Intelligent underground space smoke exhaust system

Abstract

This invention discloses an intelligent underground space smoke extraction system, comprising: a mechanical smoke extraction device including a smoke extraction window with an adjustable opening; a natural smoke extraction device including a smoke extraction fan with an adjustable fan speed; a sensing module including a sensor device for real-time monitoring of environmental parameters within the underground space; an identification module for determining a smoke distribution matrix based on the coordinate parameters of the sensor device and the environmental parameters; and a control module for dynamically adjusting the fan speed of the smoke extraction fan and the opening of the smoke extraction window based on the smoke distribution matrix, data from the mechanical smoke extraction device, and data from the natural smoke extraction device, thereby reducing mutual interference between the natural and mechanical smoke extraction devices.

Description

Technical Field

[0001] This invention relates to the field of fire smoke extraction, and more particularly to an intelligent underground space smoke extraction system. Background Technology

[0002] Smoke extraction systems are an important component of building fire protection facilities. They are mainly used to quickly remove smoke, high-temperature gases, and toxic gases during a fire, ensuring the safe evacuation of personnel and creating favorable conditions for firefighters to fight fires and conduct rescue operations. Underground spaces, such as underground parking garages, subway stations, civil defense projects, and tunnels, face many unique challenges in the design and operation of their smoke extraction systems due to their enclosed structures and poor ventilation. Underground spaces lack direct external openings, have a limited number of natural smoke extraction vents, and experience slow smoke extraction speeds. Natural smoke extraction relies on the natural ascent of smoke, but the low ceiling heights and winding passageways in underground spaces make it easy for smoke to accumulate in dead corners. Therefore, mechanical smoke extraction equipment plays a crucial role in underground spaces.

[0003] However, due to the relatively enclosed environment of underground spaces, the simultaneous operation of mechanical and natural smoke extraction equipment can lead to several problems: the high-speed suction of mechanical smoke extraction fans may disrupt the thermal pressure difference upon which natural smoke extraction relies, causing some of the exhausted smoke to be re-drawn into the space; the forced airflow from the mechanical system and natural convection can create unpredictable vortices in the complex space, delaying smoke extraction; and when natural and mechanical smoke extraction equipment are both in an enclosed environment, the natural smoke extraction outlet may become an air inlet. In short, the coordinated operation of natural and mechanical smoke extraction requires overcoming challenges such as airflow coupling instability and control timing conflicts.

[0004] Therefore, designing an intelligent underground space smoke exhaust system is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide an intelligent underground space smoke exhaust system that dynamically controls the operation of natural smoke exhaust equipment and mechanical smoke exhaust equipment, reduces mutual interference between the natural smoke exhaust equipment and mechanical smoke exhaust equipment, and improves smoke exhaust efficiency.

[0006] To achieve this objective, the present invention adopts the following technical solution:

[0007] An intelligent underground space smoke exhaust system, characterized in that it includes:

[0008] Mechanical smoke extraction equipment, including a smoke extraction window, said smoke extraction window having an adjustable opening degree;

[0009] Natural smoke extraction equipment, including a smoke extraction fan, said smoke extraction fan having an adjustable fan speed;

[0010] The sensing module includes sensor devices for real-time monitoring of environmental parameters within the underground space.

[0011] The identification module is used to determine the flue gas distribution matrix based on the coordinate parameters of the sensor device and the environmental parameters;

[0012] The control module is used to dynamically adjust the fan speed of the smoke exhaust fan and the opening of the smoke exhaust window based on the data of the smoke distribution matrix, the mechanical smoke exhaust equipment and the natural smoke exhaust equipment, so as to reduce the mutual interference between the natural smoke exhaust equipment and the mechanical smoke exhaust equipment.

[0013] Optionally, the sensor device includes a temperature sensor and a smoke sensor, which are arranged on the ceiling of the underground space to cover the entire underground space. The sensing module acquires real-time environmental parameters at various locations in the underground space.

