Method and system for correcting basic resistance of railway vehicle

Abstract

The invention relates to a correction method and system for the basic resistance of a railway vehicle, and relates to the technical field of railway vehicle control, and the method comprises the steps: collecting the real-time operation data of the vehicle, and calculating the detection value of the basic resistance of the vehicle based on the real-time operation data; inputting the data of the influence factors of the basic resistance of the vehicle into a neural network model, training the neural network model, and outputting a predicted value of the basic resistance of the vehicle; comparing the predicted value of the basic resistance of the vehicle with the detected value of the basic resistance of the vehicle, and correcting the neural network model according to a comparison result; and outputting a corrected value of the basic resistance of the vehicle through the corrected neural network model. According to the method provided by the invention, the basic resistance of the vehicle can be predicted and corrected according to the operation data of the railway vehicle, so that the precision of train traction calculation and energy consumption simulation is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of rail vehicle control, and in particular to a rail vehicle basic resistance correction method and system. BACKGROUND

[0002] The rail vehicle basic resistance is generated in the process of mutual friction or impact between the rail vehicle and air, steel rail, parts, etc., and is generally composed of mechanical resistance and air resistance. It is influenced by multiple factors and is very complex, and cannot be directly measured, so it is difficult to calculate its size using accurate data formulas or theories. In the actual running process, it is calculated by an empirical formula. The empirical formula is a monomial quadratic equation of the running speed of the rail vehicle, which includes multiple empirical coefficients. By setting the value of the empirical coefficient and detecting the running speed of the vehicle, the basic resistance of the rail vehicle is calculated.

[0003] In the calculation of rail vehicle traction and energy consumption simulation, the empirical coefficient is often set as a fixed value. The use of a fixed value will lead to a decrease in calculation accuracy and will not fully consider the various factors of the actual operation, such as airflow, weather, and track conditions. This fixed coefficient model will lead to prediction deviations in energy consumption and traction force, and the simulation results may overestimate or underestimate the energy consumption of the train, affecting system design and energy optimization, and thus affecting energy-saving strategies and equipment maintenance. Moreover, the fixed coefficient model lacks real-time adjustment capability and cannot adapt to various complex working conditions faced by the vehicle.

[0004] Therefore, there is a need for a method for calculating and correcting the basic resistance of a rail vehicle, which can dynamically adjust the value of the empirical coefficient according to the running conditions of the train to more accurately calculate the basic resistance of the rail vehicle in complex environments, thereby improving the accuracy of train traction calculation and energy consumption simulation. SUMMARY

[0005] The present application at least partly solves one of the technical problems in the related art, and provides a rail vehicle basic resistance correction method that predicts and corrects the basic resistance of a vehicle based on the running data of the rail vehicle, thereby improving the accuracy of train traction calculation and energy consumption simulation.

[0006] To achieve the above-mentioned purpose, the first aspect of the present application provides a rail vehicle basic resistance correction method, comprising: collecting real-time running data of the vehicle, calculating a detection value of the basic resistance of the vehicle based on the real-time running data; inputting data of factors influencing the basic resistance of the vehicle into a neural network model, training the neural network model, and outputting a prediction value of the basic resistance of the vehicle; comparing the prediction value of the basic resistance of the vehicle with the detection value of the basic resistance of the vehicle, and correcting the neural network model according to the comparison result; and outputting a correction value of the basic resistance of the vehicle through the corrected neural network model.

[0007] In this embodiment, the detection value calculated by real-time running data is taken as a true reference, which can reflect the true resistance state of the vehicle in actual working conditions. After the neural network model is trained and compared and corrected, it can dynamically adapt to different operating conditions, solve the error problem caused by ignoring real-time factors (such as wind speed and track state) in traditional fixed coefficient models, and make the output correction value closer to the actual resistance. At the same time, accurate basic resistance correction value can improve the accuracy of train energy consumption simulation and traction force calculation, avoid overestimation or underestimation of energy consumption caused by resistance prediction deviation, and provide more scientific support for train energy saving strategy formulation (such as speed regulation and power distribution) and equipment maintenance.

