CN104183134A - Expressway short-time traffic flow forecast method based on intelligent car type classification - Google Patents

Abstract

Translated fromChinese

本发明公开了一种基于智能分车型的高速公路短时交通流量预测方法，包括如下步骤：步骤一、获取数据；步骤二、数据预处理；步骤三、对经过预处理后得到的总车流量数据序列进行平稳性检验，若总车流量数据序列为平稳数据序列，则采用时间序列法预测交通流量；若总车流量数据序列为非平稳数据序列，则分两类车型进行预测：对小型车和集装箱型车采用时间序列法进行交通流量预测；对中型车和大型车采用二次指数平滑法进行交通流量预测；步骤四、按照不同车型的车辆折算系数将预测结果折算成标准车辆，计算得到总车流量预测值；步骤五、等待至下一次数据更新时，执行步骤一。本发明能够提高预测精度，帮助交通管理者或出行者更好地把握路况。

The invention discloses a method for predicting short-term traffic flow of expressways based on intelligent classification of vehicle types, comprising the following steps: step 1, acquiring data; step 2, data preprocessing; step 3, calculating the total traffic flow obtained after preprocessing The data sequence is tested for stationarity. If the total traffic flow data sequence is a stationary data sequence, the time series method is used to predict the traffic flow; if the total traffic flow data sequence is a non-stationary data sequence, two types of vehicle types are used for prediction: The time series method is used to predict the traffic flow of container-type vehicles; the quadratic exponential smoothing method is used to predict the traffic flow of medium-sized vehicles and large vehicles; step 4, according to the vehicle conversion coefficients of different models, the prediction results are converted into standard vehicles, and the calculation is Predicted value of total traffic flow; step 5, wait until the next data update, and then execute step 1. The invention can improve prediction accuracy and help traffic managers or travelers to better grasp road conditions.

Description

Translated fromChinese

基于智能分车型的高速公路短时交通流量预测方法Short-term traffic flow forecasting method for expressway based on intelligent vehicle classification

技术领域technical field

本发明属于道路车辆的交通控制系统技术领域，具体的为一种基于智能分车型的高速公路短时交通流量预测方法。The invention belongs to the technical field of traffic control systems for road vehicles, in particular to a short-term traffic flow prediction method for expressways based on intelligent vehicle classification.

背景技术Background technique

随着高速公路里程的增加，交通管理者对交通的信息化管理要求以及出行者对交通的信息化服务需求也在逐渐增加，实时准确的把握未来路段的交通状况是高效管理与服务的前提，而对短时交通流量的预测是对路段未来交通情况把握的关键。如何准确把握交通流的规律，提高流量预测的精度，是现有研究的重要方向。With the increase of highway mileage, traffic managers' requirements for traffic information management and travelers' demand for traffic information services are gradually increasing. Real-time and accurate grasp of traffic conditions in future road sections is the premise of efficient management and service. The prediction of short-term traffic flow is the key to grasp the future traffic conditions of road sections. How to accurately grasp the law of traffic flow and improve the accuracy of flow forecast is an important direction of existing research.

现有的高速公路短时交通流量的预测方法主要涉及基于线性系统理论的方法、基于非线性理论的方法、基于知识发现的智能预测方法、基于组合预测的方法以及基于交通流模态的预测方法等理论基础。基于这些理论基础，目前已研发出了多种高速公路短时交通流量的预测方法：Existing prediction methods for short-term traffic flow on expressways mainly involve methods based on linear system theory, methods based on nonlinear theory, intelligent prediction methods based on knowledge discovery, methods based on combination prediction, and prediction methods based on traffic flow modes. and other theoretical basis. Based on these theoretical foundations, a variety of prediction methods for short-term traffic flow on expressways have been developed:

(1)五邑大学学报(自然科学版)(第18卷第3期，2004年9月)公开了一种交通流的时间序列建模及预测方法，其采用时间序列模型对短时交通流量进行了预测，实际数据验证结果表明该模型能够较好地拟合交通流时间序列，并可获得较高的预测精度；(1) Journal of Wuyi University (Natural Science Edition) (Volume 18, No. 3, September 2004) discloses a time series modeling and forecasting method of traffic flow, which uses a time series model to analyze short-term traffic flow The prediction is made, and the actual data verification results show that the model can better fit the time series of traffic flow and obtain higher prediction accuracy;

(2)交通运输工程与信息学报(第8卷第1期，2010年3月)公开了一种基于混沌时间序列的道路断面短时交通流预测模型，其针对道路断面的短时交通流量序列的混沌特性进行了分析，实测交通流量数据验证结果表明了该预测模型在一定程度上具有效性；(2) Journal of Traffic and Transportation Engineering and Information (Volume 8, Issue 1, March 2010) discloses a short-term traffic flow prediction model for road sections based on chaotic time series, which aims at the short-term traffic flow sequence of road sections The chaotic characteristics of the model are analyzed, and the verification results of the measured traffic flow data show that the prediction model is effective to a certain extent;

(3)系统工程理论与实践(第30卷第2期，2010年2月)公开了一种基于K近邻非参数回归的短时交通流预测方法，并将利用K值构造的预测区间用于特殊路况的预测中，也得到了明显的改进效果；(3) System Engineering Theory and Practice (Volume 30, No. 2, February 2010) discloses a short-term traffic flow prediction method based on K-nearest neighbor non-parametric regression, and uses the prediction interval constructed by K value for In the prediction of special road conditions, a significant improvement has also been obtained;

(4)吉林大学学报(工学版)(第40卷第5期，2010年9月)公开了一种基于小波分析的交通参数组合预测方法，该方法更加准确地预测交通参数的变化趋势，具有普适性，且比传统的基于小波分析的组合预测过程简单，为大运算量的实时应用提供了可能；(4) Journal of Jilin University (Engineering Edition) (Volume 40, No. 5, September 2010) discloses a traffic parameter combination prediction method based on wavelet analysis, which can more accurately predict the change trend of traffic parameters and has Universal, and simpler than the traditional combined forecasting process based on wavelet analysis, it provides the possibility for real-time applications with a large amount of calculation;

(5)北京交通大学董春娇等，针对城市快速路自由流状态、拥挤流状态和阻塞流状态下交通流参数的时间和空间分布特性，提出了一种多状态下城市快速路网交通流短时预测理论与方法研究，其研究了混合状态下城市快速路交通流短时预测，实验表明，相比较传统的预测方法具有更高的预测精度。(5) Dong Chunjiao of Beijing Jiaotong University, etc., aiming at the time and space distribution characteristics of traffic flow parameters in the free flow state, congested flow state and blocked flow state of urban expressways, proposed a short-term traffic flow analysis method for urban expressway networks in multiple states. Forecasting theory and method research, which studies the short-term forecasting of urban expressway traffic flow under mixed conditions. Experiments show that compared with traditional forecasting methods, it has higher forecasting accuracy.

