Dynamic optimization scheduling method for deep tunnel drainage system based on data driving
By adopting a data-driven dynamic optimization scheduling method, the problem of delayed response of scheduling strategies in deep tunnel drainage systems has been solved, achieving efficient and reliable operation of the system and improving the adaptability and accuracy of scheduling strategies.
Patent Information
- Application Number
- CN202511821799.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-05
- Publication Date
- 2026-03-17
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing deep tunnel drainage system scheduling strategies rely on manual judgment, resulting in delayed response, low system utilization, and difficulty in fully realizing their effectiveness.
By acquiring and processing real-time data, a data-driven dynamic optimization scheduling method is established, including data cleaning, verification, fusion, credibility assessment, proactive probing scheduling, and risk assessment. This constructs an intelligent scheduling decision-making system and optimizes drainage strategies.
It significantly improves the response speed and adaptability of deep tunnel drainage systems, avoids the limitations of human experience, improves system utilization and reliability, and enables precise scheduling for complex working conditions.
Smart Images
Figure CN121684445A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of data processing, and in particular relates to a data-driven dynamic optimization scheduling method for deep tunnel drainage systems. Background Technology
[0002] Deep tunnel drainage systems play a crucial role in addressing urban flooding and controlling combined sewer overflow pollution. However, due to the system's complex structure, current scheduling strategies still rely heavily on manual judgment (e.g., increasing the tunnel's water intake only after an overflow is detected at a key location). This approach results in a delayed response, hindering the full utilization of the deep tunnel drainage system's overall effectiveness and leading to low system utilization rates, thus requiring improvement. Summary of the Invention
[0003] Therefore, it is necessary to provide a data-driven dynamic optimization scheduling method for deep tunnel drainage systems to address the aforementioned problems.
[0004] This invention is implemented as follows: a data-driven dynamic optimization scheduling method for deep tunnel drainage systems includes the following steps:
[0005] Real-time monitoring data (liquid level, flow rate, rainfall, etc.) are collected from various areas (deep tunnels, key pipeline nodes, rivers and key rainfall areas, etc.) to obtain a real-time monitoring dataset. All real-time monitoring data are cleaned, verified and merged to form a unified and usable scheduling dataset.
[0006] Establish a data credibility assessment mechanism to perform real-time cross-validation of scheduling data. Based on the scheduling data of the surrounding areas of the target area, determine whether the scheduling data of the target area is credible. If the scheduling data of the target area is determined to be unreliable, it is marked as abnormal data.
[0007] The current reliable scheduling data is output to the scheduling model to generate a scheduling strategy. In the scheduling strategy, for areas that have been marked as abnormal data, drainage scheduling is not performed. Under the premise of ensuring the completion of the set task (such as draining a specified amount of water within one hour), an active detection scheduling based on risk assessment is initiated for the area to observe the water level changes in the area and the water level changes in the surrounding areas to determine whether there is a data acquisition failure in the area.
[0008] If a data acquisition failure is confirmed in the area, the scheduling data of the area is reconstructed based on the scheduling data of the surrounding areas to obtain an updated scheduling dataset; based on the updated scheduling dataset, the optimized scheduling strategy is recalculated and output through the scheduling model.
[0009] In one embodiment, the present invention provides a data-driven dynamic optimization scheduling method for deep tunnel drainage systems, further comprising:
[0010] The scheduling model is periodically trained and calibrated based on historical operational data (including monitoring data, drainage scheduling strategies and their implementation effects), enabling the scheduling model to continuously learn response patterns under different operating conditions, thereby optimizing the decision-making algorithm and improving the adaptability and accuracy of future drainage scheduling strategies.
[0011] In one embodiment, the present invention provides a data-driven dynamic optimization scheduling method for deep tunnel drainage systems, further comprising:
[0012] The scheduling strategy distinguishes between drainage scheduling and active detection scheduling. For active detection scheduling, stricter control parameters (such as higher water level safety redundancy) are set compared to drainage scheduling, and the state sampling and feedback cycle is shortened.
