Metro space smoke exhaust air velocity detection optimization method and system based on multiple sensors

Abstract

The application provides a subway space smoke exhaust air velocity detection optimization method and system based on multiple sensors, and relates to the technical field of subway ventilation. The method comprises the following steps: obtaining subway space design information and subway space real-time data; setting each fan according to the fan parameters of each fan determined by the fan parameter control model of the i-1th optimization period in the i th optimization period; obtaining wind speed data and smoke concentration data; determining estimated wind speed data and estimated smoke concentration data; determining the overall smoke exhaust effect coefficient; judging whether the fan parameter control model of the i-1th optimization period needs to be trained; determining the training loss function; obtaining the fan parameter control model of the i th optimization period, and setting each fan according to the fan parameters of each fan determined by the fan parameter control model of the i th optimization period in the i+1th optimization period. According to the application, the stability and accuracy of the subway space smoke exhaust air velocity detection optimization can be improved.

Description

Technical Field

[0001] This invention relates to the field of subway ventilation technology, and in particular to a method and system for optimizing the detection of smoke exhaust wind speed in subway spaces based on multiple sensors. Background Technology

[0002] In related technologies, the detection and optimization of smoke exhaust velocity in subway spaces mainly relies on professional security personnel to perform manual optimization using sensors. In other words, it is mainly dependent on human factors, and the workload of detection and optimization is enormous. Therefore, over-reliance on human factors may make it difficult to accurately identify safety hazards, resulting in low efficiency in the detection and optimization of smoke exhaust velocity.

[0003] The information disclosed in the background section of this application is intended only to enhance the understanding of the general background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0004] This invention provides a multi-sensor-based method and system for detecting and optimizing smoke exhaust velocity in subway spaces, which can solve the technical problems of difficulty in accurately identifying safety hazards and low efficiency in detecting and optimizing smoke exhaust velocity.

[0005] According to a first aspect of the present invention, a multi-sensor-based method for optimizing smoke exhaust velocity detection in subway spaces is provided, comprising:

[0006] Obtain subway space design information and real-time subway space data;

[0007] In the i-th optimization cycle, each fan is set according to the fan parameters determined by the fan parameter control model of the (i-1)-th optimization cycle, wherein the fan parameters include: fan speed and damper opening.

[0008] In the i-th optimization cycle, wind speed data and smoke concentration data are acquired by a distributed multi-sensor array set on key nodes;

[0009] Based on the wind speed data and the smoke concentration data, the estimated wind speed data and estimated smoke concentration data for the simulated measurement points are determined;

[0010] Based on the subway space design information, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, the overall smoke extraction effect coefficients at multiple moments in the i-th optimization cycle are determined.

[0011] Based on the overall smoke extraction effect coefficient, determine whether the fan parameter control model for the (i-1)th optimization cycle needs to be trained.

[0012] If the fan parameter control model for the (i-1)th optimization cycle needs to be trained, the training loss function of the fan parameter control model for the (i-1)th optimization cycle is determined based on the subway space design information, the real-time subway space data, and the overall smoke exhaust effect coefficient.

[0013] The wind turbine parameter control model for the (i-1)th optimization period is trained using the training loss function of the wind turbine parameter control model for the (i-1)th optimization period to obtain the wind turbine parameter control model for the ith optimization period. In the (i+1)th optimization period, each wind turbine is set according to the wind turbine parameters determined by the wind turbine parameter control model for the ith optimization period.

[0014] According to the present invention, determining the estimated wind speed data and estimated smoke concentration data for a simulated measurement point based on the wind speed data and the smoke concentration data includes:

[0015] Establish a BIM model of the subway space;

[0016] By using BIM models and spatial interpolation algorithms, estimated wind speed and smoke concentration data were determined for each simulated measurement point in the subway space.

[0017] According to the present invention, based on the subway space design information, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, the overall smoke extraction effect coefficient at multiple moments in the i-th optimization cycle is determined, including:

[0018] Based on the subway space design information, determine the first smoke exhaust distance from each key node to the nearest smoke exhaust outlet and the second smoke exhaust distance from each simulated measurement point to the nearest smoke exhaust outlet.

[0019] Based on the wind speed data and the estimated wind speed data, a first average wind speed, a first wind speed standard deviation, a second average wind speed, and a second wind speed standard deviation are determined.

[0020] Based on the first smoke exhaust distance, the second smoke exhaust distance, the first average wind speed, the first wind speed standard deviation, the second average wind speed, the second wind speed standard deviation, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, the overall smoke exhaust effect coefficient at multiple moments in the i-th optimization cycle is determined.

[0021] According to the present invention, the overall smoke extraction effect coefficient at multiple moments in the i-th optimization period is determined based on the first smoke extraction distance, the second smoke extraction distance, the first average wind speed, the first wind speed standard deviation, the second average wind speed, the second wind speed standard deviation, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, including: according to the formula: ,

[0022] Determine the overall smoke extraction efficiency coefficient at time k in the i-th optimization cycle. ,in, Let the standard deviation of the first wind speed be the value at the k-th moment of the i-th optimization period. Let be the first average wind speed at the k-th moment of the i-th optimization cycle. Let the standard deviation of the second wind speed be the value at the k-th moment of the i-th optimization period. Let be the second average wind speed at time k in the i-th optimization cycle, and let 'if' be a conditional function. The smoke concentration data for the e-th critical node at time k in the i-th optimization cycle is given. To preset the smoke concentration threshold, Let the wind speed data for the e-th critical node at the k-th time in the i-th optimization cycle be given. To preset the wind speed data threshold, The distance to the first smoke exhaust at the e-th critical node. To preset the smoke exhaust distance threshold, For the estimated smoke concentration data of the r-th simulated measurement point at the k-th time in the i-th optimization cycle, For the estimated wind speed data of the r-th simulated measurement point at the k-th time in the i-th optimization cycle, Let R be the second smoke exhaust distance of the r-th simulated measurement point, R be the number of simulated measurement points, r ≤ R, E be the number of key nodes, e ≤ E, and e, E, r and R are all positive integers.

[0023] According to the present invention, determining whether the fan parameter control model for the (i-1)th optimization cycle needs to be trained based on the overall smoke extraction effect coefficient includes:

[0024] The first judgment condition is determined based on the minimum value of the overall smoke extraction effect coefficient at multiple moments in the i-th optimization cycle.

