data anomalies and missing data may occur. For example, when measuring the draft of a ship, data anomalies and missing data may occur due to the influence of waves. For outliers, they can be eliminated according to the Laida criterion, and for missing data, the mean method can be used to fill them.

It should be noted that the mean method means that for missing data, the average value of the data from two previous and subsequent time points can be used to fill the missing data.

It should be noted that the Laida criterion (3σ criterion) refers to the assumption that a set of data contains only random errors for samples that are large enough to be considered as normally distributed or approximately normally distributed. The standard deviation is calculated and processed, and an interval is determined according to a certain probability. It can be considered that any error exceeding this interval is not a random error but a gross error, and the data containing this error should be eliminated. In addition, due to the different ranges and units of various data, some data have large differences in values, and it is necessary to standardize each parameter so that each parameter is at the same order of magnitude. The standardization formula is as follows:

Where x is the initial data, x ^′ is the standardized data, μ is the mean of the initial data, and σ is the variance of the initial data.

In a specific embodiment of the present invention, performing noise reduction processing on the standardized data to obtain the training data set includes:

The correlation coefficient between parameters is calculated based on the Pearson correlation coefficient principle and according to the following formula:

Among them, μ _x , μ _y represent the means of two different parameters, σ _x , σ _y represent the standard deviations of the means of two different parameters, and x, y represent two different characteristic parameters;

Parameters with low correlation coefficients are removed according to the correlation coefficients between the parameters.

Specifically, the navigation of a ship is affected by many factors. Considering the redundancy between these parameters, if the original data is directly used for modeling, the complexity of subsequent working condition identification and modeling will increase. Therefore, before modeling, it is necessary to perform noise reduction on the collected data and remove data with low correlation. The Pearson correlation coefficient principle is used here.

Among them, the correlation coefficient | _xy |≤1. When | _xy |≤0.2, it is uncorrelated, when 0.2＜| _xy |≤0.4, it is weakly correlated, when 0.4＜| _xy |≤0.6, it is moderately correlated, when 0.6＜| _xy |≤0.8, it is strongly correlated, and when 0.8＜| _xy |≤1, it is extremely strongly correlated. According to the correlation coefficient, the correlation between the parameters can be calculated, and the parameters with poor correlation can be removed.

In a specific embodiment of the present invention, the ship online monitoring data includes wind speed, navigation speed, flow rate and ship draft; the ship static data includes tide level and load capacity.

Specifically, in the process of collecting ship data, data such as wind speed, ship speed, current velocity and ship draft, which are likely to change with sailing time and tide during navigation, are monitored online through sensors, while data such as tide level and loading capacity that are not affected by the ship's navigation status are obtained through manual static input.

In a specific embodiment of the present invention, the creating an initial convolutional neural network model includes:

Create a multilayer perceptron network consisting of an input layer, a hidden layer, and an output layer, where each neuron in the input layer, hidden layer, and output layer is connected to all neurons in the adjacent layer, but there is no connection between neurons in the same layer, and establish the following mapping relationship and activation function:

g(x)＝

Among them, a _i , b _j , c _t represent the units of the input layer, hidden layer, and output layer respectively, w _ij represents the connection weight from the i-th input layer unit to the j-th hidden layer unit; v _jt represents the connection weight from the j-th hidden layer unit to the t-th output layer unit, θ _j represents the bias of the j-th hidden layer unit; γ _t represents the bias of the t-th output layer unit; f represents the activation function of the hidden layer unit, g represents the activation function of the output layer unit, and x represents the input amount of the hidden layer unit activation function f or the output layer activation function g.

Specifically, in the process of constructing the network model, the multilayer perceptron network in the artificial neural network is selected as the constructed network model. The model has good nonlinear mapping ability, adaptive learning ability and parallel processing ability, and has a good prediction effect in the fuel consumption prediction problem with many influencing factors and the causal relationship of related parameters that is difficult to accurately express. The constructed multilayer perceptron network consists of three connection layers: input layer, hidden layer, and output layer. All neurons in the input layer are connected to all neurons in the hidden layer, and all neurons in the hidden layer are connected to all neurons in the output layer. There is no connection between neurons in the same layer. Each input layer unit of the network corresponds to a dimension of the input variable and there is no bias, which only plays the role of data transmission. Data propagates from the input layer to the output layer in the network, thereby establishing a mapping from N-dimensional input data to Q-dimensional output data.