[0008] In some embodiments of the present application, the influencing factors of the basic resistance of the vehicle include environmental factors, and the data of the influencing factors of the basic resistance of the vehicle include environmental data; the method for training the neural network model comprises: dividing the environmental data into different environmental type data according to environmental types, and dividing index intervals for each environmental type data; collecting real-time environmental data of the vehicle, and inputting the real-time environmental data into the neural network model; the neural network model adjusts the weights of each environmental type in calculating the predicted value of the basic resistance of the vehicle according to the index intervals to which the indexes of each type of real-time environmental data belong.

[0009] In this embodiment, environmental factors are key influencing factors of basic resistance, and modular division and interval processing can "decouple" complex environmental variables and reduce the interference of data noise on the model. At the same time, the dynamic weight adjustment mechanism can respond to environmental changes in real time, solving the problem that the traditional fixed coefficient model cannot adapt to environmental fluctuations in real time, and making the predicted value closer to the actual resistance detection value.

[0010] In some embodiments of the present application, the environmental types at least include altitude, temperature, humidity and weather, and the method for training the neural network model further comprises: dividing altitude index intervals with reference to altitude, dividing temperature index intervals with reference to annual average temperature, dividing humidity index intervals with reference to annual average humidity, and dividing weather index intervals of at least rain, snow, sunny, cloudy and cloudy; collecting the altitude, annual average temperature, annual average humidity and real-time weather of the area where the vehicle is located, and inputting them into the neural network model; the neural network model adjusts the weights of each environmental type in calculating the predicted value of the basic resistance of the vehicle according to the altitude index intervals to which the altitude of the area where the vehicle is located belongs, the temperature index intervals to which the annual average temperature belongs, the humidity index intervals to which the annual average humidity belongs, and the weather index intervals to which the real-time weather belongs.

[0011] In this embodiment, the altitude, the annual average temperature, and the annual average humidity are taken as reference to divide the index interval, which can accurately capture the environmental characteristics of different geographical regions. The weather is divided into types such as rain and snow, which can directly associate the real-time influence of weather on air resistance and wheel-rail resistance. This division method enables the neural network to focus on the core influence logic of each environmental type, avoids the confusion of the association relationship caused by the mixed input of different environmental characteristics, and significantly reduces the deviation between the predicted value and the actual detection value.

[0012] In some embodiments of the present application, the method for correcting the basic resistance of the rail vehicle further comprises: setting initial weights of each environmental type in calculating the basic resistance of the vehicle according to the region, and the neural network model adjusts the initial weights of each environmental type in calculating the basic resistance of the vehicle according to the index interval to which the index of the real-time environmental data belongs.

[0013] In this embodiment, the initial weights of each environmental type are set according to the region, which can make the model fit the inherent environmental laws of different regions in advance. For example, the initial weight of the "altitude" factor can be increased in high-altitude areas, and the initial weight of the "humidity" factor can be increased in humid coastal areas, thereby reducing the prediction deviation from the basic level. At the same time, the neural network further adjusts the weights based on the initial weights in combination with the index interval to which the real-time environmental data belongs, which not only retains the inherent influence of the regional environment, but also accurately responds to real-time fluctuations, making the resistance calculation more in line with the actual working conditions.

[0014] In some embodiments of the present application, the influencing factors of the basic resistance of the vehicle include quantitative factors and variable factors, the quantitative factors at least include the characteristic attributes of the vehicle, and the environmental factors are included in the variable factors, the variable factors at least further include the running speed of the vehicle and the load of the vehicle; the method for training the neural network model comprises: the neural network model receives the data of the quantitative factors and the variable factors, takes the detection value of the basic resistance of the vehicle as a label, adjusts the weights and biases of the neural network model, so that the predicted value of the basic resistance calculated by the model approaches the detection value of the basic resistance, thereby training the model.

[0015] In this embodiment, the quantitative factors and the variable factors are combined and input into the model, which can completely cover the core influence dimensions of the basic resistance, and avoid the prediction deviation caused by missing key factors. At the same time, the model is trained by taking the detection value of the basic resistance as a label, and the predicted value is made close to the detection value by adjusting the weights and biases, which can make the model accurately learn the mapping relationship between the "quantitative + variable factors" and the basic resistance, and minimize the prediction error.

