CN101419679A - Intelligent identification Method for running state of hybrid electric automobile - Google Patents

Abstract

Translated fromChinese

本发明公开了一种混合电动汽车行驶状态的智能识别方法，属于电动汽车控制技术领域。该方法主要采用神经网络进行识别，其工作过程主要分为学习期和工作期两个阶段。在学习期，首先采集汽车的标准行驶工况，并进行分段，再对每个小样本进行计算，得到一系列样本参数，然后代入神经网络计算公式得到智能识别控制所需的参数；在工作期，首先对采集得到的速度进行限幅滤波法处理，再进行递推平均滤波法处理，然后存储一定时间的速度并进行计算，再把计算结果代入神经网络计算公式便可以得到汽车当前的行驶状态。本发明对汽车行驶状态识别效果明显，因此可以帮助电动汽车控制器合理有效地调节控制参数，从而进一步提高电动汽车的燃油经济性和排放件。

The invention discloses an intelligent identification method for the driving state of a hybrid electric vehicle, belonging to the technical field of electric vehicle control. This method mainly adopts neural network for recognition, and its working process is mainly divided into two stages: learning period and working period. During the learning period, the standard driving conditions of the car are first collected and divided into sections, and then each small sample is calculated to obtain a series of sample parameters, which are then substituted into the neural network calculation formula to obtain the parameters required for intelligent identification and control; In the first period, the collected speed is firstly processed by the limiting filter method, and then processed by the recursive average filter method, and then the speed for a certain period of time is stored and calculated, and then the calculation result is substituted into the neural network calculation formula to obtain the current driving speed of the car. state. The invention has obvious recognition effect on the driving state of the vehicle, and thus can help the controller of the electric vehicle to adjust the control parameters reasonably and effectively, so as to further improve the fuel economy and emission parts of the electric vehicle.

Description

Translated fromChinese

一种混合动力电动汽车行驶状态的智能识别方法An intelligent identification method for the driving state of a hybrid electric vehicle

技术领域technical field

本发明涉及一种混合电动汽车的当前行驶状态的智能识别方法，属于电动汽车控制技术领域。The invention relates to an intelligent identification method for the current driving state of a hybrid electric vehicle, belonging to the technical field of electric vehicle control.

背景技术Background technique

电动汽车是我国汽车发展的一个重要方向，对混合动力电动汽车控制策略的研究已经成为汽车研究的热点之一，而汽车行驶时的能源分配问题又是电动汽车控制策略研究的重点。与传统发动机汽车相比，混合动力电动汽车的控制策略与汽车行驶状态紧密联系，而目前国内还没有对汽车行驶状态进行识别的设备和方法，因此混合动力电动汽车在燃油经济性和排放性能方面的优势没有得到最大体现。本发明通过对汽车的行驶车速进行处理和计算，从而有效地识别出了汽车的行驶状态，帮助电动汽车整车控制器更好的对汽车的能量进行分配。Electric vehicle is an important direction of automobile development in our country. The research on the control strategy of hybrid electric vehicle has become one of the hotspots in automobile research. Compared with traditional engine vehicles, the control strategy of hybrid electric vehicles is closely related to the driving state of the vehicle. At present, there is no equipment and method for identifying the driving state of the vehicle in China. Therefore, the fuel economy and emission performance of hybrid electric vehicles advantages are not fully realized. The invention processes and calculates the running speed of the car, thereby effectively identifying the running state of the car and helping the controller of the electric car to better distribute the energy of the car.

发明内容：Invention content:

本发明的目的是通过如下技术方案实现的：The purpose of the present invention is achieved through the following technical solutions:

由于目前没有对汽车行驶状态进行识别的方法和技术，在一定程度上影响了混合动力电动汽车的能量分配以及汽车燃油经济性和排放性能的提高，为了弥补这一不足，本发明提供一种智能识别混合动力电动汽车行驶状态的方法。通过采集汽车一段时间的行驶速度，然后采用神经网络方法对其进行计算和分析，从而识别出汽车当前的行驶状态，帮助汽车自身的整车控制系统调整参数，使得混合动力电动汽车的燃油经济性和排放性能得到更大的提高。Since there is no method and technology for identifying the driving state of the vehicle at present, it affects the energy distribution of the hybrid electric vehicle and the improvement of the fuel economy and emission performance of the vehicle to a certain extent. In order to make up for this deficiency, the invention provides an intelligent A method for identifying the driving state of a hybrid electric vehicle. By collecting the driving speed of the car for a period of time, and then using the neural network method to calculate and analyze it, so as to identify the current driving state of the car, help the car's own vehicle control system to adjust parameters, and make the fuel economy of hybrid electric vehicles and emissions performance are even more improved.