[0014] Optionally, the identification module acquires the spatial information of the underground space, determines the shortest path distance between any two sensor devices based on the spatial information of the underground space and the coordinate parameters of any two sensor devices, and determines the shortest path distance between each prediction point and all sensor devices based on the spatial information of the underground space, the coordinate parameters of the sensor devices, and the coordinate parameters of the prediction point.

[0015] Optionally, the identification module determines a weight vector based on the shortest path distance between any two sensor devices and the environmental parameters, obtains the predicted environmental parameters of the predicted point based on the shortest path distance between each predicted point and all sensor devices and the weight vector, and obtains the flue gas distribution matrix based on the predicted environmental parameters and coordinate parameters of the predicted point, as well as the environmental parameters and coordinate parameters of the sensors.

[0016] Optionally, the control module includes a spatial coding module, a temporal feature extraction module, a cross-attention module, an interference suppression module, and a collaborative decision-making module. The control module inputs multiple flue gas distribution matrices at consecutive different time points into the spatial coding module to obtain a spatial attention feature map.

[0017] Optionally, the time-series feature extraction module is used to obtain the time-series features of the mechanical smoke exhaust device and the natural smoke exhaust device based on the data of the mechanical smoke exhaust device and the natural smoke exhaust device, respectively. The data of the mechanical smoke exhaust device includes the opening degree of the smoke exhaust window, the wind speed outside the window, the indoor and outdoor temperature difference, and the historical smoke exhaust efficiency at different time points. The data of the natural smoke exhaust device includes the fan speed, current, static pressure, and filter pressure difference at different time points. The time-series feature extraction module includes a parallel GRU module and a TCN module. The data of the mechanical smoke exhaust device and the natural smoke exhaust device are respectively processed by the GRU module to output the long-term time features of the mechanical smoke exhaust device and the natural smoke exhaust device. The data of the mechanical smoke exhaust device and the natural smoke exhaust device are respectively processed by the TCN module to output the local time features of the mechanical smoke exhaust device and the natural smoke exhaust device. The long-term time features and local time features of the mechanical smoke exhaust device are fused together, and the long-term time features and local time features of the natural smoke exhaust device are fused together to output the time-series features of the mechanical smoke exhaust device and the natural smoke exhaust device.

[0018] Optionally, the temporal characteristics and position parameters of the mechanical smoke exhaust device and the temporal characteristics and position parameters of the natural smoke exhaust device are input to the cross-attention module to obtain a spatiotemporal coupling weight matrix representing the interference region.

[0019] Optionally, the interference suppression module is used to obtain a suppression heatmap based on the spatiotemporal coupling weight matrix, the location parameters of the natural smoke exhaust device and the mechanical smoke exhaust device. The interference suppression module includes an interference intensity analysis unit and a suppression coefficient generation unit. The interference intensity analysis unit extracts the spatiotemporal features of the spatiotemporal coupling weight matrix. The spatiotemporal features are sequentially processed by multi-head attention and max pooling operations to obtain intermediate features. A fully connected layer is used to map the intermediate features to interference intensity scores. The suppression coefficient generation unit adjusts the steepness parameter of the sigmoid function based on the average value of the interference intensity scores. The suppression coefficient generation unit uses the sigmoid function to obtain the suppression coefficient based on the steepness parameter and the interference intensity scores. The interference suppression module forms a suppression heatmap based on the suppression coefficient and the location parameters of the smoke exhaust device.

[0020] Optionally, the collaborative decision-making module determines the fan speed of each smoke exhaust fan and the opening of the smoke exhaust window based on the spatial attention feature map and the suppression heat map.