[0016] In some embodiments of the present application, the empirical formula for calculating the predicted value of the unit basic resistance of the vehicle is , is the running speed of the vehicle, , , is a coefficient of an empirical formula; the method for modifying the neural network model comprises: selecting an algorithm for modifying the neural network model; setting a decision variable of the algorithm as a coefficient of an empirical formula , , , the objective function is set as the minimum difference between the detected value and the predicted value of the basic resistance of the vehicle, and the coefficient of the empirical formula is obtained by calculating the optimal solution of the objective function , , ; according to the modified value of the coefficient of the empirical formula , , , the modified value of the basic resistance of the vehicle is calculated.

[0017] In the embodiment, the coefficients , , in the empirical formula are respectively associated with the mechanical resistance , and the air resistance , avoiding the influence of fixed coefficients in the traditional method on dynamic factors such as speed, environment and load. By setting , , as a decision variable and solving the optimal solution with the minimum difference between the detected value and the predicted value as the objective function, the coefficients can be dynamically modified according to the actual working conditions. The modified coefficients are substituted into the empirical formula, which can make the modified value of the basic resistance more consistent with the real detected value, and solve the calculation deviation problem caused by ignoring dynamic factors in the traditional fixed coefficient model, providing accurate data basis for train traction calculation and energy consumption simulation.

[0018] In some embodiments of the application, the objective function is , wherein, is the detected value of the basic resistance of the vehicle, is the predicted value of the basic resistance of the vehicle.

[0019] In some embodiments of the application, the method for inputting the real-time environmental data into the neural network model comprises: normalizing the real-time environmental data, converting the real-time environmental data into standardized real-time environmental data with a mean value of 0 and a standard deviation of 1, and inputting the standardized real-time environmental data into the neural network model.

[0020] In this embodiment, through normalization processing, all environmental data are converted to a unified scale, ensuring that each environmental factor such as altitude, temperature, humidity, etc. has equal initial influence in model training. This enables the model to more evenly learn the real influence law of different environmental factors on the basic resistance, avoids prediction bias caused by data scale differences, and improves the basic resistance prediction accuracy.

[0021] In some embodiments of the present application, the correction method of the basic resistance of the rail vehicle further comprises: dividing the data of the influencing factors of the basic resistance of the vehicle into a test set according to a set proportion, outputting the correction value of the basic resistance of the vehicle according to the test set, and calculating a loss function between the detection value and the correction value of the basic resistance of the vehicle. The accuracy of the correction value of the basic resistance of the vehicle calculated by the neural network model is judged according to the size of the calculated loss function value.

[0022] In this embodiment, the size of the loss function value directly quantifies the deviation between the correction value and the real detection value, providing a measurable index for model performance. Based on the feedback of the loss function, the model can be optimized, forming a closed loop of "training - testing - optimization", continuously improving the accuracy of the correction value, and solving the problem of difficult quantitative evaluation of resistance calculation accuracy in traditional methods.

[0023] The second aspect of the present application provides a system for implementing the above-mentioned correction method of the basic resistance of the rail vehicle, comprising a data acquisition module, a calculation module, a training module, a correction module and an output module: the data acquisition module is used to acquire real-time running data of the vehicle; the calculation module calculates the detection value of the basic resistance of the vehicle according to the real-time running data of the vehicle; the training module trains the neural network model according to the data of the influencing factors of the basic resistance of the vehicle, and calculates the prediction value of the basic resistance of the vehicle through the trained neural network model; the correction module compares the prediction value of the basic resistance of the vehicle with the detection value of the basic resistance of the vehicle, and corrects the neural network model according to the comparison result, and calculates the correction value of the basic resistance of the vehicle through the corrected neural network model; the output module outputs the correction value of the basic resistance of the vehicle.

[0024] The above description is only a summary of the technical solutions of the present disclosure. In order to more clearly understand the technical means of the present disclosure, the contents of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the present disclosure more obvious and easy to understand, the specific embodiments of the present disclosure are described below. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without any creative effort.

[0026] Fig. 1 is a flow chart of main steps of the correction method of the basic resistance of the rail vehicle according to the embodiment of the present application; Fig. 2 is a schematic diagram of modular processing of environmental factors according to the embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to make the technical problems, technical solutions and beneficial effects of the present application more clearly understood, the following will further describe the present application in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and do not limit the present application.

[0028] In the embodiments of the present application, the prefix words such as "first", "second" are only used to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. The use of ordinal words such as ordinal words in the embodiments of the present application does not limit the described objects, and the description of the described objects should be seen in the context of the claims or embodiments, and should not be limited by the use of such prefix words. In addition, in the description of the embodiments, unless otherwise stated, the meaning of "multiple" is two or more.