[0025] Obtain the standard deviation and average value of the overall smoke extraction effect coefficient at multiple times during the i-th optimization cycle;

[0026] The second judgment condition is determined based on the judgment standard deviation and the judgment average.

[0027] Based on the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, determine the smoke extraction effect coefficient at a single location for each key node and simulated measurement point at multiple times in the i-th optimization cycle;

[0028] The third judgment condition is determined based on the smoke extraction effect coefficient of the single location;

[0029] If any one of the first judgment condition, the second judgment condition, and the third judgment condition is met, it is determined that the wind turbine parameter control model for the (i-1)th optimization cycle needs to be trained.

[0030] According to the present invention, when the fan parameter control model for the (i-1)th optimization cycle needs to be trained, the training loss function of the fan parameter control model for the (i-1)th optimization cycle is determined based on the subway space design information, the real-time subway space data, and the overall smoke extraction effect coefficient, including:

[0031] Among the first judgment condition, the second judgment condition and the third judgment condition, determine the training conditions that the smoke exhaust effect coefficient of a single location and the overall smoke exhaust effect coefficient must satisfy.

[0032] Based on the subway space design information, the importance identification results of each key node and simulated measurement point are determined;

[0033] Based on the real-time subway space data, the real-time passenger flow of the subway is determined;

[0034] Based on the smoke extraction effect coefficient of a single location, the real-time passenger flow of the subway, the importance identification result, the overall smoke extraction effect coefficient, and the satisfaction of the training conditions, the training loss function of the fan parameter control model for the (i-1)th optimization cycle is determined.

[0035] According to the present invention, the training loss function of the fan parameter control model for the (i-1)th optimization cycle is determined based on the individual location smoke extraction effect coefficient, the real-time passenger flow of the subway, the importance identification result, the overall smoke extraction effect coefficient, and the satisfaction of the training conditions, including: according to the formula: ,

[0036] Determine the training loss function of the wind turbine parameter control model for the (i-1)th optimization cycle. Where, min is the minimum value function, and if is the conditional function, where, This is the first condition for judgment. This is the second condition for judgment. This is the third condition for judgment. Let be the overall smoke extraction efficiency coefficient at time k in the i-th optimization cycle. To preset the overall smoke extraction effect coefficient threshold, For the real-time passenger flow of the subway in the i-th optimization period, To preset the passenger flow threshold, Let be the standard deviation of the overall smoke extraction efficiency coefficient at multiple moments in the i-th optimization cycle. Let be the average value of the overall smoke extraction efficiency coefficient at multiple moments during the i-th optimization cycle. Let be the smoke extraction efficiency coefficient of the e-th critical node at the k-th time in the i-th optimization cycle. Let r be the smoke extraction efficiency coefficient at a single location of r simulated measurement points at time k in the i-th optimization cycle. To preset the threshold value of the smoke extraction effect coefficient for a single location, The importance identification result corresponding to the smallest single location smoke extraction effect coefficient is given by R, where R is the number of simulated measurement points, r≤R, E is the number of key nodes, e≤E, and K is the number of optimization cycle moments, k≤K. e, E, r, R, k, and K are all positive integers.

[0037] According to a second aspect of the present invention, a multi-sensor-based subway space smoke exhaust wind speed detection and optimization system is provided, comprising:

[0038] The subway data module acquires subway space design information and real-time subway space data.

[0039] In the initial setup module, during the i-th optimization cycle, each fan is set according to the fan parameters determined by the fan parameter control model of the (i-1)-th optimization cycle. The fan parameters include: fan speed and damper opening.

[0040] The detection data module is used to acquire wind speed data and smoke concentration data through a distributed multi-sensor array set on key nodes during the i-th optimization cycle.

[0041] The estimation data module is used to determine the estimated wind speed data and estimated smoke concentration data of the simulated measurement point based on the wind speed data and the smoke concentration data.

[0042] The smoke extraction coefficient module is used to determine the overall smoke extraction effect coefficient at multiple moments in the i-th optimization cycle based on the subway space design information, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data.

[0043] The training module is used to determine whether the fan parameter control model for the (i-1)th optimization cycle needs to be trained based on the overall smoke exhaust effect coefficient.

[0044] The loss function module is used to determine the training loss function of the fan parameter control model for the (i-1)th optimization period based on the subway space design information, the real-time subway space data, and the overall smoke exhaust effect coefficient, when the fan parameter control model needs to be trained in the (i-1)th optimization period.

[0045] The optimization setting module is used to train the wind turbine parameter control model for the (i-1)th optimization period according to the training loss function of the wind turbine parameter control model for the (i-1)th optimization period, to obtain the wind turbine parameter control model for the ith optimization period, and in the (i+1)th optimization period, to set the wind turbine parameters of each wind turbine according to the wind turbine parameter control model for the ith optimization period.

[0046] Technical Effects: According to the present invention, wind speed data and smoke concentration data of key nodes in the subway space, as well as estimated wind speed data and estimated smoke concentration data of simulated measurement points, can be accurately obtained. Based on the wind speed data, smoke concentration data, estimated wind speed data, and estimated smoke concentration data, the overall wind speed smoke extraction effect in the subway space can be evaluated, and the overall smoke extraction effect coefficient can be determined. Furthermore, based on the overall smoke extraction effect coefficient, the training loss function of the fan parameter control model can be determined. The fan parameter control model is trained based on the training loss function, so that when the fan parameters are controlled by the fan parameter control model, the wind speed smoke extraction effect in the subway space reaches the optimal state. When determining the overall smoke extraction effect coefficient, the overall smoke extraction effect coefficient for multiple moments in the i-th optimization cycle is determined based on the first smoke extraction distance, the second smoke extraction distance, the first average wind speed, the first wind speed standard deviation, the second average wind speed, the second wind speed standard deviation, wind speed data, smoke concentration data, estimated wind speed data, and estimated smoke concentration data. During the calculation process, the uniformity of wind speed distribution can be accurately analyzed, and the smoke extraction effect of wind speed is analyzed separately for cases where the smoke concentration meets or does not meet the standard. Furthermore, based on the uniformity of wind speed distribution and the smoke extraction effect of wind speed, the overall smoke extraction effect coefficient is determined, improving the comprehensiveness and accuracy of the overall smoke extraction effect coefficient. When determining the training loss function of the fan parameter control model, the training loss function of the fan parameter control model for the (i-1)th optimization cycle can be determined based on the smoke exhaust effect coefficient of a single location, the real-time passenger flow of the subway, the importance identification results, the overall smoke exhaust effect coefficient, and the satisfaction of the training conditions. Training the fan parameter control model according to the training loss function can make the fan parameter control model more stable, make the fan parameter control model more adaptable to peak passenger flow conditions, and improve the accuracy of the fan parameter control model for fan parameter control at more important locations.