[0013] In one embodiment, the present invention provides a data-driven dynamic optimization scheduling method for deep tunnel drainage systems, further comprising:
[0014] When there are multiple regions marked as anomalous data, a comprehensive risk assessment model is constructed. The priority of each anomalous region is dynamically calculated based on its risk factors (such as geographical location, importance in the pipeline topology, and potential overflow risk scale), and the execution order of proactive detection scheduling for each anomalous region is determined based on the priority.
[0015] In one embodiment, the present invention provides a data-driven dynamic optimization scheduling method for deep tunnel drainage systems, further comprising:
[0016] After determining that the scheduling data in the target area is unreliable and marking it as abnormal data, external data sources (such as high-precision weather radar and real-time rainfall grid data from urban weather early warning systems) are queried for scenario verification. If the external data confirms that there is a specific environment in the target area that causes the scheduling data to deviate (specific weather events such as localized heavy rainfall), the abnormal data is determined to be a reliable special operating condition, thus avoiding unnecessary active detection scheduling.
[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention obtains an optimized scheduling strategy through real-time data acquisition and processing, which significantly advances the response nodes of the deep tunnel drainage system (such as moving them from "after overflow occurs" to "when rainfall occurs" or even "before rainfall occurs"). This not only effectively avoids the problem of low utilization rate of the deep tunnel drainage system caused by the limitations of human experience and judgment delays, but also significantly enhances the adaptability and reliability of the deep tunnel drainage system under complex working conditions such as data anomalies through proactive predictive scheduling and scheduling data update mechanisms. Attached Figure Description
[0018] Figure 1 This is a schematic diagram of the first part of a data-driven dynamic optimization scheduling method for a deep tunnel drainage system, provided in an embodiment of the present invention.
[0019] Figure 2 This is a schematic diagram of the second part of a data-driven dynamic optimization scheduling method for deep tunnel drainage systems provided in an embodiment of the present invention.
[0020] Figure 3 This is a schematic diagram of the third part of a data-driven dynamic optimization scheduling method for a deep tunnel drainage system, provided as an embodiment of the present invention.
[0021] Figure 4 This is a schematic diagram of the fourth part of a data-driven dynamic optimization scheduling method for a deep tunnel drainage system, provided in an embodiment of the present invention.
[0022] Figure 5 This is a schematic diagram of the fifth part of a data-driven dynamic optimization scheduling method for a deep tunnel drainage system, provided as an embodiment of the present invention. Detailed Implementation
[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0024] In one embodiment, such as Figure 1 As shown, a data-driven dynamic optimization scheduling method for deep tunnel drainage systems includes the following steps:
[0025] Step S1: Collect real-time monitoring data (liquid level, flow rate, rainfall, etc.) from various areas (deep tunnels, key pipeline nodes, rivers and key rainfall areas, etc.) to obtain a real-time monitoring dataset. Clean, verify and merge all real-time monitoring data to form a unified and usable scheduling dataset.
[0026] Step S2: Establish a data credibility assessment mechanism to perform real-time cross-validation on scheduling data. Based on the scheduling data of the surrounding areas of the target area, determine whether the scheduling data of the target area is credible. If the scheduling data of the target area is determined to be unreliable, it is marked as abnormal data.
[0027] Step S3: Output the current reliable scheduling data to the scheduling model to generate a scheduling strategy. In the scheduling strategy, for areas that have been marked as abnormal data, drainage scheduling is not performed. Under the premise of ensuring the completion of the set task (such as draining a specified amount of water within one hour), an active detection scheduling based on risk assessment is initiated for the area to observe the water level changes in the area and the water level changes in the surrounding areas to determine whether there is a data acquisition failure in the area.
[0028] Step S4: If it is confirmed that there is a data acquisition failure in the area, the scheduling data of the area is reconstructed based on the scheduling data of the surrounding areas to obtain an updated scheduling dataset; based on the updated scheduling dataset, the optimized scheduling strategy is recalculated and output through the scheduling model.
[0029] A data-driven dynamic optimization scheduling method for deep tunnel drainage systems is proposed. This method, centered on data-driven approaches, constructs an intelligent scheduling decision-making system to achieve precise control over the entire process of "collection, storage, and discharge" in deep tunnels. Through multi-dimensional data analysis and scheduling model calculations, the method dynamically optimizes scheduling strategies to ensure maximum utilization efficiency of deep tunnels while maintaining flood control and storage safety, effectively achieving the operational goal of "collecting all that should be collected and controlling emergency discharge to the maximum extent."