[0029] The technical solutions in the embodiments of the present application will be described in combination with the drawings in the embodiments of the present application. In the description of the embodiments of the present application, unless otherwise stated, " / " represents the meaning of or, for example, A / B can represent A or B; "and / or" in this paper is only a description of the association between the associated objects, which means that there can be three kinds of relationships, for example, A and / or B, which can represent three cases: A exists alone, A and B exist together, and B exists alone.

[0030] It should be understood that the disclosed system and method can be implemented in other ways. For example, the device embodiments described above are merely illustrative, and the division of the units is merely a logical function division. In actual implementation, another division manner can be used, for example, a plurality of units or components can be combined or integrated into another system, or some features can be omitted or not executed. In addition, the coupling or direct coupling or communication connection between the units or components shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0031] In this application, the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the application. In this specification, the illustrative representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the skilled person in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples without contradiction.

[0032] The basic resistance of the rail vehicle is generated in the process of mutual friction or impact between the rail vehicle and air, steel rail, parts, etc. It is generally composed of mechanical resistance and air resistance, which is influenced by many factors and is very complex, and cannot be directly measured, and it is difficult to calculate its size using accurate data formula or theory. In the actual running process, it is calculated by an empirical formula. The empirical formula is a monomial quadratic equation of the running speed of the rail vehicle, which includes a plurality of empirical coefficients. By setting the value of the empirical coefficient and detecting the running speed of the vehicle, the basic resistance of the rail vehicle is calculated.

[0033] In the calculation of rail vehicle traction and energy consumption simulation, the empirical coefficient is often set as a fixed value. The use of fixed values will lead to a decrease in calculation accuracy and cannot fully consider the variable factors of actual operation, such as airflow, weather, track conditions, etc. This fixed coefficient model will lead to prediction deviation of energy consumption and traction force, and the simulation result may overestimate or underestimate the energy consumption of the train, affecting system design and energy saving optimization, and further affecting energy saving strategies and equipment maintenance. Moreover, the fixed coefficient model lacks real-time adjustment capability and cannot adapt to various complex working conditions faced by the vehicle.

[0034] In order to solve the above problems, the application provides a correction method for basic resistance of a rail vehicle, which dynamically adjusts the value of an empirical coefficient according to the running condition of the train, so as to more accurately calculate the basic resistance of the rail vehicle in a complex environment, thereby improving the accuracy of train traction calculation and energy consumption simulation.

[0035] In the following, the embodiments of the application will be described in detail with reference to the accompanying drawings.

[0036] As shown in the accompanying Figs. 1-2 In an exemplary embodiment of the application, the correction method for basic resistance of a rail vehicle mainly comprises the following steps: S1: collecting real-time running data of the vehicle, and calculating a detection value of the basic resistance of the vehicle based on the real-time running data; S2: inputting data of influencing factors of the basic resistance of the vehicle into a neural network model, training the neural network model, and outputting a prediction value of the basic resistance of the vehicle; S3: comparing the prediction value of the basic resistance of the vehicle with the detection value of the basic resistance of the vehicle, and correcting the neural network model according to the comparison result; S4: outputting a correction value of the basic resistance of the vehicle through the corrected neural network model.

[0037] The basic running resistance of a train can be specifically divided into mechanical resistance and air resistance. The resistance generated by the wheel-rail coupling (including the pantograph coupling and the bearing friction) is referred to as mechanical resistance, which includes bearing friction resistance, rolling friction resistance and sliding friction resistance, and is approximately expressed by a linear function of speed. The resistance generated by the coupling relationship between the outer layer and the air is referred to as air resistance, which is approximately expressed by a square relationship of speed.

[0038] The unit basic running resistance of a rail vehicle running on a straight track on an open line is generally obtained through a coasting test, and the calculation formula of the unit basic running resistance is as follows: .

[0039] In the formula, is the unit basic running resistance (N / t), is the running speed of the vehicle (km / h); , , is a regression undetermined coefficient; represents the mechanical resistance, and the coefficient , The size of the coefficient is mainly affected by the rolling friction resistance, the sliding friction resistance and the bearing friction resistance; represents the air resistance, which is mainly affected by the air resistance.

