Pavement ice-condensation time prediction method, system, device and storage medium
阅读说明：本技术 一种路面凝冰时间预测方法、系统、装置及存储介质 (Pavement ice-condensation time prediction method, system, device and storage medium ) 是由 王秋才 韩博 闫茂德 左磊 杨盼盼 于 2021-06-24 设计创作，主要内容包括：本发明公开了一种路面凝冰时间预测方法、系统、装置及存储介质,路面凝冰时间预测方法包括：获取最近一组滚动窗口路面温度数据,进行归一化数据处理,初始化一个预测计步器p＝0,并记录当前时间值t-c；将归一化的时序路面温度数据输入到已训练好的BP神经网络的时序回归预测模型,计算得到下个时刻的路面温度预测值为x-(predict)(n+p)；在x-(predict)(n+p)＞x-(ice)时,使预测计步器p＝p+1,x-(ice)为路面凝冰温度实际值；将下个时刻的路面温度预测值作为实时时序路面温度数据的最后一个时序数据,得到当前时序的时序路面温度数据x-(window),当x-(predict)(n+p)≤x-(ice)时,输出此时预测计步器p的值作为路面凝冰时间预测值。(The invention discloses a method, a system, a device and a storage medium for predicting pavement ice-condensation time, wherein the method for predicting pavement ice-condensation time comprises the following steps: acquiring a recent group of rolling window road surface temperature data, carrying out normalization data processing, initializing a predicted pedometer p to be 0, and recording the current time value t c (ii) a Inputting the normalized time sequence road surface temperature data into a trained time sequence regression prediction model of a BP neural network, and calculating to obtain a road surface temperature predicted value x at the next moment predict (n + p); at x predict (n+p)＞x ice Then, let the prediction pedometer p be p +1, x ice The actual value of the pavement ice-condensing temperature is obtained; taking the predicted value of the road surface temperature at the next moment as the last time sequence data of the real-time sequence road surface temperature data to obtain the time sequence road surface temperature data x of the current time sequence window When x is predict (n+p)≤x ice Then, the predictor at that time is outputAnd taking the value of the step p as the predicted value of the road surface ice-setting time.)
1. A pavement ice-setting time prediction method is characterized by comprising the following steps:
acquiring time sequence road surface temperature data of a latest group of rolling windows as xwindow＝(xn-w，xn-w+1，…，xn-1)；
To time sequence road surface temperature data xwindow＝(xn-w，xn-w+1，…，xn-1) Carrying out normalization data processing to obtain normalized time sequence pavement temperature data;
initializing a predicted pedometer p as 0 and recording the current time value tc(ii) a Inputting the normalized time sequence road surface temperature data into a trained time sequence regression prediction model of a BP neural network, and calculating to obtain a road surface temperature predicted value x at the next momentpredict(n+p)；
At xpredict(n+p)＞xiceThen, let the prediction pedometer p be p +1, xiceThe actual value of the pavement ice-condensing temperature is obtained; taking the predicted road surface temperature value at the next moment as the last time sequence data of the real-time sequence road surface temperature data to obtain the time sequence of the current time sequenceSequential road surface temperature data xwindowWhen x ispredict(n+p)≤xiceAnd outputting the value of the predicted pedometer p at the moment as the predicted road surface ice-setting time value.
2. The method according to claim 1, wherein time-series road surface temperature data x ═ x (x) is collected before prediction0，x1，…，xn-1) If the time series road surface temperature data x is equal to (x)0，x1，…，xn-1) If the variation trend of (2) is slowly reduced, the prediction condition is satisfied, and prediction is performed.
3. The method for predicting the road ice-freezing time according to claim 1, wherein the BP neural network model is obtained as follows:
s201, establishing a time sequence regression prediction model of the BP neural network, and establishing a BP neural network model with rolling window size, a plurality of hidden layers and single output by setting the number of network layers of the BP neural network and the number of neurons of each layer, wherein the output is a pavement temperature value;
s202, presetting a time sequence regression prediction model target parameter of a BP neural network;
s203, taking next pavement temperature real data in the time sequence training sample as a BP neural network label value, taking continuous pavement temperature time sequence data with the size of a rolling window in the time sequence training sample as an input value of the BP neural network, and training a time sequence regression prediction model of the BP neural network;
s204, judging whether the evaluation index of the time sequence regression prediction model of the BP neural network meets a preset target parameter, if not, returning to the step S203, and continuing to iteratively train the time sequence regression prediction model of the BP neural network; if yes, continue to step S205;
and S205, storing the trained time sequence regression prediction model of the BP neural network.