[0021] Optionally, the collaborative decision-making module includes a strategy module and a value evaluation module. The strategy module equips each natural smoke exhaust device and mechanical smoke exhaust device with an independent strategy sub-network. The spatial attention feature map, suppression heat map, and position parameters of the smoke exhaust device are input into the strategy sub-network to obtain the action probability distribution of the smoke exhaust device. The value evaluation module establishes an independent value assessment model for each smoke exhaust device, calculates the advantage function value and objective function of each state-action pair, updates the parameters of the strategy sub-network according to the advantage function and objective function, and after multiple iterations, outputs the fan speed of each smoke exhaust fan and the opening degree of the smoke exhaust window that maximizes the objective function.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] The intelligent underground space smoke extraction system provided by this invention predicts the smoke distribution of the entire underground space more accurately based on the shortest path distance between sensors in the underground space, the shortest path distance between the prediction point and the sensor, and the data obtained from the sensor data. This provides a solid data foundation for subsequent smoke extraction equipment control. Furthermore, it determines the interference between smoke extraction equipment based on historical data and location distribution of natural and mechanical smoke extraction equipment, and generates a suppression heat map for equipment control. This provides data suggestions for the final equipment control decision, thereby reducing the mutual influence between smoke extraction equipment and improving the smoke extraction efficiency in enclosed underground spaces.

[0024] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0025] The above and other objects, features, and advantages of the present invention will become more apparent from the more detailed description of the embodiments of the invention in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same parts or steps.

[0026] Figure 1 This is a schematic diagram of the structure of an intelligent underground space smoke exhaust system provided in an exemplary embodiment of the present invention.

[0027] Figure 2 This is a schematic diagram of the structure of a control module provided in an exemplary embodiment of the present invention. Detailed Implementation

[0028] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0029] It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention.

[0030] Those skilled in the art will understand that the terms "first," "second," etc., in the embodiments of the present invention are only used to distinguish different steps, devices, or modules, and do not represent any specific technical meaning, nor do they indicate a necessary logical order between them.

[0031] It should also be understood that in the embodiments of the present invention, "multiple" can refer to two or more, and "at least one" can refer to one, two or more.

[0032] It should also be understood that any component, data or structure mentioned in the embodiments of the present invention can generally be understood as one or more unless explicitly defined or given contrary instructions in the context.

[0033] Example 1

[0034] like Figure 1 As shown, the intelligent underground space smoke exhaust system 100 provided in this embodiment includes:

[0035] Mechanical smoke extraction equipment 110 includes a smoke extraction window, the smoke extraction window having an adjustable opening degree;

[0036] Natural smoke exhaust device 120 includes a smoke exhaust fan, said smoke exhaust fan having an adjustable fan speed;

[0037] The sensing module 130 includes a sensor device, which is used to monitor environmental parameters in the underground space in real time.

[0038] The identification module 140 is used to determine the flue gas distribution matrix based on the coordinate parameters of the sensor device and the environmental parameters;

[0039] The control module 150 is used to dynamically adjust the fan speed of the smoke exhaust fan and the opening of the smoke exhaust window based on the data of the smoke distribution matrix, the mechanical smoke exhaust equipment and the natural smoke exhaust equipment, so as to reduce the mutual interference between the natural smoke exhaust equipment and the mechanical smoke exhaust equipment.

[0040] Among them, the smoke exhaust fan of the mechanical smoke exhaust equipment 110 can quickly expel fire smoke from the building by forced extraction, ensuring the safe evacuation of personnel and the smooth progress of fire rescue. It can quickly extract high-temperature toxic smoke during a fire, prevent smoke accumulation, reduce the temperature of the fire scene, and suppress the spread of fire. It can form an effective airflow organization through negative pressure smoke exhaust and positive pressure air supply.

[0041] The 120-type natural smoke exhaust system features operable exterior windows, skylights, and louvered windows specifically designed for fire smoke extraction. Relying on thermal pressure (chimney effect) and wind pressure, it utilizes the principle of rising hot air currents generated by a fire to expel high-temperature smoke, reducing smoke accumulation within the building and improving visibility. The operable exterior windows, skylights, and louvered windows can all be electrically controlled to ensure effective ventilation area.

[0042] The sensor equipment includes a temperature sensor and a smoke sensor. The sensor equipment is installed on the ceiling of the underground space, covering the entire underground space. The sensing module acquires real-time environmental parameters at various locations in the underground space.