[0047] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Other features and aspects of the invention will become clearer from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.

[0049] Figure 1 An exemplary flowchart of a multi-sensor-based method for optimizing smoke exhaust velocity detection in subway spaces according to an embodiment of the present invention is shown.

[0050] Figure 2 An exemplary schematic diagram illustrating the determination of estimated wind speed data and estimated smoke concentration data at simulated measurement points according to an embodiment of the present invention is shown.

[0051] Figure 3 An exemplary schematic diagram illustrating the determination of the overall smoke extraction efficiency coefficient according to an embodiment of the present invention is shown;

[0052] Figure 4 An exemplary schematic diagram illustrating the determination of whether a wind turbine parameter control model needs training according to an embodiment of the present invention is provided.

[0053] Figure 5 An exemplary diagram illustrates the training loss function of the wind turbine parameter control model for the (i-1)th optimization cycle according to an embodiment of the present invention;

[0054] Figure 6 A block diagram of a multi-sensor-based subway space smoke exhaust wind speed detection and optimization system according to an embodiment of the present invention is shown as an example. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0057] Figure 1 An exemplary flowchart illustrates a multi-sensor-based method for optimizing smoke exhaust velocity detection in subway spaces according to an embodiment of the present invention. The method includes:

[0058] Step S1: Obtain subway space design information and real-time subway space data;

[0059] Step S2: In the i-th optimization cycle, set the fan parameters of each fan according to the fan parameter control model determined in the (i-1)-th optimization cycle. The fan parameters include: fan speed and damper opening.

[0060] Step S3: In the i-th optimization cycle, wind speed data and smoke concentration data are acquired by a distributed multi-sensor array set on key nodes.

[0061] Step S4: Based on the wind speed data and the smoke concentration data, determine the estimated wind speed data and estimated smoke concentration data for the simulated measurement point;

[0062] Step S5: Based on the subway space design information, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, determine the overall smoke extraction effect coefficient for multiple moments in the i-th optimization cycle;

[0063] Step S6: Based on the overall smoke extraction effect coefficient, determine whether the fan parameter control model for the (i-1)th optimization cycle needs to be trained.

[0064] Step S7: If the fan parameter control model for the (i-1)th optimization cycle needs to be trained, determine the training loss function of the fan parameter control model for the (i-1)th optimization cycle based on the subway space design information, the real-time subway space data, and the overall smoke exhaust effect coefficient.

[0065] Step S8: Train the wind turbine parameter control model for the (i-1)th optimization period according to the training loss function of the wind turbine parameter control model for the (i-1)th optimization period to obtain the wind turbine parameter control model for the ith optimization period. In the (i+1)th optimization period, set the wind turbine parameters of each wind turbine according to the wind turbine parameter control model for the ith optimization period.

[0066] According to the multi-sensor-based subway space smoke exhaust wind speed detection optimization method of the present invention, the wind speed data and smoke concentration data of key nodes in the subway space, as well as the estimated wind speed data and estimated smoke concentration data of simulated measurement points, can be accurately acquired. Based on the wind speed data, smoke concentration data, estimated wind speed data, and estimated smoke concentration data, the overall wind speed smoke exhaust effect in the subway space is evaluated, and the overall smoke exhaust effect coefficient is determined. Furthermore, based on the overall smoke exhaust effect coefficient, the training loss function of the fan parameter control model is determined, and the fan parameter control model is trained according to the training loss function, so that when the fan parameters are controlled by the fan parameter control model, the wind speed smoke exhaust effect in the subway space reaches the optimal state.

[0067] According to one embodiment of the present invention, in step S1, subway space design information and real-time subway space data are obtained.

[0068] For example, subway space design information (such as the distance from various locations in the subway space to the ventilation vents) can be obtained through subway design documents, and real-time subway space data (such as real-time subway passenger flow) can be obtained through official subway platforms.

[0069] According to an embodiment of the present invention, in step S2, in the i-th optimization cycle, each fan is set according to the fan parameters of each fan determined by the fan parameter control model of the (i-1)-th optimization cycle, wherein the fan parameters include: fan speed and air valve opening.

[0070] For example, in the second optimization cycle, each fan is set according to the fan parameters determined by the fan parameter control model of the (i-1)th optimization cycle. In the first optimization cycle, each fan is set according to the initial fan parameters. An optimization cycle can be set to 15 minutes. The fan parameter control model is a BP neural network model. The wind speed data and smoke concentration data in the subway space are input into the fan parameter control model. The fan parameter control model can output the fan speed and air valve opening of each fan in the subway space.

[0071] According to an embodiment of the present invention, in step S3, during the i-th optimization cycle, wind speed data and smoke concentration data are acquired by a distributed multi-sensor array set on key nodes.

[0072] For example, a distributed multi-sensor array is deployed along the longitudinal direction of the tunnel and at key nodes (such as platform ends, tunnel midpoints, curves, connecting passages, and near smoke exhaust outlets). Each array integrates a wind speed sensor and a smoke concentration sensor. Through the distributed multi-sensor array, wind speed data and smoke concentration data of each key node are obtained.

[0073] According to one embodiment of the present invention, in step S4, the estimated wind speed data and estimated smoke concentration data of the simulated measurement point are determined based on the wind speed data and the smoke concentration data.

[0074] Figure 2 A schematic diagram illustrating the determination of estimated wind speed data and estimated smoke concentration data at simulated measurement points according to an embodiment of the present invention is shown.

[0075] According to an embodiment of the present invention, step S4 includes:

[0076] Step S41: Establish a BIM model of the subway space;

[0077] Step S42: Using the BIM model and spatial interpolation algorithm, determine the estimated wind speed data and estimated smoke concentration data for each simulated measurement point in the subway space.

[0078] For example, a three-dimensional digital module of the subway space, namely a BIM model, is constructed based on the subway space design information. The tunnel space is discretized into multiple virtual grid points. Then, using a spatiotemporal interpolation algorithm (such as Kriging interpolation), the estimated values ​​on each virtual grid point (i.e., the simulated measurement point) are calculated based on the wind speed data and smoke concentration data of the surrounding known key nodes. That is, the estimated wind speed data and estimated smoke concentration data of each simulated measurement point.