In a specific embodiment of the present invention, the training data set is input into the initial convolutional neural network model for iterative training to predict the real-time power as output, thereby obtaining a fully trained convolutional neural network model, including:

The parameter data in the training data set are grouped according to the time primary key, and the corresponding parameters at similar time points are taken as the same group of parameters. The mean and variance of each parameter in each group of data are calculated as feature quantities and input into the initial convolutional neural network model. The initial convolutional neural network is iteratively trained using the global error function as the loss function until the loss no longer decreases, thereby obtaining a fully trained convolutional neural network model.

Specifically, in the process of training the initial neural network model, the training data set of the voyage to be trained is divided into a training set and a test set in a ratio of 8:2, where the training set is used to train and improve the initial neural network model, and the test set is used to verify the performance of the network model to test the accuracy and effectiveness of the model. In addition, since the measured data is a huge amount of time-history data, and the recorded data at a single moment cannot represent the stable input-output relationship of the system, the mean and variance are used as input-output feature quantities to train the network model. The mean can represent the average intensity of the physical quantity, and the variance represents the degree of change of the physical quantity. The calculation formula is as follows:

Among them, u represents the data type, M represents the number of data points in the time series; _xi represents the value of the i-th point of this type of data.

After inputting the training parameters, the neural network is supervised for learning based on the inverse error propagation algorithm. The idea is to use the difference between the actual output and the expected output to correct the connection weights and bias values of each layer from back to front. When the training sample is provided to the network, the error function E between the network output and the expected output is calculated, and the error signal is propagated back along the original data propagation path so that the parameters of each layer of neurons are updated according to a certain learning step size. In order to speed up the training, the network parameters are uniformly updated based on the global error after all M training sample points are provided to the network. The global error function E is defined as follows:

表示第k个学习样本期望输出第t维，

表示网络对于第k个学习样本输出的第t维，k为学习样本编号，t为期望输出或网络输出的维度，在训练过程中基于梯度下降的原理进行网络训练，在i次迭代过程中，网络参数通过以下公式进行更新：in,

It indicates that the k-th learning sample is expected to output the t-th dimension,

It represents the t-th dimension of the network output for the k-th learning sample, k is the learning sample number, t is the dimension of the expected output or network output, and the network training is performed based on the principle of gradient descent during the training process. During the i-th iteration, the network parameters are updated by the following formula:

为全局误差函数对各网络参数的偏导，可据链式求导法则得到，α为参数更新的学习效率，用于确定每次迭代中参数的更新步长。Among them, ω is the connection weight and isolation bias between network layers.

is the partial derivative of the global error function with respect to each network parameter, which can be obtained according to the chain derivation rule. α is the learning efficiency of parameter update, which is used to determine the update step size of the parameters in each iteration.

In a specific embodiment of the present invention, real-time ship operation data is obtained, and the predicted instantaneous power of the ship at each time of navigation is obtained according to the real-time ship operation data and the well-trained convolutional neural network, and the predicted fuel consumption of each time period and the predicted total fuel consumption are obtained by conversion factors according to the predicted instantaneous power, including:

The real-time operation data of the ship corresponding to different departure time windows of the voyage are obtained, and the predicted instantaneous power at each moment is obtained according to the real-time operation data of the ship and the well-trained convolutional neural network. The predicted fuel consumption of each time period is obtained by conversion factor according to the predicted instantaneous power, and the predicted fuel consumption of the entire voyage is accumulated to obtain the predicted total fuel consumption of the current time window. The predicted total fuel consumption of different departure time windows is compared, and the time window with the lowest predicted total fuel consumption is selected as the recommended time window.

Specifically, for routes that require fuel consumption prediction, the tidal forecast during the corresponding voyage in different departure time windows is used to determine environmental data such as water flow velocity and direction at each moment, and the real-time data of the ship is determined by combining the data collected by sensors and the load quantity and other data. The instantaneous power at each moment is predicted based on the real-time data of the ship and the well-trained convolutional neural network, and the fuel consumption of the corresponding time period is calculated through conversion factors such as fuel calorific value and combustion efficiency, and the total fuel consumption is accumulated for the entire voyage. According to the predicted fuel consumption corresponding to different departure time points, the recommended sailing time window is selected.