[0043] In this invention, due to cost considerations and placement requirements, sensor devices cannot cover all underground spaces with high density. Data from these sensors alone cannot directly reflect the distribution of smoke throughout the underground space, thus failing to provide a solid data foundation for subsequent smoke extraction system control. Therefore, to address this issue, the identification module 140 uses an algorithm to predict environmental parameters at other locations based on the spatial information of the underground space, the coordinate parameters of each sensor device, and environmental parameters, thereby generating a smoke distribution matrix for the entire underground space.

[0044] Specifically, the identification module 140 acquires spatial information of the underground space and determines the shortest path distance between any two sensor devices based on the spatial information and the coordinate parameters of any two sensor devices. The closer the two sensor devices are, the stronger their correlation; calculating the distance between the two sensor devices can reflect the spatial correlation between data points. However, underground spaces often have walls as obstructions. The straight-line distance between sensor devices located on either side of a wall is very small, while their actual physical paths are relatively long. Therefore, the shortest path distance between sensor devices that bypasses obstacles can better reflect their correlation.

[0045] The shortest path can be selected using algorithms such as Dijkstra's algorithm, Bellman-Ford algorithm, and TEB algorithm. Given n sensor devices in an underground space, any algorithm can be used to calculate the shortest path distance between all known sensor devices, generating an n×n distance matrix.

[0046] The identification module 140 determines the shortest path distance between each prediction point and all sensor devices based on the spatial information of the underground space, the coordinate parameters of the sensor devices, and the coordinate parameters of the prediction points. The prediction points are unknown locations in the underground space where environmental parameters need to be predicted. Predicting the environmental parameters of these prediction points will achieve full coverage of the underground space, reflecting the threat field of flue gas. Calculating the shortest path distance between each prediction point and all known sensor devices reveals the spatial correlation between data points. If there are n sensor devices and m prediction points in the underground space, an n×m distance matrix is ​​generated.

[0047] The identification module 140 determines a weight vector based on the shortest path distance between any two of the sensor devices and the environmental parameters. The weight vector represents the optimal weight of a known data point on a predicted point. The dot product of the weight vector and the matrix formed by the shortest path distance between any two of the sensor devices is equal to the matrix of the environmental parameters, as shown in the following equation:

[0048]

[0049] in, is the matrix of shortest path distances between sensor devices, F is the environmental parameter matrix of the sensor devices, and β is the weight vector.

[0050] The identification module 140 obtains the predicted environment parameters of each predicted point based on the shortest path distance between each predicted point and all sensor devices and the weight vector. The dot product of the matrix of the shortest path distance between each predicted point and all sensor devices and the weight vector is equal to the predicted environment parameter matrix of the predicted point, as shown in the following formula:

[0051] γ·β=z

[0052] Where γ is the matrix of the shortest path distance between each prediction point and all sensor devices, β is the weight vector, and z is the matrix of predicted environmental parameters.

[0053] Finally, the identification module 140 obtains the smoke distribution matrix based on the predicted environmental parameters and coordinate parameters of the predicted points, as well as the environmental parameters and coordinate parameters of the sensors. In other words, it combines the known sensor data points with the predicted points to obtain the smoke distribution matrix of the entire underground space.

[0054] like Figure 2 As shown, the control module 150 includes a spatial coding module 151, a temporal feature extraction module 152, a cross-attention module 153, an interference suppression module 154, and a collaborative decision-making module 155.

[0055] The control module 150 inputs multiple flue gas distribution matrices at different consecutive time points into the spatial encoding module 151 to obtain spatial attention feature maps. Specifically, the spatial encoding module 151 uses a Convolutional Long Short-Term Memory (ConvLSTM) network model. ConvLSTM is a hybrid architecture combining convolutional neural networks and long short-term memory networks, capable of processing spatial and temporal data simultaneously. By introducing convolutional operations into the gating mechanism of LSTM, it preserves the spatial structure of the input data and improves the model's efficiency and performance by utilizing the local receptive field and parameter sharing characteristics of convolutional operations. The spatial encoding module 151 includes dilated convolutional layers, a spatial attention mechanism module, a three-level convolutional feature extraction layer, and a ConvLSTM module.