[0079] According to an embodiment of the present invention, in step S5, the overall smoke extraction effect coefficient at multiple moments in the i-th optimization cycle is determined based on the subway space design information, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data.

[0080] Figure 3 A schematic diagram illustrating the determination of the overall smoke extraction efficiency coefficient according to an embodiment of the present invention is shown.

[0081] According to an embodiment of the present invention, step S5 includes:

[0082] Step S51: Based on the subway space design information, determine the first smoke exhaust distance from each key node to the nearest smoke exhaust outlet and the second smoke exhaust distance from each simulated measurement point to the nearest smoke exhaust outlet.

[0083] Step S52: Based on the wind speed data and the estimated wind speed data, determine the first average wind speed, the first wind speed standard deviation, the second average wind speed, and the second wind speed standard deviation.

[0084] Step S53: Based on the first smoke exhaust distance, the second smoke exhaust distance, the first average wind speed, the first wind speed standard deviation, the second average wind speed, the second wind speed standard deviation, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, determine the overall smoke exhaust effect coefficient for multiple moments in the i-th optimization cycle.

[0085] For example, based on the subway space design information, the first smoke exhaust distance from each key node to the nearest smoke exhaust outlet and the second smoke exhaust distance from each simulated measurement point to the nearest smoke exhaust outlet are determined; the average and standard deviation of wind speed data for multiple key nodes are calculated, i.e., the first average wind speed and the first standard deviation of wind speed; the average and standard deviation of wind speed for multiple simulated measurement points are calculated, i.e., the second average wind speed and the second standard deviation of wind speed; based on the first smoke exhaust distance, the second smoke exhaust distance, the first average wind speed, the first standard deviation of wind speed, the second average wind speed, the second standard deviation of wind speed, wind speed data, smoke concentration data, estimated wind speed data, and estimated smoke concentration data, the overall smoke exhaust effect of the subway space is evaluated, and the overall smoke exhaust effect coefficients for multiple moments in the i-th optimization cycle are determined.

[0086] According to an embodiment of the present invention, step S53 includes: determining the overall smoke extraction efficiency coefficient at the k-th moment of the i-th optimization period according to formula (1). , (1),

[0087] in, Let the standard deviation of the first wind speed be the value at the k-th moment of the i-th optimization period. Let be the first average wind speed at the k-th moment of the i-th optimization cycle. Let the standard deviation of the second wind speed be the value at the k-th moment of the i-th optimization period. Let be the second average wind speed at time k in the i-th optimization cycle, and let 'if' be a conditional function. The smoke concentration data for the e-th critical node at time k in the i-th optimization cycle is given. To preset the smoke concentration threshold, Let the wind speed data for the e-th critical node at the k-th time in the i-th optimization cycle be given. To preset the wind speed data threshold, The distance to the first smoke exhaust at the e-th critical node. To preset the smoke exhaust distance threshold, For the estimated smoke concentration data of the r-th simulated measurement point at the k-th time in the i-th optimization cycle, For the estimated wind speed data of the r-th simulated measurement point at the k-th time in the i-th optimization cycle, Let R be the second smoke exhaust distance of the r-th simulated measurement point, R be the number of simulated measurement points, r ≤ R, E be the number of key nodes, e ≤ E, and e, E, r and R are all positive integers.

[0088] According to one embodiment of the present invention, The ratio of the first wind speed standard deviation to the first average wind speed at time k in the i-th optimization cycle is larger. The larger this ratio is, the more uneven the distribution of wind speed data in the subway space is, which will lead to energy waste and insufficient effective air volume (the system needs to consume a lot of electrical energy to maintain a standard average wind speed (e.g., 2.30 m / s). However, this energy is mainly concentrated on the flow in the high-speed area, while the air volume in the low-speed area is seriously insufficient, resulting in a situation where some wind is idling and other wind is not available. In fact, the air volume used for effective smoke exhaust is far lower than the theoretical amount), and the smaller the overall smoke exhaust effect coefficient is.

[0089] According to an embodiment of the present invention, in formula (1), the condition function The value includes the following two cases, when the following conditions are met: When the smoke concentration data of the e-th key node at time k in the i-th optimization cycle is less than the preset smoke concentration threshold, it indicates that the smoke concentration at the e-th key node meets the preset standard, and the value of the condition function is the inner condition function. The value, when it does not satisfy When the condition is met, it indicates that the smoke concentration at e key nodes does not meet the preset standard, the smoke concentration at that key node is too high, and the smoke extraction effect of the wind speed at that location needs to be evaluated. The value of the condition function is... , This is the ratio of the wind speed data to a preset wind speed threshold. The larger the ratio, the higher the relative wind speed. When the smoke concentration is too high, a higher wind speed results in better smoke extraction. This is the ratio of the first smoke exhaust distance at the e-th critical node to the preset smoke exhaust distance threshold. The larger this ratio, the farther the first smoke exhaust distance, the more difficult it is for the smoke at the e-th critical node to be exhausted, and the greater the contribution of the wind speed at this critical node to the smoke removal task, resulting in a better smoke exhaust effect. The preset smoke exhaust distance threshold is determined based on the average distance from all critical nodes and simulated measurement points to the nearest smoke exhaust outlet. This indicates the smoke extraction effect at key nodes, determined based on the first smoke extraction distance and wind speed data.

[0090] According to an embodiment of the present invention, the inner condition function The value includes the following two cases, when the following conditions are met: When the condition is met, it means that the wind speed data of the e-th key node at time k in the i-th optimization cycle meets the preset standard, and the value of the inner condition function is... In the absence of satisfaction When the condition is met, the value of the inner condition function is... , The largest difference between the wind speed data of the e-th key node at the k-th moment of the i-th optimization cycle and the preset wind speed data threshold is the larger the ratio. The larger this ratio is, the more or less the wind speed data is too large or too small. Too large a wind speed data will cause energy waste, while too small a wind speed data will form a dead zone, which may become an area for dust and floating objects to settle and accumulate, increasing the difficulty and frequency of daily cleaning and maintenance.