In order to better implement the method for predicting fuel consumption of ships in tidal river sections in the embodiment of the present invention, based on the method for predicting fuel consumption of ships in tidal river sections, the present invention further provides a device for predicting fuel consumption of ships in tidal river sections 300, as shown in FIG. 3, comprising:

301. A data acquisition module, configured to obtain a training data set, wherein the training data set includes: ship online monitoring data, ship static data, and ship real-time power;

302. A model training module, used to create an initial convolutional neural network, and input the training data set into the initial convolutional neural network model for iterative training, so as to predict the real-time power as output, and obtain a fully trained convolutional neural network model;

303. A fuel consumption prediction module is used to obtain real-time ship operation data, obtain the predicted instantaneous power of the ship at each moment of navigation according to the real-time ship operation data and the trained convolutional neural network, and obtain the predicted fuel consumption at each moment and the predicted total fuel consumption through a conversion factor according to the predicted instantaneous power.

The device 300 for predicting fuel consumption of ships in tidal river sections provided in the above embodiment can implement the technical solution described in the above embodiment of the method for predicting fuel consumption of ships in tidal river sections. The specific implementation principles of the above modules or units can refer to the corresponding contents in the above embodiment of the method for predicting fuel consumption of ships in tidal river sections, which will not be repeated here.

The present invention also provides an electronic device based on the method for predicting fuel consumption of ships in tidal river sections, as shown in Figure 4. Figure 4 is a structural schematic diagram of an embodiment of an electronic device provided by the present invention. The electronic device 400 includes a processor 401, a memory 402, and a computer program stored in the memory 402 and executable on the processor 401. When the processor 401 executes the program, the method for predicting fuel consumption of ships in tidal river sections as described above is implemented.

The processor 401 may be an integrated circuit chip with signal processing capability. The processor 401 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC). The methods, steps and logic block diagrams disclosed in the embodiments of the present invention may be implemented or executed. The general-purpose processor may also be a microprocessor or the processor may also be any conventional processor, etc.

The memory 402 may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a secure digital (SD card), a flash card, etc. The memory 402 is used to store programs, and the processor 401 executes the programs after receiving the execution instruction. The process definition method disclosed in any of the embodiments of the present invention may be applied to the processor 401 or implemented by the processor 401.

An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for predicting fuel consumption of ships in tidal river sections as described above is implemented.

Generally speaking, the computer instructions for implementing the method of the present invention can be carried by any combination of one or more computer-readable storage media. Non-transitory computer-readable storage media can include any computer-readable media except for the signal itself that is temporarily propagating.

The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims (10)

1. A method for predicting fuel consumption of ships in tidal river sections, characterized by comprising: Acquire a training data set, wherein the training data set includes: ship online monitoring data, ship static data and ship real-time power; Creating an initial convolutional neural network model, and inputting the training data set into the initial convolutional neural network model for iterative training, so as to predict the real-time power as output, and obtain a fully trained convolutional neural network model; Real-time ship operation data is obtained, and the predicted instantaneous power of the ship at each moment of navigation is obtained according to the real-time ship operation data and the well-trained convolutional neural network. The predicted fuel consumption for each time period and the predicted total fuel consumption are obtained through conversion factors according to the predicted instantaneous power. 2. The method for predicting fuel consumption of ships in tidal river sections according to claim 1, wherein obtaining a training data set comprises: Collect ship online monitoring data, ship static data and ship real-time power based on different sensors, and merge and splice the ship online monitoring data, ship static data and ship real-time power with time as the primary key to obtain spliced data; Based on the Laida criterion, abnormal data in the spliced data are eliminated, and the vacant data are filled using the mean method to obtain post-processed data; Standardizing each parameter in the post-processing data so that each parameter is at the same order of magnitude; The standardized data is subjected to noise reduction processing to obtain the training data set. 3. The method for predicting fuel consumption of ships in tidal river sections according to claim 2, characterized in that the step of performing noise reduction processing on the standardized data to obtain the training data set comprises: The correlation coefficient between parameters is calculated based on the Pearson correlation coefficient principle and according to the following formula:

Among them, μ _x , μ _y represent the means of two different parameters, σ _x , σ _y represent the standard deviations of the means of two different parameters, and x, y represent two different characteristic parameters; Parameters with low correlation coefficients are removed according to the correlation coefficients between the parameters. 4. The method for predicting fuel consumption of ships in tidal river sections according to claim 1 is characterized in that the online monitoring data of the ship includes wind speed, navigation speed, flow velocity and ship draft; and the static data of the ship includes tide level and load capacity. 5. The method for predicting fuel consumption of ships in tidal river sections according to claim 1, characterized in that the creation of an initial convolutional neural network model comprises: Create a multilayer perceptron network consisting of an input layer, a hidden layer, and an output layer, where each neuron in the input layer, hidden layer, and output layer is connected to all neurons in the adjacent layer, but there is no connection between neurons in the same layer, and establish the following mapping relationship and activation function:

g(x)＝x Among them, a _i , b _j , c _t represent the units of the input layer, hidden layer, and output layer respectively, w _ij represents the connection weight from the i-th input layer unit to the j-th hidden layer unit; v _jt represents the connection weight from the j-th hidden layer unit to the t-th output layer unit, θ _j represents the bias of the j-th hidden layer unit; γ _t represents the bias of the t-th output layer unit; f represents the activation function of the hidden layer unit, g represents the activation function of the output layer unit, and x represents the input amount of the hidden layer unit activation function f or the output layer activation function g. 6. The method for predicting fuel consumption of ships in tidal river sections according to claim 1 is characterized in that the training data set is input into the initial convolutional neural network model for iterative training to predict the real-time power as output to obtain a fully trained convolutional neural network model, comprising: The parameter data in the training data set are grouped according to the time primary key, and the corresponding parameters at similar time points are taken as the same group of parameters. The mean and variance of each parameter in each group of data are calculated as feature quantities and input into the initial convolutional neural network model. The initial convolutional neural network is iteratively trained using the global error function as the loss function until the loss no longer decreases, thereby obtaining a fully trained convolutional neural network model. 7. The method for predicting fuel consumption of ships in tidal river sections according to claim 1 is characterized in that the real-time operation data of the ship is obtained, the predicted instantaneous power of the ship at each time of navigation is obtained according to the real-time operation data of the ship and the well-trained convolutional neural network, and the predicted fuel consumption of each time period and the predicted total fuel consumption are obtained by conversion factors according to the predicted instantaneous power, including: The real-time operation data of the ship corresponding to different departure time windows of the voyage are obtained, and the predicted instantaneous power at each moment is obtained according to the real-time operation data of the ship and the well-trained convolutional neural network. The predicted fuel consumption of each time period is obtained by conversion factor according to the predicted instantaneous power, and the predicted fuel consumption of the entire voyage is accumulated to obtain the predicted total fuel consumption of the current time window. The predicted total fuel consumption of different departure time windows is compared, and the time window with the lowest predicted total fuel consumption is selected as the recommended time window. 8. A device for predicting fuel consumption of ships in tidal river sections, characterized by comprising: A data acquisition module is used to obtain a training data set, wherein the training data set includes: ship online monitoring data, ship static data and ship real-time power; A model training module, used to create an initial convolutional neural network, and input the training data set into the initial convolutional neural network model for iterative training, so as to predict the real-time power as output, and obtain a fully trained convolutional neural network model; The fuel consumption prediction module is used to obtain real-time ship operation data, obtain the predicted instantaneous power of the ship at each moment of navigation according to the real-time ship operation data and the well-trained convolutional neural network, and obtain the predicted fuel consumption at each moment and the predicted total fuel consumption through a conversion factor according to the predicted instantaneous power. 9. An electronic device, comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the method for predicting fuel consumption of ships in tidal river sections as described in any one of claims 1 to 7 above. 10. A computer-readable storage medium, characterized in that a computer program is stored thereon, and when the computer program is executed by a processor, the method for predicting fuel consumption of ships in a tidal river section according to any one of claims 1 to 7 is implemented.

Images