[0056] The input data for the spatial coding module 151 consists of multiple smoke distribution matrices at different consecutive time points. These matrices are fed into a dilated convolutional layer, which expands the receptive field and captures smoke correlation information over a wider area. The output feature map of the dilated convolutional layer is input into the spatial attention mechanism module. This module calculates the importance weight for each location, highlighting key areas while automatically weakening sensor anomalies, such as points with a weight below 0.2. The spatial attention mechanism module performs spatial attention calculations on the output feature map of the dilated convolution, applying the weights to obtain a weighted feature map. (The text then repeats the three-level convolutional encoding method.) The feature extraction layer extracts spatiotemporal features from the feature map output by the spatial attention mechanism module, resulting in a spatiotemporal feature map. The first-level convolutional extraction layer is used to extract local concentration information, and a 3×3 convolutional kernel can be used to extract local features. The second-level convolutional extraction layer is used to capture regional diffusion patterns, and a 5×5 convolutional kernel can be used to capture regional features over a larger area. The third-level convolutional extraction layer is used to extract global flue gas movement trends, and a 7×7 convolutional kernel can be used to extract global features. The ConLSTM module is used to process multiple consecutive spatiotemporal feature maps, capture the changing trends over time, and output a spatial attention feature map for subsequent task processing.

[0057] The temporal feature extraction module 152 is used to capture the inertial characteristics of the smoke exhaust window and the smoke exhaust fan control, identify the historical dependencies of equipment operation, and analyze the time-varying interaction between natural smoke exhaust and mechanical smoke exhaust. It obtains the temporal features of the mechanical smoke exhaust equipment and the natural smoke exhaust equipment respectively based on the data of the mechanical smoke exhaust equipment and the natural smoke exhaust equipment. The data of the mechanical smoke exhaust equipment and the natural smoke exhaust equipment are the historical dynamic behavior data of the smoke exhaust fan and the smoke exhaust window, respectively. The data of the mechanical smoke exhaust equipment includes the fan speed, current, static pressure, and filter pressure difference at different time points. The data of the natural smoke exhaust equipment includes the opening degree of the smoke exhaust window, the wind speed outside the window, the indoor and outdoor temperature difference, and the historical smoke exhaust efficiency at different time points.

[0058] The temporal feature extraction module 152 includes a parallel GRU module and a TCN module. GRU can capture long-term dependencies in time series, while TCN can efficiently extract local temporal features. By combining these two models in parallel, their advantages can be fully utilized to achieve more comprehensive feature extraction. Among them, GRU (Gated Recurrent Unit) is a recurrent neural network that can solve the gradient vanishing and gradient exploding problems of traditional RNNs when processing long sequence data. By introducing a "gating mechanism" to control the flow of information, it can effectively capture long-term dependencies in time series data. TCN (Temporal Convolutional Network) is a time series modeling architecture based on convolutional neural networks, specifically designed for processing time series data.

[0059] The GRU module consists of two layers: the first layer captures short-term fluctuations, and the second layer identifies long-term patterns. The data from mechanical smoke exhaust equipment and natural smoke exhaust equipment are processed by the GRU module to output the long-term time characteristics of the mechanical smoke exhaust equipment and natural smoke exhaust equipment.

[0060] The TCN module is used to extract local time features of time series. The data from mechanical smoke exhaust equipment and natural smoke exhaust equipment are processed by the TCN module and output as local time features of mechanical smoke exhaust equipment and natural smoke exhaust equipment, respectively.

[0061] The long-term and local time features of the mechanical smoke exhaust device are fused together, and the long-term and local time features of the natural smoke exhaust device are fused together to obtain a comprehensive feature representation. The time-series features of the mechanical smoke exhaust device and the natural smoke exhaust device are then output.