纵观以上各种交通流量的预测方法，为了提高预测的效果，需要更好的挖掘交通状态的变化趋势，把握交通流内部的规律性，而上述交通流量预测方法均是以总交通流量为研究对象来分析预测交通流的变化的。Looking at the above various traffic flow forecasting methods, in order to improve the forecasting effect, it is necessary to better explore the changing trend of traffic conditions and grasp the internal regularity of traffic flow, and the above traffic flow forecasting methods are all based on the total traffic flow objects to analyze and predict changes in traffic flow.

发明内容Contents of the invention

通过对微波车检器采集的分车型流量数据分析发现：高速公路上的小型车与集装箱型车每天的流量均比较低，且一天中的流量波动较平稳；中型车每天的流量变化呈现出两个明显的波峰，但在早晚波动均比较平稳，而大型车一天的流量曲线中除了早、晚高峰分别存在一段上升与下降趋势外，其余的时间段流量变化相对比较平稳。由此可见，不同车型流量的波动变化规律是不同的，且各自车型一天中的流量变化也是随着时间的变化而改变的，相比较总车流量而言，每个车型的交通流量变化的规律性更为明显。Through the analysis of the vehicle-type flow data collected by the microwave vehicle detector, it is found that: the daily flow of small cars and container-type vehicles on the expressway is relatively low, and the flow fluctuations in a day are relatively stable; the daily flow changes of medium-sized vehicles show two There is an obvious peak, but the fluctuations are relatively stable in the morning and evening. In the flow curve of large vehicles in a day, except for the morning and evening peaks, there is a rising and falling trend respectively, and the flow changes in the rest of the time period are relatively stable. It can be seen that the fluctuations of the traffic flow of different models are different, and the flow changes of the respective models in a day also change with time. Compared with the total traffic flow, the traffic flow change law of each model is Sex is more obvious.

有鉴于此，本发明的目的在于提供一种基于智能分车型的高速公路短时交通流量预测方法，能够根据实时交通流量数据的特点，调整预测模型，从而达到智能分车型预测，提高预测精度。In view of this, the object of the present invention is to provide a short-term traffic flow prediction method for expressways based on intelligent classification of vehicle types, which can adjust the prediction model according to the characteristics of real-time traffic flow data, thereby achieving intelligent classification of vehicle types and improving prediction accuracy.

为达到上述目的，本发明提供如下技术方案：To achieve the above object, the present invention provides the following technical solutions:

一种基于智能分车型的高速公路短时交通流量预测方法，其特征在于：包括如下步骤：A kind of expressway short-term traffic flow prediction method based on intelligent vehicle classification, it is characterized in that: comprise the steps:

步骤一、获取数据：获取高速公路的总车流量数据序列、分车型流量数据序列、平均车速和平均占有率；Step 1. Obtain data: obtain the total traffic flow data sequence, vehicle type flow data sequence, average vehicle speed and average occupancy rate of the expressway;

步骤二、数据预处理：剔除不符合交通实际情况的数据；Step 2. Data preprocessing: Eliminate data that does not conform to the actual traffic situation;

步骤三、对经过预处理后得到的总车流量数据序列进行平稳性检验，若总车流量数据序列为平稳数据序列，则采用时间序列法预测交通流量；若总车流量数据序列为非平稳数据序列，则将车型分为两类，第一类车型包括小型车和集装箱型车两种车型，且该类车型的交通流量预测方法采用时间序列法进行预测；第二类车型包括中型车和大型车两种车型，且该类车型的交通流量预测方法采用二次指数平滑法进行预测；Step 3. Perform a stationarity test on the total traffic flow data sequence obtained after preprocessing. If the total traffic flow data sequence is a stationary data sequence, use the time series method to predict traffic flow; if the total traffic flow data sequence is non-stationary data The vehicle types are divided into two categories. The first type includes small cars and container vehicles, and the traffic flow prediction method of this type of vehicles is predicted by the time series method; the second type includes medium-sized cars and large-scale vehicles. There are two types of vehicles, and the traffic flow forecast method of this type of vehicle is predicted by the double exponential smoothing method;

其中，采用时间序列法预测交通流量包括以下步骤：Among them, using the time series method to predict traffic flow includes the following steps:

(1)对获取的分车型流量数据序列进行平稳性判断，若对应的分车型流量数据序列平稳，则直接执行下一步；若对应的分车型流量数据序列不平稳，则对该分车型流量数据序列进行差分处理，得到新的平稳的分车型流量数据序列后，再执行下一步；(1) To judge the stationarity of the acquired flow data series by vehicle type, if the corresponding flow data series by vehicle type is stable, then directly execute the next step; The sequence is differentially processed to obtain a new stable flow data sequence by vehicle type, and then execute the next step;

(2)采用ARIMA模型作为预测模型进行预测，且对于经过差分处理后得到的新的平稳的分车型流量数据序列，还需对其预测结果进行反变换，转换为对应车型的流量预测值；(2) The ARIMA model is used as the prediction model for forecasting, and for the new smooth flow data series by vehicle type obtained after differential processing, it is necessary to inversely transform the prediction results and convert them into the flow forecast value of the corresponding vehicle type;

采用二次指数平滑法预测交通流量包括以下步骤：Forecasting traffic flow using the quadratic exponential smoothing method includes the following steps:

(1)对获取的分车型流量数据序列进行一次累加，并在累加后的曲线上通过可决系数的计算选取组成线性度最好的K个近邻序列；(1) Carry out an accumulation of the acquired flow rate data series by vehicle type, and select the K nearest neighbor sequences with the best linearity through the calculation of the coefficient of determination on the accumulated curve;

(2)通过实验法确定与预测标准误差最小的平滑系数α为最佳平滑系数；(2) Determine the smoothing coefficient α with the minimum standard error of prediction through experiments as the best smoothing coefficient;

(3)根据二次指数平滑法预测模型进行预测，并通过转换得到对应车型的流量预测值；(3) Predict according to the prediction model of the quadratic exponential smoothing method, and obtain the flow prediction value of the corresponding vehicle type through conversion;

步骤四、按照不同车型的车辆折算系数将预测结果折算成标准车辆，计算得到总车流量预测值；Step 4: Convert the predicted results into standard vehicles according to the vehicle conversion coefficients of different models, and calculate the predicted value of the total traffic flow;

步骤五、等待至下一次数据更新时，执行步骤一。Step 5. Wait until the next data update, and then execute Step 1.

