Inlet ramp convergence control method based on time-space two-dimensional adaptive model prediction of financial risk index

Abstract

The invention discloses a time-space two-dimensional adaptive model prediction entrance ramp convergence control method based on financial risk indexes. According to the method, the traffic flow volatility and the value-at-risk index are innovatively introduced for control short view and queuing overflow risks caused by prediction time domain fixing and constraint boundary rigidity in traditional model prediction control. In the time dimension, dynamically shortening or prolonging the prediction time domain according to the volatility so as to filter long-term prediction noise; in the spatial dimension, the security constraint boundary is dynamically tightened according to the value-at-risk, and a self-adaptive buffer space is reserved for the prediction error. According to the invention, through constructing a space-time two-dimensional adaptive model prediction control framework, the traffic efficiency is improved when the traffic is stable, the intelligent balance of operation safety is ensured when the uncertainty is high, and the robustness and the anti-interference capability of the ramp control system are significantly enhanced.

Description

Technical Field

[0001] This invention belongs to the field of highway ramp control technology, and particularly relates to an entrance ramp merging control method based on a spatiotemporal two-dimensional adaptive model prediction of financial risk indicators. Background Technology

[0002] Existing highway ramp control methods have the following core flaws:

[0003] 1) Fixed prediction time domain: Traditional MPC uses a fixed prediction step size. If the time domain is too short, the control becomes "short-sighted" and cannot cope with congestion waves; if the time domain is too long, noise is introduced due to inaccurate long-term predictions, and the computational load is large.

[0004] 2) Fixed Constraint Boundaries: Existing methods typically set fixed physical constraints (e.g., maximum queue length = 50 meters). However, due to errors in the prediction model, and because these errors follow a certain probability distribution, when the predicted value is 48 meters, the actual value is very likely to exceed 50 meters (tail risk). Ignoring this uncertainty in prediction can easily lead to unexpected overflow of queues on ramps, causing severe traffic paralysis. Summary of the Invention

[0005] The purpose of this invention is to provide an entrance ramp merging control method based on a spatiotemporal two-dimensional adaptive model prediction of financial risk indicators, so as to solve the above-mentioned technical problems.

[0006] To address the aforementioned technical problems, the specific technical solution of the inlet ramp merging control method based on a spatiotemporal two-dimensional adaptive model prediction using financial risk indicators, as described in this invention, is as follows:

[0007] A method for merging control of entrance ramps based on a spatiotemporal two-dimensional adaptive model predicting financial risk indicators includes the following steps:

[0008] Step 1: Data Acquisition and Processing: Retain basic data acquisition. Construct a prediction error database;

[0009] Step 2: Key Indicator Calculation: Based on the sliding time window, calculate both the trend uncertainty indicator and the tail risk indicator.

[0010] Step 3: Parameter mapping mechanism: adopting a spatiotemporal two-dimensional adaptive mapping;

[0011] Step 4: Controller construction;

[0012] Step 5: Rolling optimization and execution.

[0013] Furthermore, step 1 includes the following steps:

[0014] The system compares the "model prediction state" of the previous moment with the "actual detection state" of the current moment in real time, calculates the prediction residual, and dynamically maintains a historical residual sequence of length L.

[0015] Furthermore, step 2 includes the following steps:

[0016] Trend uncertainty indicator – volatility: Calculates the coefficient of variation of upstream flow within window M.

[0017]

[0018] Where v(k) is the traffic flow volatility, used to quantify the system's instability. σ flow (k) represents the standard deviation of the upstream flow rate within window M, μ flow The mean of upstream flow within window M is ∈, where ∈ is a very small positive number to prevent the denominator from being zero. The higher the volatility, the shorter the prediction time domain, in order to filter out unreliable long-term forecasts.

[0019] Tail risk indicator – Value at Risk (V) a R: Calculate the value at risk at confidence level β based on the historical residual sequence.

[0020] V a R β (k)=Percentile(|Residuals|,β) (2,2)

[0021] Where β is the confidence level.

[0022] Furthermore, step 3 includes the following steps:

[0023] Spatiotemporal two-dimensional adaptive mapping:

[0024] Time dimension: Determine the prediction time domain N P Using a negative correlation mapping mechanism, the higher the volatility, the closer the correlation.