[0030] By dynamically calculating using a scheduling model, a collaborative scheduling strategy is generated when it is predicted that future inflow will exceed the storage capacity limit. This scheduling strategy is then translated into control commands for the controllable execution devices and ultimately issued to each device. For example:
[0031] (1) Gate control: The gate opening is dynamically adjusted according to the liquid level in front of the gate and the condition of the upstream flood-prone points, so as to maximize the storage within the allowable range of the reservoir capacity and avoid upstream overflow;
[0032] (2) Emptying pump scheduling: Based on the inflow prediction and real-time analysis of reservoir capacity, the system intelligently determines the optimal start-up and shutdown times of the emptying pumps and the pump operation combinations. Through the optimized operation mode of "storing and discharging simultaneously", the system achieves dynamic emptying and recycling of reservoir capacity, significantly improving the storage capacity of the deep tunnel.
[0033] (3) Drainage pump control: When the reservoir capacity approaches saturation and the risk of flooding continues to rise, the drainage pumps are automatically activated; when the rainfall weakens and the deep tunnel drainage system can achieve self-balancing, the drainage pumps are automatically shut down, and the remaining reservoir capacity is transferred to the sewage treatment plant through the emptying pumps. Through this precise control, the frequency of emergency discharges is reduced to the lowest level while ensuring flood control safety, achieving the optimal balance between flood control safety and emission reduction benefits.
[0034] Before step S1, a large number of level gauges, flow meters, rain gauges, and water quality sensors are deployed in areas such as deep tunnels, key pipeline nodes, rivers, and key rainfall points. These devices are responsible for collecting multi-dimensional data in real time, including rainfall intensity, pipeline fullness, river water level, and key water quality indicators (such as ammonia nitrogen and suspended solids).
[0035] In step S2, for example, when the level gauge of a node (node A) displays abnormal data (such as a persistent zero or excessively high level), the level and flow data of the upstream and downstream of that node will be immediately retrieved and combined with the real-time radar rainfall map for comprehensive analysis. If the meteorological data shows that the area is experiencing heavy rainfall and the water levels of both upstream and downstream nodes have risen significantly, it will be preliminarily determined that the data of node A has a high risk of distortion, and the data will be labeled as abnormal data.
[0036] In steps S3 and S4, for example, the opening of the upstream gate is slightly adjusted. By observing whether the changes in water levels upstream and downstream of the area marked as abnormal data conform to the predicted changes, and the water level response in the area, it is determined whether there is a data acquisition anomaly in the area. If an anomaly is confirmed, the scheduling data of the area is reconstructed based on the scheduling data of the upstream and downstream areas. In the scheduling strategy, steps S3 and S4 can be repeated. That is, when the optimized scheduling strategy is implemented, if a new area marked as abnormal data appears, steps S3 and S4 are repeated.
[0037] It can further differentiate the types of data anomalies, such as: persistent freeze (complete sensor failure), sudden spikes (drastic changes in readings), and slow drift (gradual data distortion). Different proactive detection schedules can be preset for each anomaly type. For example, for "slow drift," the detection command can be more subtle; for "sudden spikes," verification can be initiated immediately and the data temporarily isolated, thereby improving the accuracy and efficiency of diagnosis.
[0038] In one embodiment, such as Figure 2 As shown, a data-driven dynamic optimization scheduling method for deep tunnel drainage systems further includes:
[0039] Step S5 involves periodically training and calibrating the scheduling model based on historical operational data (including monitoring data, drainage scheduling strategies, and their execution effects). This allows the scheduling model to continuously learn response patterns under different operating conditions, thereby optimizing the decision-making algorithm and improving the adaptability and accuracy of future drainage scheduling strategies.
[0040] The scheduling model is constructed based on a digital abstraction of the physical characteristics of the drainage system, employing a hydraulic-hydrological model (such as SWMM) as its core computational engine to simulate the dynamic behavior of water flow in pipe networks and deep tunnels. During training, the scheduling model uses historical operational data as learning samples, continuously refining its internal parameters by comparing the model's predictions with the actual system response. Step S5 implements closed-loop optimization: periodically feeding new operational data back into the training process, enabling the scheduling model to continuously learn the system characteristics and strategy effects under different operating conditions, thereby iteratively improving the adaptability and accuracy of its decision-making algorithm.