[0091] According to one embodiment of the present invention, This indicates the overall smoke extraction effect at key nodes determined based on the uniformity of wind speed data distribution, wind speed data, and smoke concentration data. Similarly, This indicates the overall smoke extraction effect at the simulated measurement point determined based on the uniformity of the estimated wind speed data, the estimated wind speed data, and the estimated smoke concentration data. This indicates the overall smoke extraction efficiency in the subway space.

[0092] In this way, based on the first smoke exhaust distance, the second smoke exhaust distance, the first average wind speed, the first wind speed standard deviation, the second average wind speed, the second wind speed standard deviation, wind speed data, smoke concentration data, estimated wind speed data, and estimated smoke concentration data, the overall smoke exhaust effect coefficient at multiple moments in the i-th optimization cycle is determined. During the calculation process, the uniformity of wind speed distribution can be accurately analyzed, and the smoke exhaust effect of wind speed is analyzed separately for cases where the smoke concentration meets or does not meet the standard. Furthermore, based on the uniformity of wind speed distribution and the smoke exhaust effect of wind speed, the overall smoke exhaust effect coefficient is determined, improving the comprehensiveness and accuracy of the overall smoke exhaust effect coefficient.

[0093] According to an embodiment of the present invention, in step S6, it is determined whether the fan parameter control model for the (i-1)th optimization cycle needs to be trained based on the overall smoke exhaust effect coefficient.

[0094] Figure 4 An exemplary schematic diagram illustrating the determination of whether a wind turbine parameter control model needs training according to an embodiment of the present invention is shown.

[0095] According to an embodiment of the present invention, step S6 includes:

[0096] Step S61: Determine the first judgment condition based on the minimum value of the overall smoke extraction effect coefficient at multiple moments in the i-th optimization cycle;

[0097] Step S62: Obtain the standard deviation and average value of the overall smoke extraction effect coefficient at multiple times in the i-th optimization cycle;

[0098] Step S63: Determine the second judgment condition based on the judgment standard deviation and the judgment average.

[0099] Step S64: Based on the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, determine the smoke extraction effect coefficient of each key node and the simulated measurement point at multiple times in the i-th optimization cycle at a single location.

[0100] Step S65: Determine the third judgment condition based on the smoke extraction effect coefficient of the single location;

[0101] Step S66: If any one of the first judgment condition, the second judgment condition, and the third judgment condition is met, determine that the wind turbine parameter control model for the (i-1)th optimization cycle needs to be trained.

[0102] For example, the first judgment condition is that the minimum value of the overall smoke extraction effect coefficient at multiple moments in the i-th optimization period is less than the preset overall smoke extraction effect coefficient threshold (which can be set to 0.75), indicating that there is a situation where the overall smoke extraction effect is not up to standard in the i-th optimization period; obtain the judgment standard deviation (standard deviation of the overall smoke extraction effect coefficient at multiple moments) and judgment average (average of the overall smoke extraction effect coefficient at multiple moments) of the overall smoke extraction effect coefficient at multiple moments in the i-th optimization period; the second judgment condition is that the ratio of the judgment standard deviation to the judgment average is greater than 0.1, indicating that there is a large fluctuation in the overall smoke extraction effect in the i-th optimization period; based on wind speed data, smoke concentration data, estimated wind speed data, and estimated smoke concentration data, determine the smoke extraction effect coefficient of each key node and simulated measurement point at multiple moments in the i-th optimization period, such as, based on Determine the smoke extraction effect coefficient at a single location for each key node at multiple times during the i-th optimization cycle, based on... Determine the smoke extraction effect coefficient at a single location for each simulated measurement point at multiple times during the i-th optimization cycle, where, and All of these are part of formula (1), and the specific explanation will not be repeated here; the third judgment condition is that the minimum value of the smoke exhaust effect coefficient of each key node and the simulated measurement point at multiple times in the i-th optimization cycle is less than 0.75, indicating that there are key nodes or simulated measurement points that do not meet the standards in the i-th optimization cycle; if any one of the first judgment condition, the second judgment condition, and the third judgment condition is met, it is determined that the fan parameter control model for the i-1 optimization cycle needs to be trained.

[0103] According to an embodiment of the present invention, in step S7, when the fan parameter control model for the (i-1)th optimization cycle needs to be trained, the training loss function of the fan parameter control model for the (i-1)th optimization cycle is determined based on the subway space design information, the real-time subway space data, and the overall smoke exhaust effect coefficient.

[0104] Figure 5 An exemplary diagram is shown of the training loss function for determining the wind turbine parameter control model for the (i-1)th optimization cycle according to an embodiment of the present invention.

[0105] According to an embodiment of the present invention, step S7 includes:

[0106] Step S71: Among the first judgment condition, the second judgment condition and the third judgment condition, determine the training conditions that the smoke extraction effect coefficient of a single location and the overall smoke extraction effect coefficient satisfy.

[0107] Step S72: Based on the subway space design information, determine the importance identification results of each key node and simulated measurement point;

[0108] Step S73: Determine the real-time passenger flow of the subway based on the real-time subway space data;

[0109] Step S74: Based on the smoke exhaust effect coefficient of a single location, the real-time passenger flow of the subway, the importance identification result, the overall smoke exhaust effect coefficient, and the satisfaction of the training conditions, determine the training loss function of the fan parameter control model for the (i-1)th optimization cycle.

[0110] For example, under the first, second, and third judgment conditions, the training conditions that the smoke extraction effect coefficient of a single location and the overall smoke extraction effect coefficient must satisfy are determined; based on the subway space design information, the importance identification results of each key node and simulated measurement point are determined. For example, for key nodes and simulated measurement points of the main evacuation passage, since the evacuation passage is crucial for personnel evacuation, the importance identification result is 1.8; for key nodes and simulated measurement points of the platform area, due to the high density of people, the importance identification result is 2; for key nodes and simulated measurement points of tunnel curves and connecting passages, due to the potential for airflow four-way drive, the importance identification result is 1.2; for key nodes and simulated measurement points near equipment rooms, the importance identification result is 0.7; real-time subway passenger flow is determined through real-time subway space data; based on the real-time subway passenger flow, importance identification results, overall smoke extraction effect coefficient, and the training conditions, the training loss function of the fan parameter control model for the (i-1)th optimization cycle is determined.