采用神经网络识别汽车行驶工况的工作过程主要分为两个阶段，首先是学习期，其主要是通过对汽车标准行驶工况进行学习，从而确定状态识别程序中的各项参数：然后是工作期，主要是采集汽车的行驶车速并进行处理，再通过状态识别程序计算得到汽车的当前行驶状态是属于哪种标准行驶工况。The working process of using the neural network to identify the driving conditions of the vehicle is mainly divided into two stages. The first is the learning period, which is mainly to determine the parameters in the state recognition program by learning the standard driving conditions of the vehicle: and then the working period. In the first stage, it mainly collects the driving speed of the car and processes it, and then calculates which standard driving condition the current driving state of the car belongs to through the state recognition program.

第一个阶段学习期，主要通过在PC机中用matlab编写程序来实现，其方法步骤为：The first stage of learning is mainly realized by writing a program with matlab in a PC, and the method steps are as follows:

步骤1：首先输入现有汽车主干道、快速路等国家或者地方标准行驶工况，然后分别对汽车主干道、快速路等的行驶工况进行分段，每一小段作为一个样本，而且每个样本的时间可以重叠。若设定每个小样本的采样时间为t_s秒，每隔t_f秒钟进行一次采样，t_s＝n_it_f，则其划分结果为0～t_s秒为一段工况，t_f～(t_s+t_f)秒为一段工况，2t_f～(2t_f+t_s)秒为一段工况，......，(n·t_f)～(n·t_f+t_s)秒为一段工况，并把每种工况分段后的结果进行存储，若主干道工况分为n1段，则把主干道工况的存储为数组v_z1，v_z2，......v_zn1；若快速路工况分为n2段，快速路工况存储为数组v_k1，v_k2，......，v_kn2；Step 1: First input the national or local standard driving conditions such as the existing automobile arterial roads and expressways, and then segment the driving conditions of the automobile arterial roads and expressways respectively. Each small section is used as a sample, and each The times of the samples can overlap. If the sampling time of each small sample is set as t_s seconds, sampling is performed every t_f seconds, t_s = n_i t_f , then the division result is 0～t_s seconds as a section of working conditions, t_f ～(t_s +t_f ) seconds is a working condition, 2t_f ～(2t_f +t_s ) seconds is a working condition, ..., (n·t_f )～(n·t_f + t_s ) seconds is a working condition, and store the results of each working condition segmented, if the main road working condition is divided into n1 segments, then store the working condition of the main road as an array v_z1 , v_z2 ,. ..... v_zn1 ; If the working condition of the expressway is divided into n2 sections, the working condition of the expressway is stored as an array v_k1 , v_k2 , ..., v_kn2 ;

步骤2：对每一小段工况数组进行处理，首先通过工况的速度计算加速度，当加速度为正时存储为数组a_z1，a_z2，......，a_zn1，a_k1，a_k2，……，a_kn2，……，当加速度为负时存储为数组r_z1，r_z2，……，r_zn1，r_k1，r_k2，……，r_kn2，……，以及每个采样点的速度和加速度的乘积存储为数组va_z1，va_z2，......，va_zn1，va_k1，va_k2，……，va_kn2，......。再对速度和正加速度的数组计算其数据的最大值、平均值以及标准差，对负加速度的数组计算其数据的最小值、平均值和标准差，对加速度和速度乘积的数组计算其数据的平均值。再对速度、正加速度、负加速度、速度和加速度的乘积的数组进行节点划分和数据统计，并计算每部分占时间的百分比；Step 2: Process each small section of the working condition array, first calculate the acceleration through the speed of the working condition, and store it as an array a_z1 , a_z2 ,..., a_zn1 , a_k1 , a when the acceleration is positive_k2 ,...,a_kn2 ,..., stored as array r_z1 ,r_z2 ,...,r_zn1 ,r_k1 ,r_k2 ,...,r_kn2 ,..., when the acceleration is negative, and each sample The product of the velocity and acceleration of the point is stored as an array va_z1 , va_z2 , . . . , va_zn1 , va_k1, va_k2 , . Then calculate the maximum value, average value and standard deviation of the data for the array of velocity and positive acceleration, calculate the minimum value, average value and standard deviation of the data for the array of negative acceleration, and calculate the average of the data for the array of acceleration and velocity products value. Then perform node division and data statistics on the array of speed, positive acceleration, negative acceleration, product of speed and acceleration, and calculate the percentage of time for each part;