[0025] N P (k)=max(N min ,min(N max ,round(N max -α·V(k)))) (3,1

[0026] Where, N min To predict the minimum value in the time domain, N max The maximum value in the prediction time domain is α, which is an adjustment coefficient that controls the degree of influence of volatility on the prediction time domain.

[0027] Spatial dimension: Determining dynamic security constraints C safe Subtracting the risk value from the physical limit yields the dynamic safety boundary:

[0028] W safe (k)=W physical -γ·V a R β (k) (3,2)

[0029] Among them W physical Where γ is the actual physical length of the ramp or the maximum allowable queue length, and γ is the safety factor. This means that if the model's recent predictions are accurate, i.e., V... a A small R value makes full use of road space; if the model's prediction is poor, i.e., V... a When R is large, the system automatically reserves a large buffer zone to prevent overflow. Furthermore, step 4 includes the following steps:

[0030] Objective function: upper bound of summation N P It is a constant.

[0031]

[0032] Constraints:

[0033] CTM dynamic constraints:

[0034] x(t+1)=f CTM (x(t),u(t)) (4,2)

[0035] Physical hard constraints:

[0036] w ramp (t) <W physical (4,3)

[0037] Enforcement constraints:

[0038] u min ≤u(t)≤u max (4,4)

[0039] Robust optimization with varying dimensions:

[0040] Objective function:

[0041]

[0042] Where TTS(t) is the total travel time, which is part of the objective function and reflects traffic efficiency. λ is a weighting coefficient used to balance the importance of total travel time and changes in control actions in the objective function, u(t) is the control input, k is the current time step index, and t is the time index;

[0043] Constraints: CTM dynamic constraints:

[0044] x(t+1)=f CTM(x(t),u(t)) (4,6)

[0045] V a R-buffered queuing constraint (optimization point):

[0046] w ramp (t)≤w safe (k) (4,7)

[0047] Right now

[0048] w ramp (t)≤ <W physical -γ·V a R β (k) (4,8)

[0049] Enforcement constraints:

[0050] u min ≤u(t)≤u max (4,9).

[0051] Furthermore, step 5 includes the following steps:

[0052] Solve linear programming problems using the solver Gurobi / OSQP; obtain the optimal solution u. * (k); issue execution u * (k), repeat steps 1-5 at the next time step.

[0053] The inlet ramp merging control method based on a spatiotemporal two-dimensional adaptive model predicting financial risk indicators, as proposed in this invention, has the following advantages:

[0054] Extremely robustness: By introducing V a R essentially implements an engineering application of "distributionally Robust Control". Even if extreme weather or accidents cause the predictive model to temporarily fail, the system can ensure that the queue at the ramps does not overflow by automatically increasing the safety buffer.

[0055] Anti-interference capability: By dynamically adjusting the time domain through a volatility mechanism, false trend signals in high-noise environments are effectively filtered out, avoiding frequent oscillations of the controller.

[0056] Balancing performance and safety: When predictions are accurate and traffic is stable, the system automatically releases all physical space and long-term planning capabilities to maximize traffic efficiency; when the risk is high, it automatically switches to conservative mode, achieving an intelligent trade-off between efficiency and safety. Attached Figure Description

[0057] none. Detailed Implementation

[0058] To better understand the purpose, structure, and function of this invention, the following provides a more detailed description of the entrance ramp merging control method based on a spatiotemporal two-dimensional adaptive model prediction of financial risk indicators.

[0059] The spatiotemporal two-dimensional adaptive MPC control framework proposed in this invention is as follows:

[0060] 1) Time-dimensional adaptation: Utilizes traffic flow volatility to quantify system instability. Higher volatility corresponds to a shorter forecast timeframe, filtering out unreliable long-term predictions.

[0061] 2) Spatial Dimension Adaptation: Utilizing Value at Risk (V) a R quantifies the tail risk of the prediction model. The wider the distribution of the prediction error (V... a The larger the R value, the larger the "safety buffer" reserved in the physical constraints, thus achieving robust constraints in a probabilistic sense.

[0062] The present invention provides a method for entrance ramp merging control based on a spatiotemporal two-dimensional adaptive model prediction using financial risk indicators, comprising the following steps:

[0063] Step 1: Data Acquisition and Processing

[0064] Traditional approach: Only collects traffic status data at the current moment, including the main road flow rate q. in (k), market share out (k) and ramp queue length l ramp (k).