[0111] According to an embodiment of the present invention, step S74 includes: determining the training loss function of the wind turbine parameter control model for the (i-1)th optimization cycle according to formula (2). , (2),

[0112] Where `min` is the function to find the minimum value, and `if` is the conditional function. This is the first condition for judgment. This is the second condition for judgment. The third condition for judgment The overall smoke extraction efficiency coefficient at time k in the i-th optimization cycle. To preset the overall smoke extraction effect coefficient threshold, For the real-time passenger flow of the subway in the i-th optimization period, Preset passenger flow threshold Let be the standard deviation of the overall smoke extraction efficiency coefficient at multiple moments in the i-th optimization cycle. Let be the average value of the overall smoke extraction efficiency coefficient at multiple moments during the i-th optimization cycle. Let be the smoke extraction efficiency coefficient of the e-th critical node at the k-th time in the i-th optimization cycle. Let r be the smoke extraction efficiency coefficient at a single location of r simulated measurement points at time k in the i-th optimization cycle. To preset the threshold value of the smoke extraction effect coefficient for a single location, The importance identification result corresponding to the smallest single location smoke extraction effect coefficient is given by R, where R is the number of simulated measurement points, r≤R, E is the number of key nodes, e≤E, and K is the number of optimization cycle moments, k≤K. e, E, r, R, k, and K are all positive integers.

[0113] According to an embodiment of the present invention, the condition function The value includes the following two cases: when the first judgment condition is met, it means that the minimum value of the overall smoke extraction effect coefficient at multiple moments in the i-th optimization cycle is less than the preset overall smoke extraction effect coefficient threshold (0.75), and the value of the condition function is... That is, the difference between the minimum value of the overall smoke extraction effect coefficient at multiple times in the i-th optimization period and the preset overall smoke extraction effect coefficient threshold is used as the training loss function. During the training process, according to the preset weights... Training is performed to narrow this gap, so that the minimum value of the overall smoke extraction efficiency coefficient at multiple moments in the i-th optimization cycle approaches the preset overall smoke extraction efficiency coefficient threshold. This is the ratio of the real-time subway passenger flow to the preset passenger flow threshold during the i-th optimization period. The larger this ratio, the larger the real-time subway passenger flow, the more polluted the air in the subway space, the higher the requirement for wind speed, and the larger the loss function value. This can guide the model to pay more attention to the situation of large real-time subway passenger flow during training, and improve the adaptability of the fan parameter control model to peak passenger flow. When the first judgment condition is not met, the above condition function is set to 0, and the parameters related to wind speed in the loss function do not need to be adjusted.

[0114] According to an embodiment of the present invention, the condition function The value includes the following two cases: when the second judgment condition is met, it means that the ratio of the judgment standard deviation to the judgment mean is greater than 0.1, and the overall smoke extraction effect coefficient fluctuates significantly at multiple moments in the i-th optimization cycle. The value of the condition function is... That is, the difference between the ratio of the judgment standard deviation to the judgment mean and 0.1 is used as the training loss function. During training, the loss is calculated based on preset weights. Training is performed to reduce the difference, making the overall smoke extraction efficiency coefficient more stable at multiple moments in the i-th optimization cycle, thus improving the stability of the fan parameter control model. When the second judgment condition is not met, the above condition function is set to 0, and the parameters related to wind speed in the loss function do not need to be adjusted.

[0115] According to an embodiment of the present invention, the condition function The value includes the following two cases. When the third judgment condition is met, it means that the minimum value of the smoke extraction effect coefficient of each key node and simulated measurement point at multiple times in the i-th optimization cycle is less than the preset threshold for the smoke extraction effect coefficient of a single location (which can be set to 0.75). There is a single key node or simulated measurement point that does not meet the requirements, and the value of the condition function is... That is, the difference between the minimum value of the smoke extraction effect coefficient of a single location at multiple times during the i-th optimization cycle of each key node and simulated measurement point and the preset threshold value of the smoke extraction effect coefficient of a single location is used as the training loss function. During the training process, according to the preset weights... Training is performed to narrow this gap, ensuring that the minimum value of the smoke extraction effect coefficient at a single location at multiple times during the i-th optimization cycle for each key node and simulated measurement point approaches the preset threshold for the smoke extraction effect coefficient at a single location. The location with the smallest smoke extraction efficiency coefficient is the location whose importance is identified. The larger the value, the more important the location corresponding to the smallest single-location smoke extraction effect coefficient. The larger the training loss function, the more the model can focus on more important locations during training, improving the accuracy of the fan parameter control model for controlling fan parameters at more important locations. When the third judgment condition is not met, the above condition function is set to 0, and the parameters related to wind speed in the loss function do not need to be adjusted.

[0116] According to an embodiment of the present invention, in formula (2), the above three terms can be summed to obtain the training loss function of the wind turbine parameter control model for the (i-1)th optimization cycle. .

[0117] In this way, the training loss function of the fan parameter control model for the (i-1)th optimization cycle can be determined based on the smoke exhaust effect coefficient of a single location, real-time subway passenger flow, importance identification results, overall smoke exhaust effect coefficient, and the satisfaction of training conditions. Training the fan parameter control model based on the training loss function can make the fan parameter control model more stable, more adaptable to peak passenger flow conditions, and improve the accuracy of the fan parameter control model for fan parameter control at more important locations.

[0118] According to an embodiment of the present invention, in step S8, the wind turbine parameter control model for the (i-1)th optimization period is trained according to the training loss function of the wind turbine parameter control model for the (i-1)th optimization period to obtain the wind turbine parameter control model for the ith optimization period, and in the (i+1)th optimization period, each wind turbine is set according to the wind turbine parameters of each wind turbine determined by the wind turbine parameter control model for the ith optimization period.

[0119] For example, the wind turbine parameter control model for the (i-1)th optimization period is trained based on the training loss function of the wind turbine parameter control model for the (i-1)th optimization period, thereby improving the performance of the wind turbine parameter control model in a targeted manner, and obtaining the wind turbine parameter control model for the ith optimization period. In the (i+1)th optimization period, each wind turbine is set according to the wind turbine parameters determined by the wind turbine parameter control model for the ith optimization period.