步骤3：采用上述步骤2中计算得到的参数，作为神经网络学习的训练样本x₁、x₂……x_k，即输入层；采用y₁代表主干路工况，y₂代表快速路工况，……，y_m代表第m中工况，作为神经网络的输出层；然后采用公式(1)进行计算，求得神经网络各连接权上的权值w_im、b_m。Step 3: Use the parameters calculated in the above step 2 as the training samples x₁ , x₂ ... x_k learned by the neural network, that is, the input layer; use y₁ to represent the working condition of the trunk road, and y₂ to represent the working condition of the expressway , ..., y_m represents the working condition in the mth, as the output layer of the neural network; then use the formula (1) to calculate the weights w_im and b_m of each connection weight of the neural network.

${\mathrm{<span><span>y</span>the\; y</span>}}_{\mathrm{<span><span>m</span>m</span>}}<span><span>=</span>=</span>\underset{\mathrm{<span><span>i</span>i</span>}<span><span>=</span>=</span>\mathrm{<span><span>1</span>1</span>}}{\overset{\mathrm{<span><span>k</span>k</span>}}{\mathrm{<span><span>\&Sigma;</span>\&Sigma;</span>}}}{\mathrm{<span><span>w</span>w</span>}}_{\mathrm{<span><span>im</span>im</span>}}{\mathrm{<span><span>x</span>x</span>}}_{\mathrm{<span><span>i</span>i</span>}}<span><span>-</span>-</span>{\mathrm{<span><span>b</span>b</span>}}_{\mathrm{<span><span>m</span>m</span>}}<span><span>-</span>-</span><span><span>-</span>-</span><span><span>-</span>-</span><span><span>(</span>(</span>\mathrm{<span><span>1</span>1</span>}<span><span>)</span>)</span>$

第二个阶段工作期，主要通过在PC机中采用C编写行驶状态智能识别程序，然后下载到智能识别控制器中实现，其方法步骤为：The second stage of work is mainly realized by using C to write the driving state intelligent recognition program in the PC, and then downloading it to the intelligent recognition controller. The method steps are:

步骤4：首先对速度进行限幅滤波法处理：当前采集到车速与前一次求得的车速进行对比，若变化量大于v_c，则取前一次求得的车速，若变化小于v_c，则当前采集到车速；然后对速度进行递推平均滤波法处理：若在第一阶段中速度的采样频率为t_f秒钟一次，则取前n_c次的车速的平均值作为当前车速；Step 4: First, the speed is processed by the limiting filter method: compare the current collected vehicle speed with the previously obtained vehicle speed, if the change is greater than v_c , take the previously obtained vehicle speed, if the change is less than v_c , then The current vehicle speed is collected; then the speed is processed by the recursive average filtering method: if the sampling frequency of the speed in the first stage is once every t_f seconds, then the average value of the previous n_c times of vehicle speed is taken as the current vehicle speed;

步骤5：存储一定时间的车速，形成车速数组。若在第一阶段中选用t_s秒的样本进行神经网络权值计算，则在汽车行驶过程中在每次采样后存储前t_s秒内的车速，形成一个具有n_i个速度值的数组作为一个速度片断；Step 5: Store the vehicle speed for a certain period of time to form a vehicle speed array. If in the first stage, samples of t_s seconds are selected for neural network weight calculation, the vehicle speed in the previous t_s seconds is stored after each sampling in the driving process of the car, and an array with n_i speed values is formed as a velocity segment;

步骤6：对上述速度片断按照第一阶段中的步骤2进行计算，得到神经网络计算所需的各项参数；Step 6: Calculate the above-mentioned speed fragment according to step 2 in the first stage, and obtain various parameters required for neural network calculation;

步骤7：然后采用公式(1)进行神经网络计算，即得到当前汽车行驶状态y_m。Step 7: Then use the formula (1) to perform neural network calculation, that is, to obtain the current vehicle driving state y_m .