[0041] In one embodiment, such as Figure 3 As shown, a data-driven dynamic optimization scheduling method for deep tunnel drainage systems further includes:
[0042] Step S6: Distinguish between drainage scheduling and active detection scheduling in the scheduling strategy. For active detection scheduling, set stricter control parameters (such as higher water level safety redundancy) than drainage scheduling, and shorten the state sampling and feedback cycle.
[0043] In step S6, a tiered control mechanism ensures the safety and controllability of the active detection process. For example, when active detection needs to be initiated for a pumping station area with abnormal data, compared to the conventional drainage scheduling which allows the water level at that node to reach 90% of the warning line before triggering an alarm, the safe water level threshold for this detection will be tightened to 70% of the warning line. Simultaneously, the sampling frequency of the node and its associated pipe network status data will be increased from once every 5 minutes to once every 30 seconds. Throughout the execution of detection commands (such as fine-tuning the opening of its upstream gate from 80% to 70%), the rate of water level change at that node and the upstream and downstream pressure balance will be continuously monitored. Once the water level rise rate is detected to exceed the preset strict standard or any associated node shows abnormal fluctuations, the detection will be interrupted and the original gate state restored within seconds, thus ensuring that operational risks are always kept within safe boundaries while diagnosing data anomalies.
[0044] Step S6 aims to address the secondary operational risks that active probing may introduce. Since active probing is essentially a tentative intervention in deep tunnel drainage systems, its uncertainties are higher than conventional drainage scheduling. By establishing stricter control parameters and more intensive status monitoring, a safety boundary can be constructed to ensure that potential interference with the overall stability of the deep tunnel drainage system is minimized while diagnosing data faults.
[0045] In one embodiment, such as Figure 4 As shown, a data-driven dynamic optimization scheduling method for deep tunnel drainage systems further includes:
[0046] Step S7: When there are multiple regions marked as abnormal data, construct a comprehensive risk assessment model, dynamically calculate the priority of each abnormal region based on the risk factors of each abnormal region (such as geographical location, importance in the pipeline topology and potential overflow risk scale, etc.), and determine the execution order of active detection scheduling for each abnormal region based on the priority.
[0047] The comprehensive risk assessment model first normalizes each risk factor: It establishes a quantitative system for risk factors, assigning different scores based on regional attributes for geographical location risk—densely populated areas and transportation hubs are scored 10 points, urban parks 6 points, and industrial areas 3 points. For the importance of the pipeline topology, the betweenness centrality of nodes is calculated using graph theory algorithms, with critical nodes located at the intersection of multiple drainage paths scored 10 points and terminal nodes of branch lines scored 3 points. For potential overflow risk, it is graded based on the ratio of real-time liquid level to design capacity: 90% is scored 10 points, 80% is scored 8 points, and so on.
[0048] Subsequently, the weighting ratios are dynamically adjusted based on the real-time weather warning levels: when an orange rainstorm warning is issued, the overflow risk weight is increased to 60%, topological importance to 30%, and geographic location to 10%; during normal periods, a balanced weighting (40%, 40%, 20%) is used. Finally, a priority ranking list for each anomalous area is automatically generated using the weighted summation formula: Priority Score = Geographical Location Score × W1 + Topological Importance Score × W2 + Overflow Risk Score × W3, where W1, W2, and W3 are the weights mentioned above (40%, 40%, 20% or 60%, 30%, 10%), ensuring that the anomalous areas with the highest overall risk are addressed first with limited resources.
[0049] Step S7 aims to address resource contention and decision-making efficiency issues when multiple regions experience concurrent anomalies. When faced with multiple anomaly regions, a lack of prioritization can lead to wasted scheduling resources on less important regions. By constructing a risk assessment model and dynamically prioritizing resources, limited diagnostic and processing resources are allocated to critical nodes with the greatest impact on overall security and the highest risk, thereby optimizing emergency response.
