Background
Compared with the traditional fuel oil automobile, the electric automobile has the advantages of environmental protection, high efficiency and the like, but has the defects of long charging time and short endurance mileage under the influence of battery capacity. Therefore, in the actual daily driving process, the owner of the electric automobile often has the requirement of searching for the charging pile. In the navigation process of finding the best surrounding charging pile for the electric vehicle and planning the best path, whether the current remaining electric quantity of the electric vehicle can reach the target charging pile needs to be considered firstly. Therefore, the navigation system is required to accurately predict the speed of the automobile in the process of traveling along a planned path, further calculate the time required by the automobile to reach a destination along the planned path and the energy consumption according to the predicted speed, and finally make a decision of an optimal target charging pile and plan the optimal path.
In actual city driving, the biggest factor that can influence the running speed of an automobile is the congestion condition of a road. Therefore, prediction of the traveling speed of a vehicle based on the congestion status of a real-time road is a widely accepted criterion. The current prediction algorithms for the running speed of the automobile based on the congestion condition of the real-time road are many, such as algorithms of a neural network, Kalman filtering and the like. The neural network algorithm is based on data, a large amount of training is needed in the prediction process, the more training times, the more accurate the result is, but the calculation amount and the storage amount are excessive, and the navigation decision-making time is too long. The Kalman filtering technology is based on a model, adopts a recursion algorithm flow, and has small memory space. The volumetric kalman filtering technique can be used for non-linear systems and is accurate to a high degree, however, it has the disadvantage that the variance of the process noise and the observation noise needs to be known in the process of velocity prediction with volumetric kalman filtering. Therefore, the variance of the two needs to be estimated by using an adaptive filtering technology, and the most commonly used adaptive filtering algorithm for variance estimation includes two methods, namely, Sage-Husa and variational bayes, but in actual engineering, the estimation effect of the two methods is sometimes not ideal, the accuracy is not high, and even the filtering is scattered.
Therefore, in the electric vehicle charging navigation system, the algorithm is difficult to quickly and accurately predict the speed of the electric vehicle when the electric vehicle runs along a certain planned path in the future, and the accuracy of the speed prediction plays an important role in the navigation system.
Disclosure of Invention
The invention provides a speed prediction method of an electric vehicle charging navigation system, which can be used for rapidly and accurately predicting the speed of an electric vehicle when the electric vehicle runs along a planned path in the future, and has the advantages of small calculation amount, high calculation speed, low requirement on storage space and capability of better meeting the actual use requirement of electric vehicle charging navigation.
In order to solve the problems, the invention adopts the following technical scheme:
the invention discloses a speed prediction method of an electric automobile charging navigation system, which comprises the following steps:
s1: establishing a vehicle speed prediction system model which comprises a state equation taking the vehicle speed as a state quantity and an observation equation established according to the relation between the target road congestion condition and the vehicle speed;
s2: obtaining the congestion condition of the target road at the current time from a traffic information center so as to determine the initial value of the state quantity of the vehicle speed prediction system model
And its covariance initial value
Simultaneous setting of expected initial values of process noise for a vehicle speed prediction system model
Variance initial value
And observing a desired initial value of noise
Variance initial value
S3: using a volumetric Kalman filter, a one-step prediction of a state is calculated
And its error covariance P
k|k-1And calculating updated state volume points of the propagation of the nonlinear observation equation
Pre-measurement volumetric point
S4: calculating an expectation of observed noise
Sum variance
S5: calculating the state estimation value of the vehicle speed prediction system model at the moment
And its error covariance P
m k|kM is initially 1;
s6: judging whether m is less than N0If so, m is m +1, jumping to step S4, otherwise, executing step S7;
s7: taking the final result as a state estimation value
And its error covariance P
k|kAs a result of (1), i.e.