[0065] Innovation and optimization: Retain basic data collection. Construct a prediction error database: The system compares the "model prediction state" of the previous moment with the "actual detection state" of the current moment in real time, calculates the prediction residuals, and dynamically maintains a historical residual sequence of length L.

[0066] Step 2: Calculation of Key Indicators

[0067] Traditional approach (setting the following fixed parameters at system startup and keeping them constant during system operation):

[0068] Fixed prediction time domain N P =Constant (e.g., fixed at 20 steps). Fixed safety constraint W safe =W physical (For example, fixed at 50 meters).

[0069] Innovation optimization (based on a sliding time window, simultaneously calculating two key metrics):

[0070] Volatility – A trend uncertainty indicator: Calculates the coefficient of variation (CV) of upstream flow within a window M.

[0071]

[0072] Where v(k) is the traffic flow volatility, used to quantify the system's instability. σ flow (k) represents the standard deviation of the upstream flow rate within window M, μ flow Let be the mean of the upstream flow within window M, and ∈ be a very small positive number (e.g., 10). -6 To prevent the denominator from being zero, the higher the volatility, the shorter the forecast timeframe should be to filter out unreliable long-term forecasts.

[0073] Tail risk indicator – Value at Risk (V) a R): Calculate the Value at Risk (VaR) at a confidence level β (e.g., 95%) based on the historical residual series.

[0074] V a R β (k)=Percentile(|Residuals|,β) (2,2)

[0075] Where β is the confidence level.

[0076] Step 3: Parameter Mapping Mechanism (Added Step)

[0077] Traditional approach: This step is omitted (the fixed initialization parameters are used directly).

[0078] Innovative Optimization (Spatiotemporal Two-Dimensional Adaptive Mapping):

[0079] Time dimension: Determine the prediction time domain N P Using a negative correlation mapping mechanism, the higher the volatility, the closer the correlation.

[0080] N P (k)=max(N min ,min(N max ,round(N max -α·V(k)))) (3,1

[0081] Where, N min To predict the minimum value in the time domain, N max The maximum value in the prediction time domain is α, which is an adjustment coefficient that controls the degree of influence of volatility on the prediction time domain.

[0082] Spatial dimension: Determining dynamic security constraints Csafe Subtracting the risk value from the physical limit yields the dynamic safety boundary:

[0083] W safe (k)=W physical -γ·V a R β (k) (3,2)

[0084] Among them W physical Let γ be the actual physical length of the ramp (or the maximum allowable queue length), and γ be the safety factor. This means that if the model's recent predictions are accurate (V... a If R is small, we can make full use of road space; if the model prediction is poor (V... a With R large, the system will automatically reserve a huge buffer zone to prevent overflow.

[0085] Step 4: Controller Construction

[0086] Traditional approach (fixed-dimensional optimization):

[0087] Objective function: (summation upper bound N) P (It is a constant)

[0088]

[0089] Constraints:

[0090] CTM dynamic constraints:

[0091] x(t+1)=f CTM (x(t),u(t)) (4,2)

[0092] Physical hard constraints:

[0093] w ramp (t) <W physical (4,3)

[0094] Enforcement constraints:

[0095] u min ≤u(t)≤u max (4,4)

[0096] Innovative optimization (robust optimization with variable dimensions):

[0097] Objective function:

[0098]

[0099] Where TTS(t) is the total travel time, a part of the objective function, reflecting traffic efficiency. λ is a weighting coefficient used to balance the importance of total travel time and changes in control actions within the objective function. u(t) is the control input. k is the current time step index, and t is the time index.

[0100] Constraints: CTM dynamic constraints:

[0101] x(t+1)=f CTM (x(t),u(t)) (4,6)

[0102] V a R-buffered queuing constraint (optimization point):

[0103] w ramp (t)≤w safe (k) (4,7)

[0104] Right now

[0105] w ramp (t)≤ <W physical -γ·V a R β (k) (4,8)

[0106] Enforcement constraints:

[0107] u min ≤u(t)≤u max (4,9)

[0108] Step 5: Rolling Optimization and Execution

[0109] Traditional approach: Call the solver to solve the above linear programming problem; due to N P The dimensions of the optimization matrix remain constant (e.g., always 20 × the number of variables); the optimal solution u is obtained. * (k); issue u * (k) Give the signal to the signal machine and wait for the next moment.

[0110] Innovative Solution Optimization: Solve the above linear programming problem using a solver (Gurobi / OSQP); obtain the optimal solution u. * (k); issue execution u * (k), repeat steps 1-5 at the next time step.