[0120] According to the multi-sensor-based subway space smoke exhaust wind speed detection optimization method of the present invention, the wind speed data and smoke concentration data of key nodes in the subway space, as well as the estimated wind speed data and estimated smoke concentration data of simulated measurement points, can be accurately acquired. Based on the wind speed data, smoke concentration data, estimated wind speed data, and estimated smoke concentration data, the overall wind speed smoke exhaust effect in the subway space is evaluated, and the overall smoke exhaust effect coefficient is determined. Furthermore, based on the overall smoke exhaust effect coefficient, the training loss function of the fan parameter control model is determined, and the fan parameter control model is trained according to the training loss function, so that when the fan parameters are controlled by the fan parameter control model, the wind speed smoke exhaust effect in the subway space reaches the optimal state. When determining the overall smoke extraction effect coefficient, the overall smoke extraction effect coefficient for multiple moments in the i-th optimization cycle is determined based on the first smoke extraction distance, the second smoke extraction distance, the first average wind speed, the first wind speed standard deviation, the second average wind speed, the second wind speed standard deviation, wind speed data, smoke concentration data, estimated wind speed data, and estimated smoke concentration data. During the calculation process, the uniformity of wind speed distribution can be accurately analyzed, and the smoke extraction effect of wind speed is analyzed separately for cases where the smoke concentration meets or does not meet the standard. Furthermore, based on the uniformity of wind speed distribution and the smoke extraction effect of wind speed, the overall smoke extraction effect coefficient is determined, improving the comprehensiveness and accuracy of the overall smoke extraction effect coefficient. When determining the training loss function of the fan parameter control model, the training loss function of the fan parameter control model for the (i-1)th optimization cycle can be determined based on the smoke exhaust effect coefficient of a single location, the real-time passenger flow of the subway, the importance identification results, the overall smoke exhaust effect coefficient, and the satisfaction of the training conditions. Training the fan parameter control model according to the training loss function can make the fan parameter control model more stable, make the fan parameter control model more adaptable to peak passenger flow conditions, and improve the accuracy of the fan parameter control model for fan parameter control at more important locations.

[0121] Figure 6 An exemplary block diagram of a multi-sensor-based subway space smoke exhaust wind speed detection and optimization system according to an embodiment of the present invention is shown, the system comprising:

[0122] The subway data module acquires subway space design information and real-time subway space data.

[0123] In the initial setup module, during the i-th optimization cycle, each fan is set according to the fan parameters determined by the fan parameter control model of the (i-1)-th optimization cycle. The fan parameters include: fan speed and damper opening.

[0124] The detection data module is used to acquire wind speed data and smoke concentration data through a distributed multi-sensor array set on key nodes during the i-th optimization cycle.

[0125] The estimation data module is used to determine the estimated wind speed data and estimated smoke concentration data of the simulated measurement point based on the wind speed data and the smoke concentration data.

[0126] The smoke extraction coefficient module is used to determine the overall smoke extraction effect coefficient at multiple moments in the i-th optimization cycle based on the subway space design information, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data.

[0127] The training module is used to determine whether the fan parameter control model for the (i-1)th optimization cycle needs to be trained based on the overall smoke exhaust effect coefficient.

[0128] The loss function module is used to determine the training loss function of the fan parameter control model for the (i-1)th optimization period based on the subway space design information, the real-time subway space data, and the overall smoke exhaust effect coefficient, when the fan parameter control model needs to be trained in the (i-1)th optimization period.

[0129] The optimization setting module is used to train the wind turbine parameter control model for the (i-1)th optimization period according to the training loss function of the wind turbine parameter control model for the (i-1)th optimization period, to obtain the wind turbine parameter control model for the ith optimization period, and in the (i+1)th optimization period, to set the wind turbine parameters of each wind turbine according to the wind turbine parameter control model for the ith optimization period.

[0130] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0131] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and any variations or modifications may be made to the implementation of the present invention without departing from the stated principles.

Claims

1. A multi-sensor-based optimization method for detecting and optimizing smoke exhaust wind speed in subway spaces, characterized in that, include: Obtain subway space design information and real-time subway space data; In the i-th optimization cycle, each fan is set according to the fan parameters determined by the fan parameter control model of the (i-1)-th optimization cycle, wherein the fan parameters include: fan speed and damper opening. In the i-th optimization cycle, wind speed data and smoke concentration data are acquired by a distributed multi-sensor array set on key nodes; Based on the wind speed data and the smoke concentration data, the estimated wind speed data and estimated smoke concentration data for the simulated measurement points are determined; Based on the subway space design information, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, the overall smoke extraction effect coefficients at multiple moments in the i-th optimization cycle are determined. Based on the overall smoke extraction effect coefficient, determine whether the fan parameter control model for the (i-1)th optimization cycle needs to be trained. If the fan parameter control model for the (i-1)th optimization cycle needs to be trained, the training loss function of the fan parameter control model for the (i-1)th optimization cycle is determined based on the subway space design information, the real-time subway space data, and the overall smoke exhaust effect coefficient. The wind turbine parameter control model for the (i-1)th optimization period is trained according to the training loss function of the wind turbine parameter control model for the (i-1)th optimization period to obtain the wind turbine parameter control model for the ith optimization period. In the (i+1)th optimization period, each wind turbine is set according to the wind turbine parameters of each wind turbine determined by the wind turbine parameter control model for the ith optimization period. Based on the subway space design information, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, the overall smoke extraction effect coefficients for multiple moments in the i-th optimization cycle are determined, including: Based on the subway space design information, determine the first smoke exhaust distance from each key node to the nearest smoke exhaust outlet and the second smoke exhaust distance from each simulated measurement point to the nearest smoke exhaust outlet. Based on the wind speed data and the estimated wind speed data, a first average wind speed, a first wind speed standard deviation, a second average wind speed, and a second wind speed standard deviation are determined. Based on the first smoke exhaust distance, the second smoke exhaust distance, the first average wind speed, the first wind speed standard deviation, the second average wind speed, the second wind speed standard deviation, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, the overall smoke exhaust effect coefficient at multiple moments in the i-th optimization cycle is determined. Based on the first smoke exhaust distance, the second smoke exhaust distance, the first average wind speed, the first wind speed standard deviation, the second average wind speed, the second wind speed standard deviation, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, the overall smoke exhaust effect coefficient at multiple moments in the i-th optimization period is determined, including: according to the formula: ， Determine the overall smoke extraction efficiency coefficient at time k in the i-th optimization cycle. ,in, Let the standard deviation of the first wind speed be the value at the k-th moment of the i-th optimization period. Let be the first average wind speed at the k-th moment of the i-th optimization cycle. Let the standard deviation of the second wind speed be the value at the k-th moment of the i-th optimization period. Let be the second average wind speed at time k in the i-th optimization cycle, and let 'if' be a conditional function. For the smoke concentration data of the e-th critical node at time k in the i-th optimization cycle, To preset the smoke concentration threshold, Let the wind speed data for the e-th critical node at the k-th time in the i-th optimization cycle be given. To preset the wind speed data threshold, Let the distance be the first smoke exhaust distance of the e-th critical node. To preset the smoke exhaust distance threshold, For the estimated smoke concentration data of the r-th simulated measurement point at the k-th time in the i-th optimization cycle, For the estimated wind speed data of the r-th simulated measurement point at the k-th time in the i-th optimization cycle, Let R be the second smoke exhaust distance of the r-th simulated measurement point, R be the number of simulated measurement points, r ≤ R, E be the number of key nodes, e ≤ E, and e, E, r and R are all positive integers.