本发明的有益效果是：对汽车行驶状态识别效果明显，可以帮助电动汽车控制器合理有效地调节控制参数，从而进一步提高电动汽车的燃油经济性和排放性能。The beneficial effect of the invention is that: the recognition effect on the driving state of the vehicle is obvious, and it can help the controller of the electric vehicle to adjust the control parameters reasonably and effectively, so as to further improve the fuel economy and emission performance of the electric vehicle.

附图说明：Description of drawings:

图1为本发明采用神经网络实现智能识别流程图；Fig. 1 is that the present invention adopts neural network to realize intelligent recognition flowchart;

图2为某城市主干道行驶工况；Figure 2 shows the driving conditions of a city's main road;

图3为某城市快速路行驶工况；Figure 3 shows the driving conditions of an urban expressway;

图4为本发明采用本智能识别方法的识别效果图。Fig. 4 is a recognition effect diagram of the present invention using the intelligent recognition method.

具体实施方式：Detailed ways:

下面结合附图对本发明作进一步描述如下：Below in conjunction with accompanying drawing, the present invention is further described as follows:

如图1所示，采用神经网络进行智能识别需要进行两个阶段，第一个阶段是学习期，主要是计算智能识别程序所需的控制参数，第二个阶段就是编写智能识别程序，然后下载到控制器中就可以实时的对汽车的行驶状态进行识别。下面我们将结合实例对上述两个阶段分别进行详细的阐述：As shown in Figure 1, the use of neural networks for intelligent recognition requires two stages. The first stage is the learning period, which is mainly to calculate the control parameters required by the intelligent recognition program. The second stage is to write the intelligent recognition program and then download it. In the controller, the driving state of the car can be identified in real time. In the following, we will elaborate on the above two stages in detail with examples:

第一个阶段，学习期如图1的上半部所示：The first phase, the learning period, is shown in the upper half of Figure 1:

步骤1：采集汽车主干路和快速路标准行驶工况。然后对汽车行驶工况进行分段，得到小速度片段，即小样本。在本次实例分析中，每个小样本的采样时间为t_s选用180秒，t_f选用1秒钟，因此n_i＝180，t_s＝n_it_f。分段结果为：0～180为一个样本小段，1～181为一个样本小段，依次类推。并把每种工况分段后的结果进行存储，主干道工况分为800段，存储为数组v_z1，v_z2，......v₈₀₀；快速路工况分为64段，存储为数组v_k1，v_k2，......，v₆₄；Step 1: Collect standard driving conditions of automobile trunk roads and expressways. Then segment the driving conditions of the vehicle to obtain small speed segments, that is, small samples. In this example analysis, the sampling time of each small sample is 180 seconds for t_s and 1 second for t_f , so n_i =180, t_s =n_itf . The segmentation result is: 0~180 is a sample segment, 1~181 is a sample segment, and so on. And store the results after each working condition is segmented, the working condition of the main road is divided into 800 sections, stored as an array v_z1 , v_z2 , ... v₈₀₀ ; the working condition of the expressway is divided into 64 sections, stored as arrays v_k1 , v_k2 , . . . , v₆₄ ;