进一步，所述步骤二中，分别采用阈值理论与交通流理论剔除不符合交通实际情况的数据；Further, in said step 2, threshold value theory and traffic flow theory are adopted respectively to remove data that do not conform to actual traffic conditions;

所述阈值理论为：在一个数据更新周期内，设定总车流量数据的阀值范围为[0,Q_max]，平均车速的阀值范围为[0,V_max]；若采集到的总车流量数据或平均车速的数据不在对应的阀值范围内时，则表明该组数据不可靠，并将其剔除；若采集到的总车流量数据和平均车数的数据均落在对应的阀值范围内时，则表明该组数据可靠，保留该组数据；其中，Q_max、V_max分别表示在一个数据更新周期内的流量最大值和速度最大值；The threshold theory is: within a data update cycle, set the threshold range of the total traffic flow data to [0, Q_max ], and the threshold range of the average vehicle speed to [0, V_max ]; if the collected total When the traffic flow data or average vehicle speed data are not within the corresponding threshold range, it indicates that the group of data is unreliable and will be eliminated; if the collected total traffic flow data and average vehicle number data fall within the corresponding threshold When it is within the value range, it indicates that the set of data is reliable and the set of data is retained; among them, Q_max and V_max respectively represent the maximum flow rate and maximum speed within a data update cycle;

所述交通流理论为：首先，根据交通流理论建立错误数据判断规则，即剔除规则；然后，判断采集的数据序列是否满足剔除规则；当满足剔除规则时，将对应的数据需剔除；当不满足剔除规则时，保留对应的数据。The traffic flow theory is as follows: firstly, according to the traffic flow theory, establish the error data judgment rule, that is, the elimination rule; then, judge whether the collected data sequence satisfies the elimination rule; when the elimination rule is satisfied, the corresponding data needs to be eliminated; When the elimination rules are met, the corresponding data is retained.

进一步，所述步骤三中，总车流量数据序列的平稳性检验方法为：Further, in the step 3, the stationarity test method of the total traffic flow data sequence is:

(1)获取经过预处理的总车流量数据序列X_t，并延迟k得到X_t+k，计算其各自均值μ_t、μ_t+k；(1) Obtain the preprocessed total traffic flow data sequence X_t , delay k to obtain X_t+k , and calculate their respective mean values μ_t and μ_t+k ;

(2)根据自相关函数公式计算其自相关函数R(k)：(2) Calculate its autocorrelation function R(k) according to the autocorrelation function formula:

$\mathrm{<span><span>R</span>R</span>}<span><span>(</span>(</span>\mathrm{<span><span>k</span>k</span>}<span><span>)</span>)</span><span><span>=</span>=</span>\frac{\mathrm{<span><span>E</span>E.</span>}<span><span>[</span>[</span><span><span>(</span>(</span>{\mathrm{<span><span>X</span>x</span>}}_{\mathrm{<span><span>t</span>t</span>}}<span><span>-</span>-</span>{\mathrm{<span><span>\&mu;</span>\&mu;</span>}}_{\mathrm{<span><span>t</span>t</span>}}<span><span>)</span>)</span><span><span>(</span>(</span>{\mathrm{<span><span>X</span>x</span>}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>k</span>k</span>}}<span><span>-</span>-</span>{\mathrm{<span><span>\&mu;</span>\&mu;</span>}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>k</span>k</span>}}<span><span>)</span>)</span><span><span>]</span>]</span>}{{\mathrm{<span><span>\&sigma;</span>\&sigma;</span>}}^{\mathrm{<span><span>2</span>2</span>}}}$

其中，σ²为方差，k为滞后期；Among them,^σ2 is the variance, and k is the lag period;

(3)当自相关函数R(k)不能快速衰减趋近于0或在0附近波动，则总车流量数据序列属于非平稳序列；当自相关函数R(k)能够快速衰减到0，则总车流量数据序列属于平稳序列。(3) When the autocorrelation function R(k) cannot quickly decay to 0 or fluctuate around 0, the total traffic flow data sequence is a non-stationary sequence; when the autocorrelation function R(k) can quickly decay to 0, then The total traffic flow data series is a stationary series.

进一步，所述步骤三中，所述ARIMA模型的公式如下：Further, in the step 3, the formula of the ARIMA model is as follows:

其中，x_t为分车型流量数据序列，ε_t为白噪声，B为延迟算子(B^jx_t＝x_t-j)，d为差分次数；p为模型的阶次；Among them, x_t is the flow data sequence by vehicle type, ε_t is white noise, B is the delay operator (B^j x_t = x_tj ), d is the order of difference; p is the order of the model;

利用相关矩估计法及BIC准则进行模型参数的估计及模型阶次p的确定，具体公式如下：The correlation moment estimation method and the BIC criterion are used to estimate the model parameters and determine the model order p. The specific formula is as follows:

其中，和分别表示分车型流量数据序列的自相关函数和模型参数；p表示模型的阶次，在计算自相关函数时，一般取为其中N表示样本量。in, and Respectively represent the autocorrelation function and model parameters of the vehicle type flow data sequence; p represents the order of the model, when calculating the autocorrelation function, it is generally taken as where N is the sample size.

进一步，所述步骤三中，二次指数平滑法预测模型的公式如下：Further, in the third step, the formula of the quadratic exponential smoothing prediction model is as follows:

${\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}}<span><span>=</span>=</span>{\mathrm{<span><span>a</span>a</span>}}_{\mathrm{<span><span>t</span>t</span>}}<span><span>+</span>+</span>{\mathrm{<span><span>b</span>b</span>}}_{\mathrm{<span><span>t</span>t</span>}}\mathrm{<span><span>T</span>T</span>}$

其中：为第t+T期预测值；in: is the predicted value of period t+T;

a_t和b_t分别为模型参数，且$a_t=2S_t^1-S_t^2,$$b_t=\frac{\mathrm{\&alpha;}}{1-\mathrm{\&alpha;}}(S_t^1-S_t^2);$a_t and b_t are model parameters respectively, and$a_t=2S_t^1-S_t^2,$$b_t=\frac{\mathrm{\&alpha;}}{1-\mathrm{\&alpha;}}(S_t^1-S_t^2);$

为第t期的一次指数平滑值，且 is the exponential smoothing value of period t, and

为第t期的二次指数平滑值，且 is the quadratic exponential smoothing value of period t, and

α为平滑系数，X_t-1为初始值；α is the smoothing coefficient, and X_t-1 is the initial value;

流量预测值的转换公式为：traffic forecast The conversion formula for is:

${\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>q</span>q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}}<span><span>=</span>=</span>{\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}}<span><span>-</span>-</span>{\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}<span><span>-</span>-</span>\mathrm{<span><span>1</span>1</span>}}$

其中，和为利用二次指数平滑法对最优K近邻序列t+T-1时刻和t+T时刻的流量预测值。in, and It is the flow forecast value of the optimal K-nearest neighbor sequence at time t+T-1 and time t+T by using the quadratic exponential smoothing method.