2. The method for optimizing smoke exhaust velocity detection in subway spaces based on multiple sensors according to claim 1, characterized in that, Based on the wind speed data and the smoke concentration data, the estimated wind speed data and estimated smoke concentration data for the simulated measurement points are determined, including: Establish a BIM model of the subway space; By using BIM models and spatial interpolation algorithms, estimated wind speed and smoke concentration data were determined for each simulated measurement point in the subway space.

3. The method for optimizing smoke exhaust velocity detection in subway spaces based on multiple sensors according to claim 1, characterized in that, Based on the overall smoke extraction efficiency coefficient, determine whether the fan parameter control model for the (i-1)th optimization cycle needs to be trained, including: The first judgment condition is determined based on the minimum value of the overall smoke extraction effect coefficient at multiple moments in the i-th optimization cycle. Obtain the standard deviation and average value of the overall smoke extraction effect coefficient at multiple times during the i-th optimization cycle; The second judgment condition is determined based on the judgment standard deviation and the judgment average. Based on the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data, determine the smoke extraction effect coefficient at a single location for each key node and simulated measurement point at multiple times in the i-th optimization cycle; The third judgment condition is determined based on the smoke extraction effect coefficient of the single location; If any one of the first judgment condition, the second judgment condition, and the third judgment condition is met, it is determined that the wind turbine parameter control model for the (i-1)th optimization cycle needs to be trained.

4. The method for optimizing smoke exhaust velocity detection in subway spaces based on multiple sensors according to claim 3, characterized in that, When the fan parameter control model for the (i-1)th optimization cycle needs to be trained, the training loss function for the fan parameter control model for the (i-1)th optimization cycle is determined based on the subway space design information, the real-time subway space data, and the overall smoke extraction effect coefficient, including: Among the first judgment condition, the second judgment condition and the third judgment condition, determine the training conditions that the smoke exhaust effect coefficient of a single location and the overall smoke exhaust effect coefficient must satisfy. Based on the subway space design information, the importance identification results of each key node and simulated measurement point are determined; Based on the real-time subway space data, the real-time passenger flow of the subway is determined; Based on the smoke extraction effect coefficient of a single location, the real-time passenger flow of the subway, the importance identification result, the overall smoke extraction effect coefficient, and the satisfaction of the training conditions, the training loss function of the fan parameter control model for the (i-1)th optimization cycle is determined.

5. The method for optimizing smoke exhaust velocity detection in subway spaces based on multiple sensors according to claim 4, characterized in that, Based on the smoke extraction effect coefficient at a single location, the real-time passenger flow of the subway, the importance identification result, the overall smoke extraction effect coefficient, and the satisfaction of training conditions, the training loss function of the fan parameter control model for the (i-1)th optimization cycle is determined, including: according to the formula: ， Determine the training loss function of the wind turbine parameter control model for the (i-1)th optimization cycle. Where, min is the minimum value function, and if is the conditional function, where, This is the first condition for judgment. This is the second condition for judgment. This is the third condition for judgment. Let be the overall smoke extraction efficiency coefficient at time k in the i-th optimization cycle. To preset the overall smoke extraction effect coefficient threshold, For the real-time passenger flow of the subway in the i-th optimization period, To preset the passenger flow threshold, Let be the standard deviation of the overall smoke extraction efficiency coefficient at multiple moments in the i-th optimization cycle. Let be the average value of the overall smoke extraction efficiency coefficient at multiple moments in the i-th optimization cycle. Let be the smoke extraction efficiency coefficient of the e-th critical node at the k-th time in the i-th optimization cycle. Let r be the smoke extraction efficiency coefficient at a single location of r simulated measurement points at time k in the i-th optimization cycle. To preset the threshold value of the smoke extraction effect coefficient for a single location, The importance identification result corresponding to the smallest single location smoke extraction effect coefficient is given by R, where R is the number of simulated measurement points, r≤R, E is the number of key nodes, e≤E, and K is the number of optimization cycle moments, k≤K. e, E, r, R, k, and K are all positive integers.

6. A multi-sensor-based subway space smoke exhaust wind speed detection and optimization system, characterized in that, For performing the method according to any one of claims 1-5, comprising: The subway data module acquires subway space design information and real-time subway space data. In the initial setup module, during the i-th optimization cycle, each fan is set according to the fan parameters determined by the fan parameter control model of the (i-1)-th optimization cycle. The fan parameters include: fan speed and damper opening. The detection data module is used to acquire wind speed data and smoke concentration data through a distributed multi-sensor array set on key nodes during the i-th optimization cycle. The estimation data module is used to determine the estimated wind speed data and estimated smoke concentration data of the simulated measurement point based on the wind speed data and the smoke concentration data. The smoke extraction coefficient module is used to determine the overall smoke extraction effect coefficient at multiple moments in the i-th optimization cycle based on the subway space design information, the wind speed data, the smoke concentration data, the estimated wind speed data, and the estimated smoke concentration data. The training module is used to determine whether the fan parameter control model for the (i-1)th optimization cycle needs to be trained based on the overall smoke exhaust effect coefficient. The loss function module is used to determine the training loss function of the fan parameter control model for the (i-1)th optimization period based on the subway space design information, the real-time subway space data, and the overall smoke exhaust effect coefficient, when the fan parameter control model needs to be trained in the (i-1)th optimization period. The optimization setting module is used to train the wind turbine parameter control model for the (i-1)th optimization period according to the training loss function of the wind turbine parameter control model for the (i-1)th optimization period, to obtain the wind turbine parameter control model for the ith optimization period, and in the (i+1)th optimization period, to set the wind turbine parameters of each wind turbine according to the wind turbine parameter control model for the ith optimization period.

Images