The speed predicted value at the final k moment is obtained;
s8: estimating the period of process noise at time k using the Sage-Husa methodInspection of
Sum variance
Preferably, the formula of the vehicle speed prediction system model is as follows:
wherein the state quantity XkIndicating the speed of the vehicle at time k, the observed quantity ZkIndicating the number of vehicles observed on the road at time K, L indicating the length of the road, K*A congestion coefficient indicating a maximum traffic flow; v*Represents the maximum speed limit, w, of the road vehiclek-1And vkRepresenting process noise and observation noise.
Preferably, the congestion coefficient K of the link at the current time is calculated according to the length of the link and the number of vehicles at the current time*。
Preferably, the step S4 includes the steps of:
s41: calculating an expectation of observing noise using the Sage-Husa method
The variance of the observed noise is calculated by using the Sage-Husa method and is recorded as
S42: calculating the variance of the observed noise by using a variational Bayes method, and recording the variance as
S43: will be provided with
And
the result is summed as the final estimate of the observed noise variance at that time
The method estimates the process noise variance of the system by using a Sage-Husa method, and jointly estimates the observation noise variance of the system by using the Sage-Husa and a variational Bayes method. And then, based on the current congestion condition of the road, predicting the speed of the electric automobile running along the road in a future period by using a cubature Kalman filtering method.
The invention has the beneficial effects that: (1) the speed of the electric automobile when driving along a certain planned path in the future can be predicted quickly and accurately, the calculation amount is small, the calculation speed is high, meanwhile, the requirement on the storage space is low, and the actual use requirement of the electric automobile charging navigation is met better. (2) By means of joint estimation of the Sage-Husa method and the variational Bayesian method, the problem of poor estimation effect of a single method is solved, and variance estimation accuracy is improved.
Example (b): the speed prediction method of the electric vehicle charging navigation system of the embodiment, as shown in fig. 1, includes the following steps:
s1: establishing a vehicle speed prediction system model, wherein the formula of the vehicle speed prediction system model is as follows:
wherein the state quantity XkIndicating the speed of the vehicle at time k, the observed quantity ZkRepresents kThe number of vehicles observed on the road at the time, L representing the length of the road, K*A congestion coefficient (for example, 30) indicating a maximum traffic flow; v*Representing the maximum speed limit (e.g. 60KM/h), w of the road vehiclek-1And vkRepresenting process noise and observation noise (whose statistical properties are unknown and which need to be estimated using adaptive filtering);
s2: obtaining the congestion condition of the target road at the current time from a traffic information center so as to determine the initial value of the state quantity of the vehicle speed prediction system model
And its covariance initial value
Simultaneous setting of expected initial values of process noise for a vehicle speed prediction system model
Variance initial value
And observing a desired initial value of noise
Variance initial value
S3: using a volumetric Kalman filter, a one-step prediction of a state is calculated
And its error covariance P
k|k-1And calculating updated state volume points of the propagation of the nonlinear observation equation
Pre-measurement volumetric point
S4: calculating an expectation of observed noise
Sum variance
S5: calculating the state estimation value of the vehicle speed prediction system model at the moment
And its error covariance P
m k|kM is initially 1;
s6: judging whether m is less than N0If so, m is m +1, jumping to step S4, otherwise, executing step S7;
s7: taking the final result as a state estimation value
And its error covariance P
k|kAs a result of (1), i.e.
The speed predicted value at the final k moment is obtained;
s8: estimating process noise expectation at time k using Sage-Husa method
Sum variance
Calculating the congestion coefficient K of the road section at the current moment according to the length of the road section and the number of vehicles at the current moment*。
Step S4 includes the following steps:
s41: calculation Using Sage-Husa methodTo come out of expectation of observation noise
The variance of the observed noise is calculated by using the Sage-Husa method and is recorded as
S42: calculating the variance of the observed noise by using a variational Bayes method, and recording the variance as
S43: will be provided with
And
the result is summed as the final estimate of the observed noise variance at that time
The method estimates the process noise variance of the system by using a Sage-Husa method, and jointly estimates the observation noise variance of the system by using the Sage-Husa and a variational Bayes method. And then, based on the current congestion condition of the road, predicting the speed of the electric automobile running along the road in a future period by using a cubature Kalman filtering method.