CN-121985455-A - Tunnel partition lighting dynamic optimal control method and system based on traffic space-time distribution prediction

CN121985455ACN 121985455 ACN121985455 ACN 121985455ACN-121985455-A

Abstract

The invention relates to a tunnel partition lighting dynamic optimal control method and system based on traffic space-time distribution prediction, and belongs to the technical field of tunnel lighting energy-saving operation. The method aims to solve the problems of black hole effect caused by response lag, trailing illumination caused by rough partition and lack of predictive capability in the prior control technology. The technical scheme is that data are collected through a multi-source traffic perception layer, a traffic flow space-time track prediction layer fuses a macroscopic microscopic model and deep learning to conduct advanced prediction, a dynamic partition optimization decision layer maps a prediction result into a virtual light packet, optimal dimming instructions meeting the constraint of illumination, uniformity and vehicle speed associated brightness change rate are solved through model prediction control, and finally the optimal dimming instructions are output through a smooth execution and closed loop correction layer. The system can fundamentally eliminate the black hole effect, improve the safety, realize the fine energy-saving control of the lamp running along with the vehicle, ensure the visual comfort, and has strong self-adaptability and robustness to multiple traffic scenes.

Inventors

SHI MINGJUN
CHEN CHEN
YUAN FUSONG
WANG HAOHUAN
TIAN FUZHI
WU JINSUO
Li Shicao
XIAO YAO
LI ZAIYONG
SHI LINGNA
MA LANG

Assignees

贵州省公路工程集团有限公司
招商局重庆交通科研设计院有限公司

Dates

Publication Date: 20260505
Application Date: 20260318

Claims (9)