[0040] Therefore, in some embodiments, the calculation method of the detection value of the basic resistance of the vehicle in step S1 is: Obtain the real-time running data of the vehicle in sunny weather and wind force below level 4, filter out the data during the vehicle's inertial period according to the conditions that the traction force is 0, the braking force is 0, and the speed is not 0, and obtain the acceleration a of the vehicle at the current time during the vehicle's inertial period according to the data during the vehicle's inertial period , wherein, is the running speed of the vehicle at the current time, is the running speed of the vehicle at the next time, is the time interval between the current time and the next time.

[0041] The resultant force C suffered by the vehicle at the current time can be obtained from the acceleration a of the vehicle at the current time ; wherein M is the load of the vehicle, is the rotational mass coefficient.

[0042] According to the real-time running data of the vehicle, the running additional resistance W1 of the vehicle is calculated, and the running additional resistance of the vehicle is mainly the additional resistance caused by the line conditions (slope, curve, tunnel). At the same time, combined with the resultant force C suffered by the vehicle at the current time, the actual basic resistance W0 suffered by the vehicle at the current time can be obtained The actual basic resistance suffered by the vehicle at the current time calculated by this method is the detection value of the basic resistance of the vehicle, and at the same time, the detection value is also regarded as the actual value of the basic resistance of the vehicle, which is used to train and optimize the value predicted by the neural network model.

[0043] In some embodiments, the method of inputting the data of the influencing factors of the basic resistance of the vehicle into the neural network model in step S2 includes: The influencing factors of the basic resistance of the vehicle are decoupled by the first principle, as shown in Table 1, the influencing factors of the basic resistance of the vehicle can be divided into variable factors and quantitative factors, wherein the quantitative factors at least include the characteristic attributes of the vehicle, after determining the train model and state value, the variable factors can be formed into a data set and input into the neural network model; the variable factors change with the running of the vehicle, and need to deploy a vehicle-mounted sensor network to collect real-time data and form a data set to input into the neural network.

[0044] Table 1 Basic resistance influencing factor analysis table

[0045] Considering the dynamic disturbance of the environmental changes to the air fluid density and wind speed in the variable factors during the long-distance running of the railway vehicle. Therefore, in some embodiments, the influencing factors of the basic resistance of the vehicle include environmental factors, and the data of the influencing factors of the basic resistance of the vehicle include environmental data. The method of training the neural network model in step S2 includes: dividing the environment data into different environment type data according to the environment type, and dividing index intervals for each environment type data respectively; collecting real-time environment data of the vehicle, and inputting the real-time environment data into the neural network model; The neural network model adjusts the weight of each environment type in calculating the predicted value of the basic resistance of the vehicle according to the index interval to which the index of each type of real-time environment data belongs, using a transfer learning strategy.

[0046] In some embodiments, the method of training the neural network model in step S2 further comprises setting an initial weight of each environment type in calculating the basic resistance of the vehicle according to the region, and the neural network model adjusts the initial weight of each environment type in calculating the basic resistance of the vehicle according to the index interval to which the index of each type of real-time environment data belongs.

[0047] Specifically, in some embodiments, the above-mentioned environment types at least include altitude, temperature, humidity and weather, and the method of training the neural network model further comprises: dividing the altitude index interval with reference to the altitude, dividing the temperature index interval with reference to the annual average temperature, dividing the humidity index interval with reference to the annual average humidity, and dividing the weather data into at least the weather index intervals of rain, snow, sunny, cloudy, and overcast; collecting the altitude, annual average temperature, annual average humidity and real-time weather of the region where the vehicle is located, and inputting them into the neural network model; The neural network model adjusts the weight of each environment type in calculating the predicted value of the basic resistance of the vehicle according to the altitude index interval to which the altitude of the region where the vehicle is located belongs, the temperature index interval to which the annual average temperature belongs, the humidity index interval to which the annual average humidity belongs, and the weather index interval to which the real-time weather belongs.

[0048] Illustratively, the setting of the initial weight and the adjustment of the weight can be understood by the following steps: (1) The types of environmental factors affecting the basic resistance of the train are divided into 4 modules, and each module contains specific indexes: Altitude module: index is divided into low altitude (<1000m), medium altitude (1000-3000m), high altitude (>3000m); Temperature module: index is divided into low temperature (<0℃), normal temperature (0-25℃), high temperature (>25℃); Humidity module: index is divided into dry (<40%), humid (40%-70%), and humid (>70%); Weather type module: index is divided into sunny day, cloudy day, rainy day, snowy day, and cloudy.