4. The method for predicting the road surface icing time according to claim 3, wherein the time series training samples are obtained by: acquiring historical road surface temperature data of a monitored road section; and generating an actual sample by the historical road surface temperature data through a rolling window data processing technology, and normalizing to obtain a time sequence training sample.
5. The method according to claim 1, wherein the time-series training samples include a plurality of continuous road surface temperature time-series data having a size of a rolling window and road surface temperature true data at the next time.
6. A system for predicting the ice-setting time of a road surface according to any one of claims 1 to 5, comprising:
a first module for obtaining time sequence road surface temperature data of a latest group of rolling windows as xwindow＝(xn-w，xn-w+1，…，xn-1)；
A second module for comparing time sequence road surface temperature data xwindow＝(xn-w，xn-w+1，…，xn-1) Carrying out normalization data processing to obtain normalized time sequence pavement temperature data;
a third module for initializing a predicted pedometer p to 0 and recording the current time value tc(ii) a Inputting the normalized time sequence road surface temperature data into a trained time sequence regression prediction model of a BP neural network, and calculating to obtain a road surface temperature predicted value x at the next momentpredict(n+p)；
A fourth module for applying a voltage at xpredict(n+p)＞xiceThen, let the prediction pedometer p be p +1, xiceThe actual value of the pavement ice-condensing temperature is obtained; taking the predicted value of the road surface temperature at the next moment as the last time sequence data of the real-time sequence road surface temperature data to obtain the time sequence road surface temperature data x of the current time sequencewindowWhen x ispredict(n+p)≤xiceAnd outputting the value of the predicted pedometer p at the moment as the predicted road surface ice-setting time value.
7. An apparatus for predicting a road surface freezing time according to any one of claims 1 to 5, comprising: a memory and a processor;
the memory for storing a computer program; the processor, when executing the computer program, is configured to implement the road surface freezing time prediction method according to any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that the storage medium has stored thereon a computer program which, when executed by a processor, implements the road surface freezing time prediction method according to any one of claims 1 to 5.
Technical Field
The invention belongs to the technical field of highway detection, and particularly relates to a method, a system and a device for predicting pavement ice-freezing time and a storage medium.
Background
The changeable meteorological environment along the highway is always a main factor which hinders the safe and efficient operation of the highway system, and the influence of severe meteorological conditions such as ice, snow, rain and the like on the highway is particularly serious. In winter, due to snowfall and rainfall or large humid and foggy air, under the condition of sudden drop of air temperature, the ice is very easily condensed on the bridge surfaces of special highway sections in high-latitude and high-altitude humid areas, such as mountain super-large bridges, cloudy slope bends and other road surfaces.
The road surface ice condensation not only can influence the passing efficiency of the highway, but also brings great inconvenience to the trip of people. At present, the research on the problem of road ice condensation at home and abroad mainly focuses on the road ice condensation detection and the deicing and snow melting technology, and the research on the aspect of predicting the impending road ice condensation time is relatively deficient.
In winter, the actual variation trend of the highway pavement temperature is not ideally uniform and straight-line descending, but fluctuates randomly up and down and slowly descends, as shown in fig. 1 and 5, so that the prediction stability of the patent CN104134098B is poor, and meanwhile, the predicted result has a large error, which is not beneficial to practical engineering popularization and application.
Disclosure of Invention
Aiming at the defects of the existing prediction method, the invention provides a pavement ice-setting time prediction method, a pavement ice-setting time prediction system, a pavement ice-setting time prediction device and a storage medium based on a time sequence regression neural network, so that the stability, the applicability and the accuracy of a prediction model are improved.