进一步，所述步骤四中，t时刻的折算后的总车流量预测值为：Further, in said step 4, the total traffic flow forecast value after conversion at time t for:

$\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>=</span>=</span>\mathrm{<span><span>1</span>1</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>1</span>1</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>+</span>+</span>\mathrm{<span><span>1.5</span>1.5</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>2</span>2</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>+</span>+</span>\mathrm{<span><span>2</span>2</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>3</span>3</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>+</span>+</span>\mathrm{<span><span>3</span>3</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>4</span>4</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span>$

其中，q₁(t)、q₂(t)、q₃(t)、q₄(t)分别为t时刻小型车、中型车、大型车、集装箱型车的流量预测值。Among them, q₁ (t), q₂ (t), q₃ (t), and q₄ (t) are the flow forecast values of small vehicles, medium vehicles, large vehicles, and container vehicles at time t, respectively.

本发明的有益效果在于：The beneficial effects of the present invention are:

本发明基于智能分车型的高速公路短时交通流量预测方法，从实时交通流量变化特点和一天中不同车型的流量变化特点的角度出发，通过实时的判断交通流量序列的平稳性，确定是否需要进行分车型预测；该方法即把握了实时的流量特点，也考虑了整体的流量变化趋势和不同车型车流量的变化特点，相比较不考虑车型的预测方法，其更能表现出交通流内部的规律性，能够获得更高的预测精度。The present invention is based on the short-term traffic flow prediction method of the expressway based on the intelligent vehicle classification, starting from the real-time traffic flow change characteristics and the flow change characteristics of different vehicle types in a day, and determining whether it is necessary to carry out Prediction by vehicle type; this method not only grasps the real-time traffic characteristics, but also considers the overall flow change trend and the change characteristics of different vehicle types. Compared with the prediction method that does not consider the vehicle type, it can better show the internal laws of traffic flow can obtain higher prediction accuracy.

附图说明Description of drawings

为了使本发明的目的、技术方案和有益效果更加清楚，本发明提供如下附图进行说明：In order to make the purpose, technical scheme and beneficial effect of the present invention clearer, the present invention provides the following drawings for illustration:

图1为本发明基于智能分车型的高速公路短时交通流量预测方法实施例的流程图；Fig. 1 is the flow chart of the embodiment of the expressway short-term traffic flow forecasting method based on intelligent vehicle classification in the present invention;

图2为本实施例对总车流量数据序列进行平稳性检验的流程图；Fig. 2 is the flow chart that the present embodiment carries out stationarity test to total traffic flow data series;

图3为本实施例采用时间序列法预测交通流量的流程图；Fig. 3 adopts the flow chart of time series method to predict traffic flow for the present embodiment;

图4为本实施例采用二次指数平滑法预测交通流量的流程图。FIG. 4 is a flow chart of forecasting traffic flow using the double exponential smoothing method in this embodiment.

具体实施方式Detailed ways

下面结合附图和具体实施例对本发明作进一步说明，以使本领域的技术人员可以更好的理解本发明并能予以实施，但所举实施例不作为对本发明的限定。The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, so that those skilled in the art can better understand the present invention and implement it, but the examples given are not intended to limit the present invention.

如图1所示，为本发明基于智能分车型的高速公路短时交通流量预测方法实施例的流程图。本实施例基于智能分车型的高速公路短时交通流量预测方法，其特征在于：包括如下步骤：As shown in FIG. 1 , it is a flow chart of an embodiment of the method for predicting short-term traffic flow of expressways based on intelligent vehicle classification according to the present invention. The present embodiment is based on the expressway short-term traffic flow forecasting method of intelligent vehicle classification, is characterized in that: comprise the following steps:

步骤一、获取数据：读取经高速公路微波车检器采集得到的总车流量数据序列、分车型流量数据序列、平均车速和平均占有率，本实施例将车型分为四种，分别为小型车、中型车、大型车和集装箱型车，具体的，本实施例高速公路交通数据字段定义如表1所示：Step 1. Obtain data: read the total traffic flow data sequence, vehicle type flow data sequence, average vehicle speed and average occupancy rate collected by the expressway microwave vehicle detector. In this embodiment, the vehicle types are divided into four types, namely small Cars, medium-sized cars, large cars and container cars, specifically, the definition of the expressway traffic data field in this embodiment is as shown in Table 1:

表1：高速公路交通数据字段定义表Table 1: Freeway traffic data field definition table

序列的平稳性不仅受到数据本身的影响，还受到样本量大小的影响。随着数据序列样本量的增加，序列的平稳性往往随之下降，而且样本量过大会增加序列的自相关系数和偏相关系数的计算量；又由于交通流量变化的不确定性和随机性，样本量太小的流量序列不仅难以通过平稳性检验，而且也无法反映交通流的变化趋势，进行预测时就会产生较大的误差。通过多次实验，本实施例的流量数据序列的样本量选取25个。The stationarity of the series is not only affected by the data itself, but also by the size of the sample. As the sample size of the data sequence increases, the stationarity of the sequence tends to decrease, and the calculation of the autocorrelation coefficient and partial correlation coefficient of the sequence will increase if the sample size is too large; and due to the uncertainty and randomness of traffic flow changes, A traffic sequence with too small a sample size is not only difficult to pass the stationarity test, but also cannot reflect the changing trend of traffic flow, resulting in large errors in forecasting. Through multiple experiments, 25 samples are selected for the traffic data sequence in this embodiment.

步骤二、数据预处理：剔除不符合交通实际情况的数据；本实施例分别采用阈值理论与交通流理论剔除不符合交通实际情况的数据；Step 2, data preprocessing: remove the data that does not conform to the actual traffic situation; this embodiment adopts the threshold value theory and the traffic flow theory to remove the data that does not conform to the actual traffic situation;

阈值理论为：读取距当前时刻a×m分钟，长度为m的数据序列，这里m取值与步骤1中的样本量一样为25，a为一个数据更新周期的时间间隔；设定在一个数据更新周期内，总车流量数据的阀值范围为[0,Q_max]，平均车速的阀值范围为[0,V_max]；若采集到的总车流量数据或平均车速的数据不在对应的阀值范围内时，则表明该组数据不可靠，并将其剔除；若采集到的总车流量数据和平均车数的数据均落在对应的阀值范围内时，则表明该组数据可靠，保留该组数据；其中，Q_max、V_max分别表示在一个数据更新周期内的流量最大值和速度最大值，本实施例Q_max取值为300辆，V_max取值为150km/h。The threshold theory is as follows: read a data sequence of a×m minutes from the current moment and a length of m, where the value of m is the same as the sample size in step 1, which is 25, and a is the time interval of a data update cycle; set at a In the data update cycle, the threshold range of the total traffic flow data is [0, Q_max ], and the threshold value range of the average vehicle speed is [0, V_max ]; if the collected total traffic flow data or average vehicle speed data is not corresponding If it is within the threshold range, it indicates that the group of data is unreliable, and it will be removed; Reliable, keep this set of data; wherein, Q_max and V_max respectively represent the maximum flow rate and the maximum speed within a data update cycle. In this embodiment, the value of Q_max is 300 vehicles, and the value of V_max is 150km/h .