1. A tunnel partition lighting dynamic optimal control method based on traffic space-time distribution prediction is characterized by comprising the following steps: a multi-source traffic sensing step, namely arranging detection devices in front of a tunnel entrance and in the tunnel, and acquiring and fusing traffic information in real time to form a unified traffic state vector; a traffic flow space-time track prediction step, based on the traffic state vector, predicting the traffic flow space-time distribution of the full tunnel in the future prediction time domain by adopting a layered hybrid prediction architecture, and outputting a traffic flow space-time distribution prediction matrix; A dynamic partition optimization decision step, wherein the traffic flow space-time distribution prediction matrix is mapped into illumination demand weights of all illumination partitions, a model prediction control MPC optimization model is constructed, total energy consumption minimization and dimming smoothness are taken as targets, and an optimal dimming instruction sequence of each illumination loop in the future prediction time domain is solved under the condition that illumination constraint, brightness uniformity constraint and vehicle speed association type brightness change rate constraint are met; And performing smoothing and closed loop correction, namely performing time domain smoothing on the optimal dimming command sequence to obtain a final dimming command, sending the final dimming command to the lighting controller, simultaneously comparing the actual measured vehicle position in the hole with the predicted position, and compensating the predicted error through a rolling time domain corrector.
2. The method for dynamically and optimally controlling illumination of a tunnel partition based on traffic space-time distribution prediction as set forth in claim 1, wherein the traffic state vector is , wherein, The unit is the section flow at the moment t, the unit is vehicle/hour, The spatial average density at time t, in units of vehicles/km, The time average speed at time t is expressed in kilometers per hour, The average headway at time t, in seconds, And classifying vectors for the vehicle types at the time t.
3. The method for dynamically and optimally controlling the illumination of the tunnel partition based on traffic space-time distribution prediction according to claim 1 or 2, wherein the traffic space-time trajectory prediction step specifically comprises the following steps: traffic state discrimination step, according to real-time traffic density And the time interval of the head of a vehicle And a preset threshold value 、 The comparison result of the (a) is adaptively switched to a micro track tracking mode or a macro density wave prediction mode; A microscopic track tracking step, in which a state estimator is established for each vehicle in a microscopic track tracking mode, the position and the speed of the bicycle in the tunnel are predicted by adopting improved Kalman filtering, the improved Kalman filtering introduces an adaptive noise covariance adjustment mechanism, a process noise covariance matrix Q is dynamically adjusted according to the variance of the speed variation in adjacent control periods, and the adjustment rule is that , wherein, As a forgetting factor, For the Kalman gain matrix, Is an information vector; A macroscopic density wave prediction step, in which a macroscopic density wave predictor based on a traffic flow fluid dynamic model is adopted to predict traffic density waves in a tunnel under a macroscopic density wave prediction mode, and the correction speed-density relation of the model is as follows , wherein, In order to achieve a free flow velocity, In order to achieve a blocking density, As a function of the shape parameter(s), Correcting coefficients for the closed environment of the tunnel; And an auxiliary prediction and fusion step, namely introducing a deep learning auxiliary predictor based on a transducer architecture, learning historical traffic flow time sequence data, generating an auxiliary prediction result, and carrying out weighted fusion on the output of the micro track tracking step or the macro density wave prediction step and the auxiliary prediction result to generate the traffic flow space-time distribution prediction matrix.
4. The method for dynamically and optimally controlling illumination of a tunnel partition based on traffic space-time distribution prediction as set forth in claim 1, wherein the rate of change of brightness in the vehicle speed correlation is constrained to be , wherein, The average brightness value of the road surface corresponding to the jth loop, Is defined as a maximum allowable brightness change rate function of vehicle speed , wherein, For the intra-tunnel predicted average speed at time t, For the maximum value of the design luminance, Is a low-speed reference rate of change coefficient, For the velocity-coupled attenuation coefficient, Is the free flow velocity.
5. The method for dynamically and optimally controlling illumination of a tunnel partition based on traffic space-time distribution prediction as set forth in claim 1, wherein said time-domain smoothing process employs an exponentially weighted moving average method with the formula , wherein, The j-th loop optimal dimming proportion solved for the dynamic partition optimization decision step, For the final dimming instruction after the smoothing process, Is a smoothing coefficient.
6. A tunnel partition lighting dynamic optimal control system based on traffic space-time distribution prediction is characterized by comprising the following components: The multi-source traffic perception layer comprises a radar detection unit and a high-definition camera which are arranged in front of a tunnel entrance, and an in-tunnel section detector which is arranged in a plurality of sections in the tunnel and is used for acquiring and fusing original traffic information to generate traffic state vectors; the traffic space-time track prediction layer is connected with the multi-source traffic perception layer and is used for receiving the traffic state vector and outputting a traffic space-time distribution prediction matrix in the future prediction domain by adopting a layered hybrid prediction architecture; The dynamic partition optimization decision layer is connected with the traffic space-time track prediction layer and is used for mapping the traffic space-time distribution prediction matrix into illumination demand weight, and generating an optimal dimming instruction sequence of each illumination loop by solving a model prediction control MPC optimization model; the smooth execution and closed loop correction layer is connected with the dynamic partition optimization decision layer, the traffic space-time track prediction layer and the illumination controller and is used for carrying out smooth processing on the optimal dimming instruction sequence and issuing the optimal dimming instruction sequence, and meanwhile, rolling time domain compensation is carried out on the prediction error according to the measured data of the in-hole section detector.
7. The tunnel partition lighting dynamic optimal control system based on traffic space-time distribution prediction according to claim 6, wherein the traffic space-time track prediction layer comprises a traffic state discriminator, a micro track tracking module, a macroscopic density wave prediction module, a deep learning auxiliary predictor and a weighted fusion processor, wherein the traffic state discriminator selectively activates the micro track tracking module or the macroscopic density wave prediction module according to real-time traffic density and headway, and the weighted fusion processor performs weighted fusion on the output of the micro track tracking module or the macroscopic density wave prediction module and the output of the deep learning auxiliary predictor.
8. The tunnel partition lighting dynamic optimal control system based on traffic space-time distribution prediction according to claim 6 or 7, wherein the dynamic partition optimization decision layer comprises a virtual light Bao Yingshe unit and a model prediction control MPC optimization solver, wherein the virtual light Bao Yingshe unit is used for converting the traffic space-time distribution prediction matrix into dynamic lighting requirements of each lighting partition, and the model prediction control MPC optimization solver is used for solving a quadratic programming problem comprising a total energy consumption term, a dimming smoothing term, an illumination lower limit constraint, a brightness uniformity constraint and a vehicle speed-related brightness change rate constraint.
9. The tunnel partition lighting dynamic optimal control system based on traffic space-time distribution prediction according to claim 6, wherein the smoothing execution and closed loop correction layer comprises a time domain smoothing module, an error evaluation module, a rolling time domain corrector and a safety redundancy protection module, the time domain smoothing module is used for smoothing a dimming instruction, the error evaluation module is used for calculating deviation of a vehicle position predicted value and an actual measurement value, the rolling time domain corrector is used for feeding the deviation back to the traffic space-time track prediction layer to correct the prediction, and the safety redundancy protection module is used for triggering a forced brightening instruction when the deviation exceeds a safety threshold.