步骤2：计算各个速度片段的参数。首先通过工况的速度计算加速度，当加速度为正时存储为数组a_z1，a_z2，......，a_zn1，a_k1，a_k2，……，a_kn2，当加速度为负时存储为数组r_z1，r_z2，……，r_zn1，r_k1，r_k2，……，r_kn2，以及每个采样点的速度和加速度的乘积存储为数组va_z1，va_z2，......，va_zn1，va_k1，va_k2，……，va_kn2。再对速度和正加速度的数组计算其数据的最大值、平均值以及标准差，对负加速度的数组计算其数据的最小值、平均值和标准差，对速度和加速度乘积的数组计算其数据的平均值。再对速度、正加速度、负加速度、速度和加速度的乘积的数组进行节点划分和数据统计，并计算每部分占时间的百分比，例如：若对速度划分节点为4个，即1m/s、5m/s、10m/s、15m/s，则计算速度小于等于1m/s的采样点数占采样总点数的比例、速度大于1小于等于5m/s的点数占总点数的比例、速度大于5小于等于10m/s点数占总点数的比例、速度大于10小于等于15m/s的点数占总点数的比例、速度大于15m/s的点数占总点数的比例；若对正加速度划分节点为1个，即7m/s²，则计算加速度大于0小于等于7m/s²的点数占总点数的比例、加速度大于7m/s²的点数占总点数的比例；若对负加速度划分节点为1个，即—7m/s²，则计算减速度大于等于-7小于0m/s²的点数占总点数的比例、减速度小于-7m/s²的点数占总点数的比例；若对速度和加速度的乘积节点为3个，即3m²/s³、6m²/s³、10m²/s³，则计算速度和加速度的乘积小于0m²/s³的点数占总点数的比例、速度和加速度的乘积大于3小于等于6m²/s³的点数占总点数的比例、速度和加速度的乘积大于6小于等于10m²/s³的点数占总点数的比例、速度和加速度的乘积大于10m²/s³的点数占总点数的比例。最终主干道工况计算得到一个800*24的矩阵，快速路工况计算得到一个64*24的矩阵；Step 2: Calculate the parameters of each velocity segment. First, the acceleration is calculated through the speed of the working condition. When the acceleration is positive, it is stored as an array a_z1 , a_z2 , ..., a_zn1 , a_k1 , a_k2 , ..., a_kn2 , and when the acceleration is negative Stored as an array r_z1 , r_z2 , ..., r_zn1 , r_k1 , r_k2 , ..., r_kn2 , and the product of velocity and acceleration at each sampling point is stored as an array va_z1 , va_z2 , ... ..., va_zn1 , va_k1 , va_k2 , ..., va_kn2 . Then calculate the maximum value, average value, and standard deviation of the data for the array of speed and positive acceleration, calculate the minimum value, average value, and standard deviation of the data for the array of negative acceleration, and calculate the average of the data for the array of speed and acceleration products value. Then divide the array of speed, positive acceleration, negative acceleration, product of speed and acceleration into node division and data statistics, and calculate the percentage of time for each part, for example: if the speed is divided into 4 nodes, that is, 1m/s, 5m /s, 10m/s, 15m/s, calculate the proportion of sampling points whose speed is less than or equal to 1m/s to the total number of sampling points, the proportion of points whose speed is greater than 1 and less than or equal to 5m/s to the total number of points, and the proportion of points whose speed is greater than 5 and less than or equal to The ratio of 10m/s points to the total number of points, the ratio of points with a speed greater than 10 and less than or equal to 15m/s to the total number of points, the ratio of points with a speed greater than 15m/s to the total number of points; if the positive acceleration is divided into 1 node, that is 7m/s² , then calculate the proportion of points whose acceleration is greater than 0 and less than or equal to 7m/s² to the total number of points, and the proportion of points whose acceleration is greater than 7m/s² to the total number of points; if the negative acceleration is divided into one node, that is— 7m/s² , then calculate the proportion of points whose deceleration is greater than or equal to -7 and less than 0m/s² to the total number of points, and the proportion of points whose deceleration is less than -7m/s² to the total number of points; if the product node of velocity and acceleration is 3, that is, 3m² /s³ , 6m² /s³ , and 10m² /s³ , then the ratio of the number of points where the product of velocity and acceleration is less than 0m² /s³ to the total number of points, and the product of velocity and acceleration is greater than 3 The ratio of points less than or equal to 6m² /s³ to the total number of points, and the product of velocity and acceleration greater than 6 The ratio of points less than or equal to 10m² /s³ to the total number of points, and the product of velocity and acceleration greater than 10m² /s³ Points as a percentage of total points. Finally, an 800*24 matrix is calculated for the main road condition, and a 64*24 matrix is obtained for the expressway condition;