所述交通流理论为：首先，根据交通流理论建立错误数据判断规则，即剔除规则；然后，判断采集的数据序列是否满足剔除规则；当满足剔除规则时，将对应的数据需剔除；当不满足剔除规则时，保留对应的数据。本实施例交通流理论错误数据判断规则如表2所示。The traffic flow theory is as follows: firstly, according to the traffic flow theory, establish the error data judgment rule, that is, the elimination rule; then, judge whether the collected data sequence satisfies the elimination rule; when the elimination rule is satisfied, the corresponding data needs to be eliminated; When the elimination rules are met, the corresponding data is retained. The error data judgment rules of the traffic flow theory in this embodiment are shown in Table 2.

表2：基于交通流理论的错误数据判别规则Table 2: Discrimination rules for wrong data based on traffic flow theory

步骤三、对经过预处理后得到的总车流量数据序列进行平稳性检验，平稳时间序列是指均值、方差和自回归函数不随时间变化的时间序列。当时间序列{x_t}为平稳随机过程时，对于任意一个时段1≤t₁＜t₂...＜t_m和的联合分布等同于的联合分布；由定义可知，平稳性等价于：所有x_t都具有相同的分布；在整个时期内，任何两个相邻项之间的相关程度都相同；用数学表达式为：Step 3. Perform a stationarity test on the preprocessed total traffic flow data series. A stationary time series refers to a time series in which the mean, variance and autoregressive function do not change with time. When the time series {x_t } is a stationary random process, for any period 1≤t₁ <t₂ ...<t_m and The joint distribution of is equivalent to The joint distribution of ; by definition, stationarity is equivalent to: all x_t have the same distribution; in the whole period, the degree of correlation between any two adjacent items is the same; the mathematical expression is:

(1)对任意t，均值恒为常数：Ex_t＝μ(与t无关的常数)。(1) For any t, the mean is always a constant: Ex_t = μ (constant independent of t).

(2)方差Var(x_t)＝σ²(与t无关的有限常数)。(2) Variance Var(x_t )=σ² (finite constant independent of t).

(3)对任意整数t和k，自相关函数r_t,t+k只与k有关，r_t,t+k＝r_k。(3) For any integer t and k, the autocorrelation function r_t,t+k is only related to k, r_t,t+k = r_k .

本实施例的总车流量数据序列的平稳性检验方法为：The stationarity test method of the total traffic flow data sequence of the present embodiment is:

(1)获取经过预处理的总车流量数据序列X_t，并延迟k得到X_t+k，计算其各自均值μ_t、μ_t+k；(1) Obtain the preprocessed total traffic flow data sequence X_t , delay k to obtain X_t+k , and calculate their respective mean values μ_t and μ_t+k ;

(2)根据自相关函数公式计算其自相关函数R(k)：(2) Calculate its autocorrelation function R(k) according to the autocorrelation function formula:

$\mathrm{<span><span>R</span>R</span>}<span><span>(</span>(</span>\mathrm{<span><span>k</span>k</span>}<span><span>)</span>)</span><span><span>=</span>=</span>\frac{\mathrm{<span><span>E</span>E.</span>}<span><span>[</span>[</span><span><span>(</span>(</span>{\mathrm{<span><span>X</span>x</span>}}_{\mathrm{<span><span>t</span>t</span>}}<span><span>-</span>-</span>{\mathrm{<span><span>\&mu;</span>\&mu;</span>}}_{\mathrm{<span><span>t</span>t</span>}}<span><span>)</span>)</span><span><span>(</span>(</span>{\mathrm{<span><span>X</span>x</span>}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>k</span>k</span>}}<span><span>-</span>-</span>{\mathrm{<span><span>\&mu;</span>\&mu;</span>}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>k</span>k</span>}}<span><span>)</span>)</span><span><span>]</span>]</span>}{{\mathrm{<span><span>\&sigma;</span>\&sigma;</span>}}^{\mathrm{<span><span>2</span>2</span>}}}$

其中，σ²为方差，k为滞后期；Among them,^σ2 is the variance, and k is the lag period;

(3)当自相关函数R(k)不能快速衰减趋近于0或在0附近波动，则总车流量数据序列属于非平稳序列；当自相关函数R(k)能够快速衰减到0，则总车流量数据序列属于平稳序列。(3) When the autocorrelation function R(k) cannot quickly decay to 0 or fluctuate around 0, the total traffic flow data sequence is a non-stationary sequence; when the autocorrelation function R(k) can quickly decay to 0, then The total traffic flow data series is a stationary series.

若总车流量数据序列为平稳数据序列，则采用时间序列法预测交通流量；若总车流量数据序列为非平稳数据序列，则将车型分为两类，第一类车型包括小型车和集装箱型车两种车型，且该类车型的交通流量预测方法采用时间序列法进行预测；第二类车型包括中型车和大型车两种车型，且该类车型的交通流量预测方法采用二次指数平滑法进行预测；If the total traffic flow data sequence is a stationary data sequence, the time series method is used to predict the traffic flow; if the total traffic flow data sequence is a non-stationary data sequence, the vehicle types are divided into two categories. The first category includes small cars and container vehicles. There are two types of vehicles, and the traffic flow forecast method of this type of vehicle is forecasted by the time series method; the second type of vehicle type includes two types of medium-sized cars and large cars, and the traffic flow forecast method of this type of vehicles uses the quadratic exponential smoothing method make predictions;

其中，采用时间序列法预测交通流量包括以下步骤：Among them, using the time series method to predict traffic flow includes the following steps:

(1)对获取的分车型流量数据序列进行平稳性判断，若对应的分车型流量数据序列平稳，则直接执行下一步；若对应的分车型流量数据序列不平稳，则对该分车型流量数据序列进行差分处理，得到新的平稳的分车型流量数据序列和差分次数d后，再执行下一步；(1) To judge the stationarity of the acquired flow data series by vehicle type, if the corresponding flow data series by vehicle type is stable, then directly execute the next step; The sequence is differentially processed to obtain a new stable vehicle-type flow data sequence and the number of differences d, and then execute the next step;

(2)采用ARIMA模型作为预测模型进行预测，且对于经过差分处理后得到的新的平稳的分车型流量数据序列，还需对其预测结果进行反变换，转换为对应车型的流量预测值。(2) The ARIMA model is used as the prediction model for prediction, and for the new smooth flow data series by vehicle type obtained after differential processing, it is necessary to inversely transform the prediction results and convert them into the flow forecast value of the corresponding vehicle type.