[0049] (2) For different regional environmental characteristics, set initial weights for each module (the sum of the weights is 1, and the greater the value, the more significant the influence of the module on the basic resistance): In coastal plain areas (such as a line in the Yangtze River Delta), due to low elevation and dense air, air resistance (humidity, weather module) has a stronger influence, and the initial weight is set as: Elevation module: 0.1 (low elevation, small influence); Temperature module: 0.2 (mainly at room temperature, stable influence); Humidity module: 0.4 (high humidity significantly increases air resistance); Weather type module: 0.3 (rainy days, cloudy days have a large influence on air resistance).

[0050] (3) When the train collects real-time environmental data, the neural network will determine the region to which each module belongs and dynamically adjust the weight: When the coastal plain area enters the plum rain season, the real-time environmental data shows "elevation 50 meters, temperature 28°C, humidity 85%, and weather is rainy."

[0051] The neural network model recognizes that "high humidity + rainy days" will significantly increase air resistance, so it increases the "humidity module" weight (e.g., from 0.4 to 0.5) and the "weather type module" weight (e.g., from 0.3 to 0.35), and reduces the "elevation module" weight (e.g., from 0.1 to 0.05). The adjusted weights are: Elevation (0.05) + Temperature (0.1) + Humidity (0.5) + Weather (0.35) = 1.

[0052] Through the above process, the neural network can dynamically adjust the weights of each module according to the environmental module division of different regions and real-time indicators, making the basic resistance prediction more in line with the actual regional and working condition characteristics.

[0053] After dividing the collected environmental data into types and indicator intervals, due to the large difference in the value range of environmental data in different regions, in some embodiments, the method of inputting real-time environmental data into the neural network model further includes: Using the Z-score method to normalize the real-time environmental data, the real-time environmental data is converted into standardized real-time environmental data with a mean of 0 and a standard deviation of 1, and the standardized real-time environmental data is input into the neural network model.

[0054] The calculation formula of the Z-score method is: ; is the original environmental data (such as the real-time detection value of temperature and humidity in a certain area), the mean of the environmental data of this type (e.g. the average of the temperature data of multiple regions); the standard deviation of the environmental data of this type (reflecting the degree of dispersion of the data); the normalized result, i.e. the data after Z-score processing.

[0055] In some embodiments, the variable factors include at least the running speed of the vehicle and the load of the vehicle in addition to the environmental factors. The number of neurons in the input layer is selected according to the number of types of variable factors to be input into the neural network model. The number of neurons in the output layer is selected according to the dimension of the output target, and the number of neurons in the hidden layer is also set as needed.

[0056] In some embodiments, the method of training the neural network model includes: The neural network model receives data of the quantitative factors and the variable factors, divides the received data into a training set and a test set in a ratio of 70% and 30%, and uses the test set to train the model.

[0057] The neural network model receives data of the quantitative factors and the variable factors (the test set), takes the detected value of the basic resistance of the vehicle as the label, continuously adjusts the weights and biases of the neural network model through the back propagation algorithm, optimizes the loss function between the detected value and the predicted value of the basic resistance, and makes the predicted value of the basic resistance calculated by the model continuously close to the detected value of the basic resistance, thereby training the model.

[0058] It should be understood that, by adjusting the weights of the neural network model through the back propagation algorithm, the internal connection weights of the neural network are adjusted, and the purpose is to minimize the error of the predicted basic resistance; and by adjusting the weights of each environment type in the environment module of the neural network model, the weight parameters of each environment type in the environment module of the neural network model are adjusted, and the purpose is to enhance the generalization ability of the model to geographical / climatic differences.

[0059] In some embodiments, the BP (back propagation) algorithm is used to train the neural network model, including two processes of forward propagation of signals and backward propagation of errors.

[0060] The forward propagation of signals refers to inputting the data of the influencing factors of the basic resistance of the rail vehicle into the input layer of the neural network, and after the data are calculated and processed by the current weights and biases in the hidden layer, the data are transmitted to the output layer and the predicted value of the basic resistance is outputted; The backward propagation of errors refers to comparing the predicted value of the basic resistance obtained by the output layer with the detected value of the real-time basic resistance, calculating the error (loss function) between the two, and transmitting the error from the output layer to the hidden layer and the input layer, and then adjusting the weights and biases of each layer according to the error size to reduce the deviation between the predicted value and the detected value.