The invention specifically adopts the following technical scheme to solve the technical problems:
a pavement ice-setting time prediction method comprises the following steps:
acquiring time sequence road surface temperature data of a latest group of rolling windows as xwindow＝(xn-w,xn-w+1,…,xn-1)；
To time sequence road surface temperature data xwindow＝(xn-w,xn-w+1,…,xn-1) Carrying out normalization data processing to obtain normalized time sequence pavement temperature data;
initializing a predicted pedometer p as 0 and recording the current time value tc(ii) a Inputting the normalized time sequence road surface temperature data into a trained time sequence regression prediction model of a BP neural network, and calculating to obtain a road surface temperature predicted value x at the next momentpredict(n+p)；
At xpredict(n+p)>xiceThen, let the prediction pedometer p be p +1, xiceThe actual value of the pavement ice-condensing temperature is obtained; taking the predicted value of the road surface temperature at the next moment as the last time sequence data of the real-time sequence road surface temperature data to obtain the time sequence road surface temperature data x of the current time sequencewindowWhen x ispredict(n+p)≤xiceAnd outputting the value of the predicted pedometer p at the moment as the predicted road surface ice-setting time value.
Furthermore, before prediction, time-series road surface temperature data x ═ x (x) is collected0,x1,…,xn-1) If the time series road surface temperature data x is equal to (x)0,x1,…,xn-1) If the variation trend of (2) is slowly reduced, the prediction condition is satisfied, and prediction is performed.
Further, the acquisition mode of the BP neural network model is as follows:
s201, establishing a time sequence regression prediction model of the BP neural network, and establishing a BP neural network model with rolling window size, a plurality of hidden layers and single output by setting the number of network layers of the BP neural network and the number of neurons of each layer, wherein the output is a pavement temperature value;
s202, presetting a time sequence regression prediction model target parameter of a BP neural network;
s203, taking next pavement temperature real data in the time sequence training sample as a BP neural network label value, taking continuous pavement temperature time sequence data with the size of a rolling window in the time sequence training sample as an input value of the BP neural network, and training a time sequence regression prediction model of the BP neural network;
s204, judging whether the evaluation index of the time sequence regression prediction model of the BP neural network meets a preset target parameter, if not, returning to the step S203, and continuing to iteratively train the time sequence regression prediction model of the BP neural network; if yes, continue to step S205;
and S205, storing the trained time sequence regression prediction model of the BP neural network.
Further, the method for obtaining the time sequence training sample is as follows: acquiring historical road surface temperature data of a monitored road section; and generating an actual sample by the historical road surface temperature data through a rolling window data processing technology, and normalizing to obtain a time sequence training sample.
Further, the time sequence training sample comprises a plurality of continuous road surface temperature time sequence data with the size of a rolling window and road surface temperature real data at the next moment.
The embodiment of the invention provides another technical scheme that:
a system for the road ice freezing time prediction method, comprising:
a first module for obtaining time sequence road surface temperature data of a latest group of rolling windows as xwindow＝(xn-w,xn-w+1,…,xn-1)；
A second module for comparing time sequence road surface temperature data xwindow＝(xn-w,xn-w+1,…,xn-1) Carrying out normalization data processing to obtain normalized time sequence pavement temperature data;
a third module for initializing a predicted pedometer p to 0 and recording the current time value tc(ii) a Inputting the normalized time sequence road surface temperature data into a trained time sequence regression prediction model of a BP neural network, and calculating to obtain a road surface temperature predicted value x at the next momentpredict(n+p)；
A fourth module for applying a voltage at xpredict(n+p)>xiceThen, let the prediction pedometer p be p +1, xiceThe actual value of the pavement ice-condensing temperature is obtained; taking the predicted value of the road surface temperature at the next moment as the last time sequence data of the real-time sequence road surface temperature data to obtain the time sequence road surface temperature of the current time sequenceDegree data xwindowWhen x ispredict(n+p)≤xiceAnd outputting the value of the predicted pedometer p at the moment as the predicted road surface ice-setting time value.
The embodiment of the invention provides another technical scheme that:
an apparatus for the road surface ice-freezing time prediction method, characterized by comprising: a memory and a processor;
the memory for storing a computer program; the processor is used for realizing the road surface ice-freezing time prediction method when the computer program is executed.
The embodiment of the invention provides another technical scheme that:
a computer-readable storage medium, characterized in that the storage medium has stored thereon a computer program which, when executed by a processor, implements the road icing time prediction method.