具体的，ARIMA模型的公式如下：Specifically, the formula of the ARIMA model is as follows:

其中，x_t为分车型流量数据序列，ε_t为白噪声，B为延迟算子(B^jx_t＝x_t-j)，d为差分次数；p为模型的阶次，为模型参数；Among them, x_t is the flow data sequence by vehicle type, ε_t is white noise, B is the delay operator (B^j x_t = x_tj ), d is the order of difference; p is the order of the model, is the model parameter;

利用相关矩估计法及BIC准则进行模型参数的估计及模型阶次p的确定，具体公式如下：The correlation moment estimation method and the BIC criterion are used to estimate the model parameters and determine the model order p. The specific formula is as follows:

其中，和分别表示分车型流量数据序列的自相关函数和模型参数；p表示模型的阶次，在计算自相关函数时，一般取为其中N表示样本量，即本实施例的模型阶次p＝5。in, and Respectively represent the autocorrelation function and model parameters of the vehicle type flow data sequence; p represents the order of the model, when calculating the autocorrelation function, it is generally taken as Where N represents the sample size, that is, the model order of this embodiment is p=5.

采用二次指数平滑法预测交通流量包括以下步骤：Forecasting traffic flow using the quadratic exponential smoothing method includes the following steps:

(1)对获取的分车型流量数据序列进行一次累加，并在累加后的曲线上通过可决系数的计算选取组成线性度最好的K个近邻序列，本实施例中，K＝25；(1) Carry out once accumulation to the obtained sub-vehicle flow data sequence, and select the K nearest neighbor sequences that form the best linearity by calculating the coefficient of determination on the curve after accumulation, in the present embodiment, K=25;

(2)通过实验法确定与预测标准误差最小的平滑系数α为最佳平滑系数，其中，α∈[0.6,1]；(2) Determine the smoothing coefficient α with the minimum standard error of prediction through experiments as the best smoothing coefficient, where α∈[0.6,1];

(3)取最优K个近邻序列前3个值的平均值作为初始值，结合最佳平滑系数α，根据二次指数平滑法预测模型进行预测，并通过转换得到对应车型的流量预测值。(3) Take the average value of the first three values of the optimal K neighbor sequences as the initial value, combined with the best smoothing coefficient α, predict according to the prediction model of the quadratic exponential smoothing method, and obtain the flow prediction value of the corresponding vehicle type through conversion.

具体的，二次指数平滑法预测模型的公式如下：Specifically, the formula of the prediction model of the double exponential smoothing method is as follows:

${\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}}<span><span>=</span>=</span>{\mathrm{<span><span>a</span>a</span>}}_{\mathrm{<span><span>t</span>t</span>}}<span><span>+</span>+</span>{\mathrm{<span><span>b</span>b</span>}}_{\mathrm{<span><span>t</span>t</span>}}\mathrm{<span><span>T</span>T</span>}$

其中：为第t+T期预测值；in: is the predicted value of period t+T;

a_t和b_t分别为模型参数，且$a_t=2S_t^1-S_t^2,$$b_t=\frac{\mathrm{\&alpha;}}{1-\mathrm{\&alpha;}}(S_t^1-S_t^2);$a_t and b_t are model parameters respectively, and$a_t=2S_t^1-S_t^2,$$b_t=\frac{\mathrm{\&alpha;}}{1-\mathrm{\&alpha;}}(S_t^1-S_t^2);$

为第t期的一次指数平滑值，且 is the exponential smoothing value of period t, and

为第t期的二次指数平滑值，且 is the quadratic exponential smoothing value of period t, and

α为平滑系数，X_t-1为初始值；α is the smoothing coefficient, and X_t-1 is the initial value;

流量预测值的转换公式为：traffic forecast The conversion formula for is:

${\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>q</span>q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}}<span><span>=</span>=</span>{\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}}<span><span>-</span>-</span>{\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}<span><span>-</span>-</span>\mathrm{<span><span>1</span>1</span>}}$

其中，和为利用二次指数平滑法对最优K近邻序列t+T-1时刻和t+T时刻的流量预测值。in, and It is the flow forecast value of the optimal K-nearest neighbor sequence at time t+T-1 and time t+T by using the quadratic exponential smoothing method.

步骤四、按照不同车型的车辆折算系数将预测结果折算成标准车辆，计算得到总车流量预测值。具体的，t时刻的折算后的总车流量预测值为：Step 4: Convert the predicted results into standard vehicles according to the vehicle conversion coefficients of different models, and calculate the predicted value of the total traffic flow. Specifically, the converted total traffic flow forecast value at time t is for:

$\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>=</span>=</span>\mathrm{<span><span>1</span>1</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>1</span>1</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>+</span>+</span>\mathrm{<span><span>1.5</span>1.5</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>2</span>2</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>+</span>+</span>\mathrm{<span><span>2</span>2</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>3</span>3</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>+</span>+</span>\mathrm{<span><span>3</span>3</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>4</span>4</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span>$

其中，q₁(t)、q₂(t)、q₃(t)、q₄(t)分别为t时刻小型车、中型车、大型车、集装箱型车的流量预测值。Among them, q₁ (t), q₂ (t), q₃ (t), and q₄ (t) are the flow forecast values of small vehicles, medium vehicles, large vehicles, and container vehicles at time t, respectively.

步骤五、等待至下一次数据更新时，执行步骤一。Step 5. Wait until the next data update, and then execute Step 1.

本实施例基于智能分车型的高速公路短时交通流量预测方法，从实时交通流量变化特点和一天中不同车型的流量变化特点的角度出发，通过实时的判断交通流量序列的平稳性，确定是否需要进行分车型预测；该方法即把握了实时的流量特点，也考虑了整体的流量变化趋势和不同车型车流量的变化特点，相比较不考虑车型的预测方法，其更能表现出交通流内部的规律性，能够获得更高的预测精度。This embodiment is based on the short-term traffic flow prediction method of the expressway based on the intelligent classification of vehicle types. From the perspective of the characteristics of real-time traffic flow changes and the flow characteristics of different types of vehicles in a day, by judging the stationarity of the traffic flow sequence in real time, it is determined whether it is necessary to Forecast by vehicle type; this method not only grasps the real-time traffic characteristics, but also considers the overall flow change trend and the change characteristics of different vehicle types. Regularity can obtain higher prediction accuracy.

以上所述实施例仅是为充分说明本发明而所举的较佳的实施例，本发明的保护范围不限于此。本技术领域的技术人员在本发明基础上所作的等同替代或变换，均在本发明的保护范围之内。本发明的保护范围以权利要求书为准。The above-mentioned embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or transformations made by those skilled in the art on the basis of the present invention are all within the protection scope of the present invention. The protection scope of the present invention shall be determined by the claims.