[0061] Through the iterative cycle of the two processes, the neural network model can continuously optimize the parameters, and finally realize the accurate prediction of the basic resistance of the rail vehicle.

[0062] In some embodiments, the empirical formula for predicting the value of the basic resistance of the vehicle is obtained according to the least square method regression based on the running speed v of the vehicle at the current time, the load M of the vehicle Therefore, the empirical formula for predicting the value of the unit basic resistance of the vehicle is . Wherein, is the running speed of the vehicle (km / h), M is the load of the vehicle, g is the acceleration of gravity, a, b, c are the coefficients of the basic resistance empirical formula.

[0063] In some embodiments, the method for modifying (optimizing) the neural network model in step S3 includes: selecting an algorithm for modifying (optimizing) the neural network model; setting the decision variables of the algorithm for modification as the coefficients of the basic resistance empirical formula 、 、 , setting the objective function as the minimum difference between the detected value and the predicted value of the basic resistance of the vehicle, and obtaining the modification value of the coefficients 、 、 of the empirical formula by calculating the optimal solution of the objective function; According to the modification value of the coefficients 、 、 of the empirical formula and the empirical formula of the basic running resistance, the modification value of the basic resistance of the vehicle is calculated.

[0064] In some embodiments, evolutionary algorithms such as particle swarm optimization (PSO) and genetic algorithm (GA) are used to modify the neural network model.

[0065] In some embodiments, the objective function is , wherein, is the detected value of the basic resistance of the vehicle, is the predicted value of the basic resistance of the vehicle.

[0066] In some embodiments, the method for calculating and modifying the basic resistance further includes: outputting the modification value of the basic resistance of the vehicle according to the test set, calculating the loss function between the detected value and the modification value of the basic resistance of the vehicle, and judging the accuracy of the modification value of the basic resistance of the vehicle calculated by the neural network model according to the size of the calculated loss function value.

[0067] Through the above steps, the neural network prediction model is established, the relationship modeling and prediction between the basic resistance formula and the train operation variable factors are realized, and thus the correction of the basic resistance formula is obtained.

[0068] In some embodiments, the application also provides a system for implementing the above-mentioned correction method of the basic resistance of the railway vehicle, comprising: a data acquisition module for acquiring real-time operation data of the vehicle; a calculation module for calculating a detected value of the basic resistance of the vehicle according to the real-time operation data of the vehicle; a training module for training a neural network model according to the data of the influencing factors of the basic resistance of the vehicle, and calculating a predicted value of the basic resistance of the vehicle through the trained neural network model; a correction module for comparing the predicted value of the basic resistance of the vehicle with the detected value of the basic resistance of the vehicle, correcting the neural network model according to the comparison result, and calculating a corrected value of the basic resistance of the vehicle through the corrected neural network model; an output module for outputting the corrected value of the basic resistance of the vehicle.

[0069] The above is only a specific embodiment of the application, but the protection scope of the application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the application, which should be covered. Therefore, the protection scope of the application should be subject to the protection scope of the claims.

Claims

1. A method for correcting the basic resistance of a rail vehicle, characterized in that, include: Collect real-time operating data of the vehicle, and calculate the detected value of the vehicle's basic resistance based on the real-time operating data; The data on the factors affecting the basic resistance of a vehicle are input into a neural network model, the neural network model is trained, and the predicted value of the basic resistance of the vehicle is output. The predicted value of the vehicle's basic resistance is compared with the detected value of the vehicle's basic resistance, and the neural network model is corrected based on the comparison result. The corrected neural network model outputs the corrected value of the vehicle's basic resistance.

2. The method for correcting the basic resistance of a rail vehicle according to claim 1, characterized in that, The factors influencing the basic resistance of the vehicle include environmental factors, and the data on the factors influencing the basic resistance of the vehicle include environmental data. The method for training the neural network model includes: The environmental data is divided into different environmental types based on the environmental type, and index ranges are defined for each environmental type. Collect real-time environmental data of the vehicle and input the real-time environmental data into the neural network model; The neural network model adjusts the weight of each environmental type in the prediction of the vehicle's basic resistance based on the index range to which the indicators of each type of real-time environmental data belong.