[0062] The cross-attention module 153 is used to obtain a spatiotemporal coupling weight matrix representing the interference region based on the temporal and spatial characteristics of the mechanical smoke exhaust equipment and the natural smoke exhaust equipment, namely the temporal characteristics and position parameters of the mechanical smoke exhaust equipment and the temporal characteristics and position parameters of the natural smoke exhaust equipment. The cross-attention module 153 allows the feature flow of the natural smoke exhaust equipment and the mechanical smoke exhaust equipment to interact with each other through multi-head attention. The output spatiotemporal coupling weight matrix displays the flow cancellation area of ​​the two systems, i.e., the interference region. Among them, the position parameters of the mechanical smoke exhaust equipment include x-coordinate, y-coordinate, distance from the nearest smoke exhaust window, and installation height, and the position parameters of the natural smoke exhaust equipment include x-coordinate, y-coordinate, and orientation angle.

[0063] The cross-attention module 153 maps the input sequence to the query, key, and value spaces respectively. It calculates the attention weights by the dot product of the query and key, and uses the attention weights to perform a weighted summation of the values ​​to obtain the spatiotemporal coupling weight matrix.

[0064] The interference suppression module 154 is used to obtain a suppression heat map based on the spatiotemporal coupling weight matrix, the position parameters of the natural smoke exhaust device and the mechanical smoke exhaust device. The suppression heat map provides a physically reasonable feature representation for the collaborative decision-making module 155, showing the degree of suppression at each location in the space.

[0065] The interference suppression module 154 includes an interference intensity analysis unit and a suppression coefficient generation unit. The spatiotemporal coupling weight matrix of the cross-attention module 154 is input to the interference intensity analysis unit. The interference intensity analysis unit uses 3D convolution to extract the spatiotemporal features of the spatiotemporal coupling weight matrix. The spatiotemporal features are sequentially processed by multi-head attention and max pooling operations to obtain intermediate features. The multi-head attention mechanism is used to analyze the relationship between devices, while the max pooling operation is used to compress the output of multi-head attention in the time and window dimensions. Finally, the interference intensity analysis unit uses a fully connected layer to map the intermediate features into an interference intensity score, with the score range being [0,1].

[0066] The suppression coefficient generation unit generates suppression coefficients based on the interference intensity score. The suppression coefficient represents the degree of interference for each smoke extraction device; a higher suppression coefficient indicates greater interference and requires more suppression, and vice versa. While the interference intensity score directly reflects the degree of interference between smoke extraction devices, these scores may be too sensitive or not sensitive enough in certain ranges. For example, the score may change too quickly or too slowly in certain ranges, making it difficult for the collaborative decision-making module to effectively utilize this information. The suppression coefficient unit maps the score to a more suitable range using a non-linear function (sigmoid function), thus better reflecting the actual interference impact. Furthermore, by setting a lower limit for the suppression coefficient, it ensures that certain feature information is retained even in high-interference regions. This is crucial in practical applications because completely suppressing certain regions may lead to information loss, affecting the overall system performance.

[0067] The suppression coefficient generation unit first adjusts the steepness of the sigmoid function based on the average value of the interference intensity scores. Specifically, it calculates the mean of the interference intensity scores to obtain a scalar, multiplies the mean by 10 to obtain an initial steepness parameter, and restricts the initial steepness parameter to a preset range, such as between 5.0 and 20.0. If the initial steepness parameter is less than the minimum value of the preset range, the minimum value is set as the steepness parameter; if the initial steepness parameter is greater than the maximum value of the preset range, the maximum value is set as the steepness parameter; if the initial steepness parameter is within the preset range, the initial steepness parameter is set as the steepness parameter. This is to ensure that the steepness parameter is not too large or too small, thereby guaranteeing the stability and effectiveness of the sigmoid function.

[0068] The suppression coefficient generation unit obtains the suppression coefficient based on the steepness parameter and the interference intensity score, as shown in the following formula:

[0069] s = sigmoid(k * (a - 0.5))

[0070] Where s is the suppression coefficient, sigmoid() is the sigmoid function, k is the steepness parameter, and a is the interference intensity score.