[0050] In one embodiment, such as Figure 5 As shown, a data-driven dynamic optimization scheduling method for deep tunnel drainage systems further includes:
[0051] Step S8: After determining that the scheduling data in the target area is unreliable and marking it as abnormal data, query external data sources (such as high-precision meteorological radar, real-time rainfall grid data of urban meteorological early warning system) for scenario verification. If the external data confirms that there is a specific environment in the target area that causes the scheduling data to deviate (specific weather events such as local heavy rainfall), the abnormal data is determined to be a reliable special working condition to avoid unnecessary active detection scheduling.
[0052] In addition, after receiving accurate short-term heavy rain forecasts from the meteorological observatory, the capacity of deep tunnels can be pre-emptively drained to prepare for flood peaks. Data interfaces and coordination mechanisms with other key urban infrastructures (such as intelligent transportation systems and power supply networks) can also be established. For example, when the transportation system detects severe water accumulation in a certain area, it can provide auxiliary verification of the ground disaster situation for the drainage system and coordinate the adjustment of traffic signals to guide vehicles to detour.
[0053] Step S8 aims to prevent the system from over-diagnosing real-world special operating conditions. While localized heavy rainfall and other weather phenomena may cause data anomalies, they are still reliable. Initiating proactive probing scheduling indiscriminately would waste resources and cause response delays. By introducing external data sources for scenario verification, the system intelligently distinguishes between data faults and real special events, avoiding misjudgments while ensuring the timeliness and effectiveness of scheduling strategies under real-world special weather conditions.
[0054] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.
[0055] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0056] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.
[0057] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
[0058] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.
Claims
1. A data-driven deep tunnel drainage system dynamic optimization scheduling method, characterized in that, The data-driven deep tunnel drainage system dynamic optimization scheduling method comprises the following steps: Collect real-time monitoring data of each region, obtain real-time monitoring data set, clean, correct and fuse all real-time monitoring data to form unified and usable scheduling data set; A data credibility evaluation mechanism is established to cross-validate the scheduling data in real time. Based on the scheduling data of the surrounding areas of the target region, it is judged whether the scheduling data of the target region is credible. If it is judged that the scheduling data of the target region is not credible, it is marked as abnormal data; The current credible scheduling data is output to the scheduling model to generate a scheduling strategy. In the scheduling strategy, for the region that has been marked as abnormal data, no drainage scheduling is performed, and an active detection type scheduling based on risk assessment is started for the region under the premise of ensuring the completion of the set task. Observe the water level change of the region and the water level change of the surrounding areas, and judge whether the region has data acquisition failure; If it is confirmed that the region has data acquisition failure, the scheduling data of the region is reconstructed based on the scheduling data of the surrounding areas of the region, and the updated scheduling data set is obtained; based on the updated scheduling data set, the scheduling model is used to recalculate and output the optimized scheduling strategy.
2. The data-driven deep tunnel drainage system dynamic optimization scheduling method according to claim 1, wherein, Further comprising: Periodically train and calibrate the scheduling model based on historical operation data, so that the scheduling model continuously learns the response law under different working conditions, thereby optimizing the decision algorithm and improving the adaptability and accuracy of future drainage scheduling strategy.
3. The data-driven deep tunnel drainage system dynamic optimization scheduling method according to claim 1, wherein, Further comprising: The drainage scheduling and active detection type scheduling of the scheduling strategy are distinguished. For the active detection type scheduling, more stringent control parameters are set compared with the drainage scheduling, and the state sampling and feedback period is shortened.
4. The data-driven deep tunnel drainage system dynamic optimization scheduling method according to any one of claims 1 to 3, characterized in that, Further comprising: When there are multiple regions that have been marked as abnormal data, a comprehensive risk assessment model is constructed, the priority of each abnormal region is dynamically calculated according to the risk factors of each abnormal region, and the execution order of the active detection type scheduling of each abnormal region is determined based on the priority.
5. The data-driven deep tunnel drainage system dynamic optimization scheduling method according to claim 4, characterized in that, Further comprising: After judging that the scheduling data of the target region is not credible and marking it as abnormal data, the external data source is queried for scenario verification. If the external data confirms that the target region has a specific environment that causes the scheduling data to deviate, the abnormal data is determined as credible special working condition to avoid unnecessary active detection type scheduling.