Compared with the prior art, the invention has the following beneficial effects:
1) according to the method, the time sequence regression prediction model of the BP neural network is constructed, random fluctuation of time sequence sample data is avoided by using a rolling window data processing technology, and the stability of model prediction is ensured.
2) According to the method, a time sequence regression prediction model based on the BP neural network can be continuously optimized by adopting a large amount of historical data of the pavement temperature, so that the accuracy of the model for predicting the pavement ice-freezing time is improved.
3) The method is particularly suitable for high-risk road sections which are easy to freeze in cold or high altitude, and is convenient for highway operation departments to master the road surface condition of the road section, and timely issue early warning information of road surface ice condensation in advance, so that the traffic accident rate is reduced, the safe traffic capacity of the highway is improved, and the increasing demands of the public on intelligent high speed are met.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:
fig. 1 is a fitting curve diagram of road surface temperature data and a time-series regression BP neural network actually acquired at a certain road section.
FIG. 2 is a schematic diagram of a time-series regression prediction model of the BP neural network according to the present invention.
FIG. 3 is a graph showing the trend of Mean Square Error (MSE) along with the variation of training period in the training process of the time-series regression prediction model of the BP neural network of the present invention.
FIG. 4 is a flow chart of the time sequence regression prediction of the road surface ice-freezing time of the BP neural network of the present invention.
FIG. 5 is a time-series regression prediction road surface ice-freezing time result diagram of the BP neural network of the present invention.
Detailed Description
The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
The following detailed description is exemplary in nature and is intended to provide further details of the invention. Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention.
The invention provides a pavement ice-freezing time prediction method based on a time sequence regression neural network, which comprises a rolling window preprocessing stage of pavement temperature historical data of a road section to be predicted, a time sequence regression prediction model training stage of a BP neural network and a pavement ice-freezing time real-time prediction stage. In the rolling window preprocessing stage of the historical road temperature data of the road section to be predicted, the historical road temperature data of the monitored road section is firstly acquired, and as shown in table 1, each time sequence in the table is spaced for 1 minute.
TABLE 1 historical road temperature data for a certain section
Generating an actual sample through a rolling window data processing technology and normalizing the actual sample to obtain a time sequence training sample; the rolling window data processing technique uses a continuous plurality (rolling window size) of road surface temperature time series data as a set of actual samples to predict next road surface temperature data; the pavement temperature is mapped to a [0,1] interval by normalization; the time sequence training sample comprises a plurality of continuous road surface temperature time sequence data with the size of a rolling window and road surface temperature real data at the next moment.
In this embodiment, the size of the rolling window is 6, and the prediction step is 1, so that the road surface temperature time sequence regression training samples shown in table 2 can be obtained, and the last one is a time sequence regression prediction sample: 196,(-2.8-2.9-2.8-2.8-2.8-3.0), -3.1.
TABLE 2 time series regression training samples
The training stage of the time sequence regression prediction model of the BP neural network comprises the following steps:
step S201: establishing a time sequence regression prediction model of the BP neural network, establishing a BP neural network model with rolling window size of neural network input, a plurality of hidden layers and single output by setting the number of network layers of the BP neural network and the number of neurons of each layer, and outputting the BP neural network model as a pavement temperature value. The structure diagram of the time-series regression prediction model of the BP neural network of this embodiment is shown in fig. 2, and the prediction model parameters are shown in table 3.
TABLE 3 chronogenesis regression prediction model parameters for BP neural networks
Step S202: target parameters of the time sequence regression prediction model of the BP neural network are preset, and the preset training target parameters of the embodiment are shown in table 4.
TABLE 4 training target parameters
Step S203: and (3) taking the next pavement temperature real data in the time sequence training sample as a BP neural network label value, taking continuous pavement temperature time sequence data with the size of a rolling window in the time sequence training sample as an input value of the BP neural network, and training a time sequence regression prediction model of the BP neural network.
Step S204: judging whether the evaluation index of the time sequence regression prediction model of the BP neural network meets any preset directory training target in the step S202, if not, returning to the step S203, and continuing to iteratively train the time sequence regression prediction model of the BP neural network; if yes, the process continues to step S205.
Step S205: the trained time sequence regression prediction model of the BP neural network is stored, a time sequence regression curve is shown in figure 1, Mean Square Error (MSE) is shown in figure 3, and the training result data of the prediction model is shown in table 5.