[0071] The interference suppression module 154 maps the coordinates of the smoke exhaust equipment and the suppression coefficient into a two-dimensional matrix based on the suppression coefficient and the position parameters of the smoke exhaust equipment, thereby forming a suppression heat map.

[0072] The collaborative decision-making module 155 determines the fan speed and smoke exhaust window opening of each smoke exhaust fan based on the spatial attention feature map and suppression heatmap. The collaborative decision-making module 155 employs a reinforcement learning network, including a policy module and a value judgment module.

[0073] The strategy module equips each natural and mechanical smoke exhaust device with an independent strategy sub-network. Spatial attention feature maps, suppression heatmaps, and the location parameters of the smoke exhaust devices are input into the strategy sub-network to obtain the action probability distribution of the smoke exhaust devices, such as changes in the opening degree of the smoke exhaust window or changes in the speed of the fan. The value assessment module establishes an independent value assessment model for each smoke exhaust device, calculating the dominance function value for each state-action pair. The dominance function is:

[0074] A(s,a)=β1Alocal(s,a)+β2Aglobal(s,a)

[0075] =β1(Qi-μ Qi ) / σ Qi +β2α·ΔCsys / Δt

[0076] Where β1, β2, and β3 are weighting coefficients, Alocal(s,a) represents the local advantage of the device, Aglobal(s,a) represents the global advantage of the system, Atime(s,a) represents the time-series advantage, Qi is the action value of smoke extraction device i, and μ Qi and σ Qi These are the mean and standard deviation of Qi, respectively; α is the weight of smoke extraction effect; ΔCsys is the rate of change of system smoke extraction efficiency; and Δt is the time interval.

[0077] Then, the calculated advantage function value is used to update the policy subnetwork, and the objective function is established:

[0078] L(θ)=E t [min(r t (θ)A(s,a),clip(r t (θ), 1-∈, 1+∈)A(s,a))]

[0079] Where θ is the updated policy network parameter, ∈ is the pruning parameter, and r t (θ) is the ratio of the new to the old strategy, clip(r) t (θ), 1-∈, 1+∈) is the clipping function, relative to the ratio r. t (θ) is pruned and restricted to the range [1-∈, 1+∈] to prevent the policy update from being too large. A(s,a) is the advantage function.

[0080] The parameters of the strategy subnetwork are updated by combining the advantage function and the objective function. After multiple iterations, the objective function is maximized. The action that maximizes the objective function is output, namely the fan speed of each smoke exhaust fan and the opening of the smoke exhaust window, which is the optimal strategy.

[0081] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0082] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For system embodiments, since they largely correspond to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0083] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0084] The methods and apparatus of this disclosure may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the methods is for illustrative purposes only, and the steps of the methods of this disclosure are not limited to the order specifically described above unless otherwise specifically stated. Furthermore, in some embodiments, this disclosure may also be implemented as a program recorded on a recording medium, the program including machine-readable instructions for implementing the methods according to this disclosure. Thus, this disclosure also covers recording media storing programs for performing the methods according to this disclosure.