[0111] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A method for entrance ramp merging control based on a spatiotemporal two-dimensional adaptive model predicting financial risk indicators, characterized in that, Includes the following steps: Step 1: Data Acquisition and Processing: Retain basic data acquisition. Construct a prediction error database; Step 2: Key Indicator Calculation: Based on the sliding time window, calculate both the trend uncertainty indicator and the tail risk indicator. Step 3: Parameter mapping mechanism: adopting a spatiotemporal two-dimensional adaptive mapping; Step 4: Controller construction; Step 5: Rolling optimization and execution.

2. The entrance ramp merging control method based on a spatiotemporal two-dimensional adaptive model predicting financial risk indicators as described in claim 1, characterized in that, Step 1 includes the following steps: The system compares the "model prediction state" of the previous moment with the "actual detection state" of the current moment in real time, calculates the prediction residual, and dynamically maintains a historical residual sequence of length L.

3. The entrance ramp merging control method based on a spatiotemporal two-dimensional adaptive model predicting financial risk indicators as described in claim 1, characterized in that, Step 2 includes the following steps: Trend uncertainty indicator – volatility: Calculates the coefficient of variation of upstream flow within window M. Where v(k) is the traffic flow volatility, used to quantify the system's instability. σ flow (k) represents the standard deviation of the upstream flow rate within window M, μ flow The mean of upstream flow within window M is ∈, where ∈ is a very small positive number to prevent the denominator from being zero. The higher the volatility, the shorter the prediction time domain, in order to filter out unreliable long-term forecasts. Tail risk indicator – Value at Risk (V) a R: Calculate the value at risk at confidence level β based on the historical residual sequence. V a R β (k)=Percentile(|Residuals|,β) (2,2) Where β is the confidence level.

4. The entrance ramp merging control method based on a spatiotemporal two-dimensional adaptive model prediction of financial risk indicators according to claim 1, characterized in that, Step 3 includes the following steps: Spatiotemporal two-dimensional adaptive mapping: Time dimension: Determine the prediction time domain N P Using a negative correlation mapping mechanism, the higher the volatility, the closer the correlation. N P (k)=max(N min ,min(N max ,round(N max -α·V(k)))) (3,1) Where, N min To predict the minimum value in the time domain, N max The maximum value in the prediction time domain is α, which is an adjustment coefficient that controls the degree of influence of volatility on the prediction time domain. Spatial dimension: Determining dynamic security constraints C safe Subtracting the risk value from the physical limit yields the dynamic safety boundary: W safe (k)=W physical -γ·V a R β (k) (3,2) Among them W physical Where γ is the actual physical length of the ramp or the maximum allowable queue length, and γ is the safety factor. This means that if the model's recent predictions are accurate, i.e., V... a A small R value makes full use of road space; if the model's prediction is poor, i.e., V... a When R is large, the system automatically provides a large buffer zone to prevent overflow.

5. The entrance ramp merging control method based on a spatiotemporal two-dimensional adaptive model prediction of financial risk indicators according to claim 1, characterized in that, Step 4 includes the following steps: Objective function: upper bound of summation N P It is a constant. Constraints: CTM dynamic constraints: x(t+1)=f CTM (x(t),u(t)) (4,2) Physical hard constraints: w ramp (t)<W physical (4,3) Enforcement constraints: u min ≤u(t)≤u max (4,4) Robust optimization with varying dimensions: Objective function: Where TTS(t) is the total travel time, which is part of the objective function and reflects traffic efficiency. λ is a weighting coefficient used to balance the importance of total travel time and changes in control actions in the objective function, u(t) is the control input, k is the current time step index, and t is the time index; Constraints: CTM dynamic constraints: x(t+1)=f CTM (x(t),u(t)) (4,6) V a R-buffered queuing constraint (optimization point): w ramp (t)≤w safe (k) (4,7) Right now w ramp (t)≤<W physical -γ·V a R β (k) (4,8) Enforcement constraints: u min ≤u(t)≤u max (4,9)。 6. The entrance ramp merging control method based on a spatiotemporal two-dimensional adaptive model prediction of financial risk indicators according to claim 1, characterized in that, Step 5 includes the following steps: Solve linear programming problems using the solver Gurobi / OSQP; obtain the optimal solution u. * (k); issue execution u * (k), repeat steps 1-5 at the next time step.

Images