Description

Tunnel partition lighting dynamic optimal control method and system based on traffic space-time distribution prediction Technical Field The invention belongs to the technical field of tunnel illumination energy-saving operation, and relates to a tunnel partition illumination dynamic optimal control method and system based on traffic space-time distribution prediction. Background Highway tunnels are a closed or semi-closed special road environment with extremely low internal light, which must be kept artificially illuminated, both day and night, which determines that the tunnel lighting system must be operated continuously for a long period of time. The illumination energy consumption accounts for 50 to 70 percent of the total energy consumption of highway tunnel operation and is the biggest single energy consumption source in the tunnel electromechanical system. Therefore, how to realize the fine energy-saving control of tunnel illumination on the premise of ensuring the driving safety visual environment is always a core technical problem in the field. At present, highway tunnel lighting control technologies are mainly divided into the following two main categories: The first type is timing control and off-hole brightness linkage control. According to the method, the brightness gears of all lamps in the tunnel are uniformly adjusted according to a preset time schedule or measured values of an external environment brightness sensor in the tunnel. Its advantages are simple implementation and low cost. However, the control method has the obvious defects that (1) the real-time traffic flow state in the tunnel cannot be perceived, serious 'ineffective lighting' waste exists in the low traffic flow period, (2) the regional granularity is rough, the tunnel is generally divided into 3 to 4 lighting areas of an inlet section, a middle section and an outlet section, the inside of each area cannot be subjected to differential dimming according to traffic flow distribution, and (3) the lighting strategy is lagged behind the actual demand, cannot cope with special traffic scenes such as holidays, construction management and control and the like, and has poor adaptability. The second type is vehicle detection trigger control. According to the method, the vehicle detectors are arranged on a plurality of sections in the tunnel, when the vehicle is detected to enter a certain zone, the corresponding illumination loop is triggered to improve the brightness, and the time delay is reduced after the vehicle leaves. The system has the following basic defects that (1) response lag is difficult to eliminate, a system triggers illumination action after a sensor senses a vehicle, the process of sensing, transmitting, responding and lighting is delayed, a driver enters an insufficiently illuminated area in the period, (2) partition control granularity is coarse, the problem of trailing illumination is prominent, the vehicle can be turned off after waiting for delay after passing, the area without the vehicle can maintain a high-brightness state for a long time, 3, sparse traffic flow and dense traffic flow are difficult to be compatible, fixed trigger logic cannot be used for adaptively distinguishing different traffic flow density scenes, and (4) the system lacks prospective prediction capability, is completely driven by the generated events and does not have the capability of predicting the space-time distribution of future traffic flow. In summary, the existing tunnel illumination control technology has three core technical bottlenecks of lag response, coarse partition granularity and lack of prejudgment capability on space-time distribution of future traffic flow, so that energy saving efficiency and driving vision safety are difficult to be compatible. Disclosure of Invention In view of the above, the present invention aims to provide a tunnel partition lighting dynamic optimal control method and system based on traffic space-time distribution prediction. In order to achieve the above purpose, the present invention provides the following technical solutions: a tunnel partition lighting dynamic optimal control method based on traffic space-time distribution prediction comprises the following steps: a multi-source traffic sensing step, namely arranging detection devices in front of a tunnel entrance and in the tunnel, and acquiring and fusing traffic information in real time to form a unified traffic state vector; a traffic flow space-time track prediction step, based on the traffic state vector, predicting the traffic flow space-time distribution of the full tunnel in the future prediction time domain by adopting a layered hybrid prediction architecture, and outputting a traffic flow space-time distribution prediction matrix; A dynamic partition optimization decision step, wherein the traffic flow space-time distribution prediction matrix is mapped into illumination demand weights of all illumination partitions, a model predictio