Claims (6)

Translated fromChinese

1.一种基于智能分车型的高速公路短时交通流量预测方法，其特征在于：包括如下步骤：1. a kind of expressway short-term traffic flow prediction method based on intelligence sub-vehicle, it is characterized in that: comprise the steps:步骤一、获取数据：获取高速公路的总车流量数据序列、分车型流量数据序列、平均车速和平均占有率；Step 1. Obtain data: obtain the total traffic flow data sequence, vehicle type flow data sequence, average vehicle speed and average occupancy rate of the expressway;步骤二、数据预处理：剔除不符合交通实际情况的数据；Step 2. Data preprocessing: Eliminate data that does not conform to the actual traffic situation;步骤三、对经过预处理后得到的总车流量数据序列进行平稳性检验，若总车流量数据序列为平稳数据序列，则采用时间序列法预测交通流量；若总车流量数据序列为非平稳数据序列，则将车型分为两类，第一类车型包括小型车和集装箱型车两种车型，且该类车型的交通流量预测方法采用时间序列法进行预测；第二类车型包括中型车和大型车两种车型，且该类车型的交通流量预测方法采用二次指数平滑法进行预测；Step 3. Perform a stationarity test on the total traffic flow data sequence obtained after preprocessing. If the total traffic flow data sequence is a stationary data sequence, use the time series method to predict traffic flow; if the total traffic flow data sequence is non-stationary data The vehicle types are divided into two categories. The first type includes small cars and container vehicles, and the traffic flow prediction method of this type of vehicles is predicted by the time series method; the second type includes medium-sized cars and large-scale vehicles. There are two types of vehicles, and the traffic flow forecast method of this type of vehicle is predicted by the double exponential smoothing method;其中，采用时间序列法预测交通流量包括以下步骤：Among them, using the time series method to predict traffic flow includes the following steps:(1)对获取的分车型流量数据序列进行平稳性判断，若对应的分车型流量数据序列平稳，则直接执行下一步；若对应的分车型流量数据序列不平稳，则对该分车型流量数据序列进行差分处理，得到新的平稳的分车型流量数据序列后，再执行下一步；(1) To judge the stationarity of the acquired flow data series by vehicle type, if the corresponding flow data series by vehicle type is stable, then directly execute the next step; The sequence is differentially processed to obtain a new stable flow data sequence by vehicle type, and then execute the next step;(2)采用ARIMA模型作为预测模型进行预测，且对于经过差分处理后得到的新的平稳的分车型流量数据序列，还需对其预测结果进行反变换，转换为对应车型的流量预测值；(2) The ARIMA model is used as the prediction model for forecasting, and for the new smooth flow data series by vehicle type obtained after differential processing, it is necessary to inversely transform the prediction results and convert them into the flow forecast value of the corresponding vehicle type;采用二次指数平滑法预测交通流量包括以下步骤：Forecasting traffic flow using the quadratic exponential smoothing method includes the following steps:(1)对获取的分车型流量数据序列进行一次累加，并在累加后的曲线上通过可决系数的计算选取组成线性度最好的K个近邻序列；(1) Carry out an accumulation of the acquired flow rate data series by vehicle type, and select the K nearest neighbor sequences with the best linearity through the calculation of the coefficient of determination on the accumulated curve;(2)通过实验法确定与预测标准误差最小的平滑系数α为最佳平滑系数；(2) Determine the smoothing coefficient α with the minimum standard error of prediction through experiments as the best smoothing coefficient;(3)根据二次指数平滑法预测模型进行预测，并通过转换得到对应车型的流量预测值；(3) Predict according to the prediction model of the quadratic exponential smoothing method, and obtain the flow prediction value of the corresponding vehicle type through conversion;步骤四、按照不同车型的车辆折算系数将预测结果折算成标准车辆，计算得到总车流量预测值；Step 4: Convert the predicted results into standard vehicles according to the vehicle conversion coefficients of different models, and calculate the predicted value of the total traffic flow;步骤五、等待至下一次数据更新时，执行步骤一。Step 5. Wait until the next data update, and then execute Step 1.2.根据权利要求1所述基于智能分车型的高速公路短时交通流量预测方法，其特征在于：所述步骤二中，分别采用阈值理论与交通流理论剔除不符合交通实际情况的数据；2. according to the described expressway short-term traffic flow prediction method based on intelligence sub-vehicle type of claim 1, it is characterized in that: in described step 2, adopt threshold value theory and traffic flow theory to reject the data that does not conform to traffic actual situation respectively;所述阈值理论为：在一个数据更新周期内，设定总车流量数据的阀值范围为[0,Q_max]，平均车速的阀值范围为[0,V_max]；若采集到的总车流量数据或平均车速的数据不在对应的阀值范围内时，则表明该组数据不可靠，并将其剔除；若采集到的总车流量数据和平均车数的数据均落在对应的阀值范围内时，则表明该组数据可靠，保留该组数据；其中，Q_max、V_max分别表示在一个数据更新周期内的流量最大值和速度最大值；The threshold theory is: within a data update cycle, set the threshold range of the total traffic flow data to [0, Q_max ], and the threshold range of the average vehicle speed to [0, V_max ]; if the collected total When the traffic flow data or average vehicle speed data are not within the corresponding threshold range, it indicates that the group of data is unreliable and will be eliminated; if the collected total traffic flow data and average vehicle number data fall within the corresponding threshold When it is within the value range, it indicates that the set of data is reliable and the set of data is retained; among them, Q_max and V_max respectively represent the maximum flow rate and maximum speed within a data update cycle;所述交通流理论为：首先，根据交通流理论建立错误数据判断规则，即剔除规则；然后，判断采集的数据序列是否满足剔除规则；当满足剔除规则时，将对应的数据需剔除；当不满足剔除规则时，保留对应的数据。The traffic flow theory is as follows: firstly, according to the traffic flow theory, establish the error data judgment rule, that is, the elimination rule; then, judge whether the collected data sequence satisfies the elimination rule; when the elimination rule is satisfied, the corresponding data needs to be eliminated; When the elimination rules are met, the corresponding data is retained.3.根据权利要求1所述基于智能分车型的高速公路短时交通流量预测方法，其特征在于：所述步骤三中，总车流量数据序列的平稳性检验方法为：3. according to the described expressway short-term traffic flow forecasting method based on intelligence sub-model of claim 1, it is characterized in that: in described step 3, the stationarity checking method of total traffic flow data sequence is:(1)获取经过预处理的总车流量数据序列X_t，并延迟k得到X_t+k，计算其各自均值μ_t、μ_t+k；(1) Obtain the preprocessed total traffic flow data sequence X_t , delay k to obtain X_t+k , and calculate their respective mean values μ_t and μ_t+k ;(2)根据自相关函数公式计算其自相关函数R(k)：(2) Calculate its autocorrelation function R(k) according to the autocorrelation function formula:$\mathrm{<span><span>R</span>R</span>}<span><span>(</span>(</span>\mathrm{<span><span>k</span>k</span>}<span><span>)</span>)</span><span><span>=</span>=</span>\frac{\mathrm{<span><span>E</span>E.