步骤3：计算神经网络识别中的参数。把步骤2中对主干道和快速路得到的两个矩阵作为输入层；然后定义主干道y₁为1，快速路y₂为0，作为输出层。通过公式(1)计算得到，w_im为矩阵[-12.214；-12.269；-12.183；6.0995；2.3818；2.3416；-13.32；-13.214；-13.924；5.8685；5.7457；-0.06623；6.0199；2.8603；1.3746；4.3112；3.712；2.9215；-5.1756；-11.405；-10.7；-3.1999；-2.5333；-3.1603]，b_m＝7.5095；Step 3: Calculate the parameters in neural network recognition. Take the two matrices obtained for the main road and express road in step 2 as the input layer; then define the main road y₁ as 1 and the express road y₂ as 0 as the output layer. Calculated by formula (1),_wim is a matrix [-12.214; -12.269; -12.183; 6.0995; 2.3818; 2.3416; -13.32; 4.3112; 3.712; 2.9215; -5.1756; -11.405; -10.7; -3.1999; -2.5333; -3.1603], b_m =7.5095;

第二个阶段，工作期如图1的下半部所示：The second stage, the working period is shown in the lower half of Figure 1:

步骤4：对采集到的车速进行滤波处理。首先对速度进行限幅滤波法处理：若v_c选用20km/h，与前一次求得的车速进行对比，若变化量大于20km/h，则取前一次求得的车速，若变化小于20km/h，则取当前采集到的车速；然后对速度进行递推平均滤波法处理：若n_c选用10，则取前10次的车速的平均值作为当前车速；Step 4: Filter the collected vehicle speed. Firstly, the speed is processed by the limiting filter method: if v_c is 20km/h, compare it with the previous vehicle speed, if the change is greater than 20km/h, take the previous vehicle speed, if the change is less than 20km/h h, then take the currently collected vehicle speed; then process the speed with the recursive average filtering method: if n_c is selected as 10, then take the average value of the previous 10 vehicle speeds as the current vehicle speed;

步骤5：存储前t_i即180秒内的车速得到一个速度数组；Step 5: Store the vehicle speed within 180 seconds before t_i to obtain a speed array;

步骤6：对上述速度组计算表1中数值，得到一个矩阵；Step 6: Calculate the values in Table 1 for the above speed group to obtain a matrix;

步骤7：把得到的矩阵代入神经网络计算公式(1)便可得到当前行驶状态。Step 7: Substituting the obtained matrix into the neural network calculation formula (1) to obtain the current driving state.

如图2所示，图中纵坐标为速度，横坐标为时间，曲线是中国某城市主干道的标准行驶工况。As shown in Figure 2, the ordinate in the figure is speed, the abscissa is time, and the curve is the standard driving condition of a main road in a certain city in China.

如图3所示，图中纵坐标为速度，横坐标为时间，曲线是中国某城市快速路的标准行驶工况。As shown in Figure 3, the ordinate in the figure is speed, the abscissa is time, and the curve is the standard driving condition of an expressway in a certain city in China.

如图4所示，图中左侧纵坐标为速度，右侧纵坐标为行驶状态，横坐标表示时间，图中实线为某型号电动汽车在某城市实际行驶速度曲线，图中虚线为识别出的行驶状态结果，在0～2132秒和2602～3035秒的上识别结果为1，表示在主干道上行驶，在其他时间段上识别结果为0，表示在快速路上行驶，与实际行驶状况基本吻合，识别效果明显。As shown in Figure 4, the ordinate on the left side of the figure is the speed, the ordinate on the right side is the driving state, and the abscissa represents the time. The solid line in the figure is the actual driving speed curve of a certain type of electric vehicle in a certain city, and the dotted line in the figure is the identification The result of the driving status is 1 at 0-2132 seconds and 2602-3035 seconds, which means driving on the main road, and 0 at other time periods, which means driving on the express road, which is different from the actual driving conditions. Basically match, the recognition effect is obvious.