[0085] It should also be noted that in the apparatus, devices, and methods of this disclosure, the components or steps are decomposable and / or recombinable. Such decomposition and / or recombination should be considered equivalent to the present disclosure. The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0086] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. An intelligent underground space smoke exhaust system, characterized in that, include: Natural smoke extraction equipment, including a smoke extraction window, said smoke extraction window having an adjustable opening degree; Mechanical smoke extraction equipment, including a smoke extraction fan, said smoke extraction fan having an adjustable fan speed; The sensing module includes sensor devices for real-time monitoring of environmental parameters within the underground space. The identification module is used to determine the flue gas distribution matrix based on the coordinate parameters of the sensor device and the environmental parameters; The control module is used to dynamically adjust the fan speed of the smoke exhaust fan and the opening of the smoke exhaust window based on the data of the smoke distribution matrix, the mechanical smoke exhaust equipment and the natural smoke exhaust equipment, so as to reduce the mutual interference between the natural smoke exhaust equipment and the mechanical smoke exhaust equipment. The control module includes a spatial coding module, a temporal feature extraction module, a cross-attention module, an interference suppression module, and a collaborative decision-making module. The control module inputs multiple flue gas distribution matrices at different consecutive time points into the spatial coding module to obtain a spatial attention feature map. The temporal feature extraction module includes a parallel GRU module and a TCN module. The data from the mechanical smoke exhaust device and the natural smoke exhaust device are respectively processed by the GRU module to output long-term temporal features of the mechanical smoke exhaust device and the natural smoke exhaust device. The data from the mechanical smoke exhaust device and the natural smoke exhaust device are respectively processed by the TCN module to output local temporal features of the mechanical smoke exhaust device and the natural smoke exhaust device. The long-term temporal features and local temporal features of the mechanical smoke exhaust device are fused together, and the long-term temporal features and local temporal features of the natural smoke exhaust device are fused together to output the temporal features of the mechanical smoke exhaust device and the temporal features of the natural smoke exhaust device. The temporal characteristics and position parameters of the mechanical smoke exhaust device and the temporal characteristics and position parameters of the natural smoke exhaust device are input into the cross-attention module to obtain a spatiotemporal coupling weight matrix representing the interference region. The interference suppression module is used to obtain a suppression heatmap based on the spatiotemporal coupling weight matrix and the position parameters of the natural smoke exhaust device and the mechanical smoke exhaust device. The collaborative decision-making module determines the fan speed of each smoke exhaust fan and the opening degree of the smoke exhaust window based on the spatial attention feature map and the suppression heat map.

2. The system according to claim 1, characterized in that, The sensor device includes a temperature sensor and a smoke sensor. The sensor device is installed on the ceiling of the underground space, covering the entire underground space. The sensing module acquires real-time environmental parameters at various locations in the underground space.

3. The system according to claim 2, characterized in that, The identification module acquires the spatial information of the underground space, determines the shortest path distance between any two sensor devices based on the spatial information of the underground space and the coordinate parameters of any two sensor devices, and determines the shortest path distance between each prediction point and all sensor devices based on the spatial information of the underground space, the coordinate parameters of the sensor devices, and the coordinate parameters of the prediction point.

4. The system according to claim 3, characterized in that, The identification module determines a weight vector based on the shortest path distance between any two sensor devices and the environmental parameters, obtains the predicted environmental parameters of the predicted point based on the shortest path distance between each predicted point and all sensor devices and the weight vector, and obtains the flue gas distribution matrix based on the predicted environmental parameters and coordinate parameters of the predicted point and the environmental parameters and coordinate parameters of the sensor devices.

5. The system according to claim 1, characterized in that, The interference suppression module includes an interference intensity analysis unit and a suppression coefficient generation unit. The interference intensity analysis unit extracts the spatiotemporal features of the spatiotemporal coupling weight matrix. The spatiotemporal features are sequentially processed by multi-head attention and max pooling operations to obtain intermediate features. A fully connected layer is used to map the intermediate features to interference intensity scores. The suppression coefficient generation unit adjusts the steepness parameter of the sigmoid function according to the average value of the interference intensity scores. The suppression coefficient generation unit uses the sigmoid function to obtain the suppression coefficient based on the steepness parameter and the interference intensity scores. The interference suppression module generates a suppression heatmap based on the suppression coefficient and the location parameters of the smoke exhaust equipment.

6. The system according to claim 1, characterized in that, The collaborative decision-making module includes a strategy module and a value evaluation module. The strategy module equips each natural smoke exhaust device and mechanical smoke exhaust device with an independent strategy sub-network. Spatial attention feature map, suppression heat map and position parameters of the smoke exhaust device are input into the strategy sub-network to obtain the action probability distribution of the smoke exhaust device. The value evaluation module establishes an independent value assessment model for each smoke exhaust device, calculates the advantage function value and objective function of each state-action pair, updates the parameters of the strategy sub-network according to the advantage function and objective function, and after multiple iterations, outputs the fan speed of each smoke exhaust fan and the opening degree of the smoke exhaust window that maximizes the objective function.

Images