TABLE 5 prediction model training results
The real-time pavement ice-condensation time prediction stage comprises the following steps:
step S301: for convenience of prediction and comparison with the conventional straight line uniform method, it is assumed that the time sequence road surface temperature data actually acquired on site is a part of data x (x) in table 10,x1,…,xn-1) Where n is 16, as shown in table 6, the time-series road surface temperature data of the latest set of rolling windows obtained is xwindow＝(-0.20-0.20-0.30-0.30-0.30-0.40)。
TABLE 6 on-site pavement temperature time series data
Time sequence
Temperature value (. degree. C.)
Time sequence
Temperature value (. degree. C.)
0
0
8
-0.3
1
0
9
-0.3
2
0
10
-0.2
3
0
11
-0.2
4
-0.1
12
-0.3
5
-0.1
13
-0.3
6
-0.2
14
-0.3
7
-0.2
15
-0.4
Step 302: initializing a predicted pedometer p as 0 and recording the current time value tc。
Step S303: time series road surface temperature data x ═ x0,x1,…,xn-1) The change tendency of (2) is generally slowly decreased, as shown in fig. 5, and the prediction condition is satisfied.
Step 304: to time sequence road surface temperature data xwindowAnd (6) carrying out normalized data processing on the obtained data (-0.20-0.20-0.30-0.30-0.40) to obtain normalized time sequence pavement temperature data.
Step S305: inputting the normalized time sequence road surface temperature data into a trained time sequence regression prediction model of a BP neural network, and calculating to obtain a road surface temperature predicted value x at the next momentpredict(n+p)＝-0.39。
Step 306: the actual value of the detected road surface ice-freezing temperature is xice2.8 ℃ C, due to xpredict(n+p)>xiceAnd the predicting pedometer p is p +1, and simultaneously, by utilizing the rolling window characteristic, the predicted value of the road surface temperature at the next moment is used as the last time sequence data of the real-time sequence road surface temperature data, and the time sequence road surface temperature data of the current time sequence is obtained to be xwindow(-0.20-0.30-0.30-0.40-0.39), and go to step 304 to continue iterative prediction; when p is 116, xpredict(n + p) — 2.80 satisfies the condition xpredict(n+p)≤xiceThen, the process goes to step S307.
Step 307: because the predicted pedometer p is 116>0, while each step represents 1 minute, i.e. the preceding time value tcThen, the second stepAfter about 116 minutes, the road surface temperature may reach the freezing temperature xice＝-2.8。
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
And (3) simulation result analysis:
the time-series regression prediction of the BP neural network of the present invention compares the road ice-freezing time with the two-point straight line prediction method of patent CN104134098B and the time required for actual occurrence of road ice as shown in table 1, and the results are shown in table 7.
TABLE 7 comparison of predicted results
As can be seen from table 7:
the time sequence regression prediction method of the BP neural network is basically close to the actually measured road ice-freezing time, and the traditional two-point direct prediction method has larger prediction error.
As can be seen from fig. 1 and 5: the traditional two-point straight line prediction method has larger deviation with the actually measured road surface temperature change trend, but the time sequence regression prediction method of the BP neural network of the invention basically and really reflects the actual nonlinear slow-falling change trend of the road surface temperature and has better prediction stability and accuracy.
In summary, the method for predicting the road ice-condensation time provided by the embodiment of the invention comprises the following steps:
1) according to the method, the time sequence regression prediction model of the BP neural network is constructed, random fluctuation of time sequence sample data is avoided by using a rolling window data processing technology, and the stability of pavement ice-setting time prediction is ensured.
2) According to the method, a time sequence regression prediction model based on the BP neural network can be continuously optimized by adopting a large amount of historical data of the pavement temperature, so that the accuracy of the model for predicting the pavement ice-freezing time is improved.
It will be appreciated by those skilled in the art that the invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The embodiments disclosed above are therefore to be considered in all respects as illustrative and not restrictive. All changes which come within the scope of or equivalence to the invention are intended to be embraced therein.
- 上一篇：一种医用注射器针头装配设备
- 下一篇：基于年降雨长期预报方法的矿区生态需水量计算方法