表1：神经网络计算中样本计算所需参数Table 1: Parameters required for sample calculation in neural network calculation

 序号参数符号意义单位1v_avg平均车速km/h2v_std车速标准差km/h3v_max最高车速km/h4a_avg平均加速度m/s²5a_std加速度标准差m/s²6a_max最大加速度m/s²7r_avg平均减速度m/s²8r_std减速度标准差m/s²9r_min最小减速度m/s²10V_0_1速度小于等于1m/s的采样点数占采样总点数的比例％11v_1_5速度大于1小于等于5m/s的点数占总点数的比例％12v_5_10速度大于5小于等于10m/s点数占总点数的比例％13v_10_15速度大于10小于等于15m/s的点数占总点数的比例％14v_15_25速度大于15m/s的点数占总点数的比例％15a_0_7加速度大于0小于等于7m/s²的点数占总点数的比例％16a_7加速度大于7m/s²的点数占总点数的比例％17r_0_7减速度大于等于-7小于0m/s²的点数占总点数的比例％18r_7减速度小于-7m/s²的点数占总点数的比例％29va_0va小于0m²/s³的点数占总点数的比例％20va_0_3va大于等于0小于等于3m²/s³的点数占总点数的比例％21va_3_6va大于3小于等于6m²/s³的点数占总点数的比例％22va_6_10va大于6小于等于10m²/s³的点数占总点数的比例％23va_10va大于10m²/s³的点数占总点数的比例％24va_avgVa的平均值m²/s³ serial number parameter symbol significance unit 1 v_avg average speed km/h 2 v_std speed standard deviation km/h 3 v_max maximum speed km/h 4 a_avg average acceleration m/s² 5 a_std Acceleration standard deviation m/s² 6 a_max maximum acceleration m/s² 7 r_avg average deceleration m/s² 8 r_std deceleration standard deviation m/s² 9 r_min Minimum deceleration m/s² 10 V_0_1 The proportion of sampling points whose speed is less than or equal to 1m/s to the total number of sampling points % 11 v_1_5 The proportion of points whose speed is greater than 1 and less than or equal to 5m/s to the total points % 12 v_5_10 Proportion of points whose speed is greater than 5 and less than or equal to 10m/s to the total points % 13 v_10_15 The proportion of points whose speed is greater than 10 and less than or equal to 15m/s to the total points % 14 v_15_25 The proportion of points whose speed is greater than 15m/s to the total points % 15 a_0_7 Proportion of points whose acceleration is greater than 0 and less than or equal to 7m/s² to the total number of points % 16 a_7 The proportion of points whose acceleration is greater than 7m/s² to the total points % 17 r_0_7 The proportion of points whose deceleration is greater than or equal to -7 and less than 0m/s² to the total points % 18 r_7 The proportion of points whose deceleration is less than -7m/s² to the total points % 29 va_0 The proportion of points with va less than 0m² /s³ to the total points % 20 va_0_3 The ratio of the number of points with va greater than or equal to 0 and less than or equal to 3m² /s³ to the total number of points % twenty one va_3_6 The proportion of points with va greater than 3 and less than or equal to 6m² /s³ in the total points % twenty two va_6_10 The proportion of points with va greater than 6 and less than or equal to 10m² /s³ in the total points % twenty three va_10 The proportion of points with va greater than 10m² /s³ to the total points % twenty four va_avg The average value of Va m² /s³

说明：va：代表加速度和速度的乘积，单位为：m²/s³。Explanation: va: stands for the product of acceleration and velocity, the unit is: m² /s³ .

Claims (5)

1, a kind of intelligent identification Method of running state of hybrid electric automobile is characterized in that: adopt neural network identification automobile running working condition, its course of work mainly is divided into two stages:

First stage is the learning period, mainly by realizing with the matlab coding that in PC its method step is:

Step 1: at first input has the country or the provincial standard driving cycle of automobile trunk road, through street etc. now, respectively the driving cycle of automobile trunk road, through street etc. is carried out segmentation then, each segment is as a sample, and the time of each sample can be overlapping; If setting the sampling time of each small sample is t_sSecond, every t_fOnce sample second, t_s=n_it_f, then its division result is 0～t_sBe one section operating mode second, t_f～(t_f+ t_s) second be one section operating mode, 2t_f～(2t_f+ t_s) second be one section operating mode ..., (nt_f)～(nt_f+ t_s) be one section operating mode, and the result after every kind of operating mode segmentation is stored second; If the major trunk roads operating mode is divided into the n1 section, then the major trunk roads operating mode is stored as array v_Z1, v_Z2... v_Zn1If the through street operating mode is divided into the n2 section, the through street operating mode is stored as array v_K1, v_K2..., v_Kn2