</span>}<span><span>[</span>[</span><span><span>(</span>(</span>{\mathrm{<span><span>X</span>x</span>}}_{\mathrm{<span><span>t</span>t</span>}}<span><span>-</span>-</span>{\mathrm{<span><span>\&mu;</span>\&mu;</span>}}_{\mathrm{<span><span>t</span>t</span>}}<span><span>)</span>)</span><span><span>(</span>(</span>{\mathrm{<span><span>X</span>x</span>}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>k</span>k</span>}}<span><span>-</span>-</span>{\mathrm{<span><span>\&mu;</span>\&mu;</span>}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>k</span>k</span>}}<span><span>)</span>)</span><span><span>]</span>]</span>}{{\mathrm{<span><span>\&sigma;</span>\&sigma;</span>}}^{\mathrm{<span><span>2</span>2</span>}}}$其中，σ²为方差，k为滞后期；Among them,^σ2 is the variance, and k is the lag period;(3)当自相关函数R(k)不能快速衰减趋近于0或在0附近波动，则总车流量数据序列属于非平稳序列；当自相关函数R(k)能够快速衰减到0，则总车流量数据序列属于平稳序列。(3) When the autocorrelation function R(k) cannot quickly decay to 0 or fluctuate around 0, the total traffic flow data sequence is a non-stationary sequence; when the autocorrelation function R(k) can quickly decay to 0, then The total traffic flow data series is a stationary series.4.根据权利要求1所述基于智能分车型的高速公路短时交通流量预测方法，其特征在于：所述步骤三中，所述ARIMA模型的公式如下：4. according to the described expressway short-term traffic flow forecasting method based on intelligent sub-model of claim 1, it is characterized in that: in described step 3, the formula of described ARIMA model is as follows:其中，x_t为分车型流量数据序列，ε_t为白噪声，B为延迟算子(B^jx_t＝x_t-j)，d为差分次数；p为模型的阶次；Among them, x_t is the flow data sequence by vehicle type, ε_t is white noise, B is the delay operator (B^j x_t = x_tj ), d is the order of difference; p is the order of the model;利用相关矩估计法及BIC准则进行模型参数的估计及模型阶次p的确定，具体公式如下：The correlation moment estimation method and the BIC criterion are used to estimate the model parameters and determine the model order p. The specific formula is as follows:其中，和分别表示分车型流量数据序列的自相关函数和模型参数；p表示模型的阶次，在计算自相关函数时，一般取为其中N表示样本量。in, and Respectively represent the autocorrelation function and model parameters of the vehicle type flow data sequence; p represents the order of the model, when calculating the autocorrelation function, it is generally taken as where N is the sample size.5.根据权利要求1所述基于智能分车型的高速公路短时交通流量预测方法，其特征在于：所述步骤三中，二次指数平滑法预测模型的公式如下：5. according to the described expressway short-term traffic flow prediction method based on intelligence sub-vehicle of claim 1, it is characterized in that: in described step 3, the formula of quadratic exponential smoothing prediction model is as follows:${\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}}<span><span>=</span>=</span>{\mathrm{<span><span>a</span>a</span>}}_{\mathrm{<span><span>t</span>t</span>}}<span><span>+</span>+</span>{\mathrm{<span><span>b</span>b</span>}}_{\mathrm{<span><span>t</span>t</span>}}\mathrm{<span><span>T</span>T</span>}$其中：为第t+T期预测值；in: is the predicted value of period t+T;a_t和b_t分别为模型参数，且$a_t=2S_t^1-S_t^2,$$b_t=\frac{\mathrm{\&alpha;}}{1-\mathrm{\&alpha;}}(S_t^1-S_t^2);$a_t and b_t are model parameters respectively, and$a_t=2S_t^1-S_t^2,$$b_t=\frac{\mathrm{\&alpha;}}{1-\mathrm{\&alpha;}}(S_t^1-S_t^2);$为第t期的一次指数平滑值，且 is the exponential smoothing value of period t, and为第t期的二次指数平滑值，且 is the quadratic exponential smoothing value of period t, andα为平滑系数，X_t-1为初始值；α is the smoothing coefficient, and X_t-1 is the initial value;流量预测值的转换公式为：traffic forecast The conversion formula for is:${\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>q</span>q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}}<span><span>=</span>=</span>{\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}}<span><span>-</span>-</span>{\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}}_{\mathrm{<span><span>t</span>t</span>}<span><span>+</span>+</span>\mathrm{<span><span>T</span>T</span>}<span><span>-</span>-</span>\mathrm{<span><span>1</span>1</span>}}$其中，和为利用二次指数平滑法对最优K近邻序列t+T-1时刻和t+T时刻的流量预测值。in, and It is the flow forecast value of the optimal K-nearest neighbor sequence at time t+T-1 and time t+T by using the quadratic exponential smoothing method.6.根据权利要求1所述基于智能分车型的高速公路短时交通流量预测方法，其特征在于：所述步骤四中，t时刻的折算后的总车流量预测值为：6. according to the described expressway short-term traffic flow forecasting method based on intelligence sub-model of claim 1, it is characterized in that: in described step 4, the total traffic flow forecast value after the conversion of t moment for:$\stackrel{<span><span>^</span>^</span>}{\mathrm{<span><span>Q</span>Q</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>=</span>=</span>\mathrm{<span><span>1</span>1</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>1</span>1</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>+</span>+</span>\mathrm{<span><span>1.5</span>1.5</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>2</span>2</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>+</span>+</span>\mathrm{<span><span>2</span>2</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>3</span>3</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span><span><span>+</span>+</span>\mathrm{<span><span>3</span>3</span>}<span><span>\&times;</span>\&times;</span>{\mathrm{<span><span>q</span>q</span>}}_{\mathrm{<span><span>4</span>4</span>}}<span><span>(</span>(</span>\mathrm{<span><span>t</span>t</span>}<span><span>)</span>)</span>$其中，q₁(t)、q₂(t)、q₃(t)、q₄(t)分别为t时刻小型车、中型车、大型车、集装箱型车的流量预测值。Among them, q₁ (t), q₂ (t), q₃ (t), and q₄ (t) are the flow forecast values of small vehicles, medium vehicles, large vehicles, and container vehicles at time t, respectively.