3. The method for correcting the basic resistance of a rail vehicle according to claim 2, characterized in that, The environment type includes at least altitude, temperature, humidity, and weather. The method for training the neural network model further includes: Altitude index ranges are divided based on altitude, temperature index ranges are divided based on annual average temperature, and humidity index ranges are divided based on annual average humidity. Weather data are divided into at least four weather index ranges: rain, snow, sunny, cloudy, and overcast. The altitude, average annual temperature, average annual humidity, and real-time weather of the area where the vehicle is located are collected and input into the neural network model. The neural network model adjusts the weight of each environmental type in the predicted value of the vehicle's basic resistance based on the altitude index range to which the vehicle's location belongs, the temperature index range to which the annual average temperature belongs, the humidity index range to which the annual average humidity belongs, and the weather index range to which the real-time weather belongs.

4. The method for correcting the basic resistance of a rail vehicle according to claim 3, characterized in that, Also includes: The neural network model adjusts the initial weights of each environmental type in the calculation of the vehicle's basic resistance based on the index range to which the indicators of each type of real-time environmental data belong.

5. The method for correcting the basic resistance of a rail vehicle according to claim 2, characterized in that, The factors influencing the basic resistance of the vehicle include quantitative factors and variable factors. The quantitative factors include at least the characteristic attributes of the vehicle. The environmental factors are included in the variable factors. The variable factors also include at least the vehicle's operating speed and the vehicle's load. The method for training the neural network model includes: The neural network model receives data on quantitative and variable factors, uses the detected value of the vehicle's basic resistance as a label, and adjusts the weights and biases of the neural network model so that the predicted value of the basic resistance calculated by the model is close to the detected value of the basic resistance, thereby training the model.

6. The method for correcting the basic resistance of a rail vehicle according to claim 2, characterized in that, The empirical formula for calculating the predicted value of a vehicle's unit basic resistance is as follows: , For the vehicle's operating speed, , , These are the coefficients of the empirical formula; The methods for modifying the neural network model include: Select an algorithm for refining the neural network model; The decision variables of the algorithm are set as coefficients of the empirical formula. , , The objective function is set as the minimum difference between the detected and predicted values ​​of the vehicle's basic resistance. The coefficients of the empirical formula are obtained by calculating the optimal solution of the objective function. , , Correction value; Based on the coefficients of the empirical formula , , The correction value is used to calculate the correction value for the vehicle's basic resistance.

7. The method for correcting the basic resistance of a rail vehicle according to claim 6, characterized in that, The objective function is: ,in, This is the measured value of the vehicle's basic resistance. This is the predicted value of the vehicle's basic resistance.

8. The method for correcting the basic resistance of a rail vehicle according to claim 2, characterized in that, The method of inputting the real-time environmental data into the neural network model includes: The real-time environmental data is normalized and converted into standardized real-time environmental data with a mean of 0 and a standard deviation of 1. The standardized real-time environmental data is then input into a neural network model.

9. The method for correcting the basic resistance of a rail vehicle according to any one of claims 1-8, characterized in that, Also includes: The data of factors affecting the basic resistance of the vehicle are divided into a test set according to a set ratio. The correction value of the basic resistance of the vehicle is output according to the test set, and the loss function between the detected value and the correction value of the basic resistance of the vehicle is calculated. The accuracy of the correction value of the basic resistance of the vehicle calculated by the neural network model is judged according to the magnitude of the calculated loss function value.

10. A system for implementing the method for correcting the basic resistance of a rail vehicle according to any one of claims 1-9, characterized in that, include: Data acquisition module: Used to collect real-time operating data of the vehicle; Calculation module: Calculates the detected value of the vehicle's basic resistance based on the vehicle's real-time operating data; Training module: Based on the data of factors affecting the basic resistance of a vehicle, the neural network model is trained, and the predicted value of the basic resistance of the vehicle is calculated through the trained neural network model; Correction module: compares the predicted value of the vehicle's basic resistance with the detected value of the vehicle's basic resistance, corrects the neural network model based on the comparison result, and calculates the corrected value of the vehicle's basic resistance using the corrected neural network model; Output module: Outputs the correction value for the vehicle's basic resistance.

Images