Step 2: each segment operating mode array is handled, at first passed through the speed calculation acceleration of operating mode, when acceleration is that timing is stored as array a_Z1, a_Z2..., a_Zn1, a_K1, a_K2..., a_Kn2..., when acceleration is stored as array r when negative_Z1, r_Z2..., r_Zn1, r_K1, r_K2..., r_Kn2..., and the product of the speed of each sampled point and acceleration is stored as array va_Z1, va_Z2..., vaz_N1, va_K1, va_K2..., va_Kn2..., again the array of speed and positive acceleration is calculated maximal value, mean value and the standard deviation of its data, the array of negative acceleration is calculated minimum value, mean value and the standard deviation of its data, the array of acceleration and speed product is calculated the mean value of its data; Again the array of the product of speed, positive acceleration, negative acceleration, speed and acceleration is carried out node division and data statistics, and calculate the number percent that every part accounts for T.T.;

Step 3: adopt the parameter that calculates in the above-mentioned steps 2, as the training sample x of neural network learning₁, x₂X_k, i.e. input layer; Adopt y₁Represent the major trunk roads operating mode, y₂Represent the through street operating mode ..., y_mRepresent m kind operating mode, as the output layer of neural network; Adopt formula (1) to calculate then, try to achieve the weight w on each connection weight of neural network_Im, b_m

$y_m=\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}{w}_{\mathrm{im}}{x}_i-b_m---\left(1\right)$

Second stage is under-stream period, mainly by adopting C to write the running status intelligent recognizer in PC, downloads to then in the Intelligent Recognition controller and realizes, its method step is:

Step 4: at first speed is carried out the limit filtration method and handle: currently collect the speed of a motor vehicle with the preceding speed of a motor vehicle of once trying to achieve compares, as if variable quantity greater than v_c, the speed of a motor vehicle that once calculates before then getting is if change less than v_c, then get the current speed of a motor vehicle that collects; Then speed being carried out the recurrence average filter method handles: as if the sample frequency at the phase one medium velocity is t_fOnce then get preceding n second_cThe mean value of the inferior speed of a motor vehicle is as current vehicle speed;

Step 5: the speed of a motor vehicle of storage certain hour forms speed of a motor vehicle array; If in the phase one, select t for use_sThe sample of second carries out neural network weight and calculates, t before then storing after each sampling in vehicle traveling process_sThe speed of a motor vehicle in second forms one and has n_iThe array of individual velocity amplitude is as a speed segment;

Step 6: above-mentioned speed segment is calculated according to the step 2 in the phase one, obtained the parameters in the table 1;

Step 7: adopt formula (1) to carry out neural network then and calculate, promptly obtain current motoring condition y_m

2, the method for a kind of Intelligent Recognition running state of hybrid electric automobile according to claim 1 is characterized in that: when adopting neural network to calculate the motoring condition parameter designing, the driving cycle of automobile is divided the time t of division_sBe 120～300 seconds, interval time t_fBe 0.5～3 second, and the time of dividing can be overlapping.

3, the intelligent identification Method of a kind of running state of hybrid electric automobile according to claim 1, it is characterized in that: when adopting neural network to calculate the motoring condition parameter, in to time segment computation process, product to speed, acceleration, speed and acceleration has carried out node division, and dividing node is 1～10.

4, the intelligent identification Method of a kind of running state of hybrid electric automobile according to claim 1, it is characterized in that: when adopting neural network to calculate the motoring condition parameter, selected average velocity for use, the standard deviation of the speed of a motor vehicle, max. speed, average acceleration, the standard deviation of acceleration, peak acceleration, average retardation rate, the minimum deceleration degree, the standard deviation of retarded velocity, the mean value of the product of speed and acceleration and to the speed of a motor vehicle, acceleration, after the product of speed and acceleration carried out node division, each several part accounted for the input of the ratio of T.T. as neural network learning.

5, the intelligent identification Method of a kind of running state of hybrid electric automobile according to claim 1, it is characterized in that: when adopting the transport condition of neural network identification automobile, at the variation characteristic of the speed of a motor vehicle in the reality, at first the speed of a motor vehicle has been carried out the limit filtration method and handled its variable quantity v_cBe 18～30km/h, again the speed of a motor vehicle carried out the recurrence average filter method then and handled, select the speed number n for use_cBe 10～25.

Images