Dangerous source isolation disposal method for lifting transverse moving type new energy stereo garage

Abstract

The invention discloses a danger source isolation disposal method for a lifting and transverse moving type new energy stereo garage, and the method comprises the steps: constructing a two-dimensional fire-fighting layout mapping model of the stereo garage, and building an explosion time prediction constraint equation based on vehicle temperature state parameters; a deep reinforcement learning algorithm is introduced to calculate an optimal evacuation path of the dangerous vehicle, a state space and an action space are defined, and a multi-target reward function fusing the moving cost of the dangerous vehicle and the scheduling cost of the obstacle vehicle is designed; the difficulty of obstacle vehicle scheduling is evaluated, and a greedy algorithm is adopted for priority scheduling according to the moving cost; and self-adaptive generation of the equipment control instruction is realized through a real-time state feedback mechanism. By means of the method, on the premise that the disposal timeliness is guaranteed, the danger source isolation evacuation efficiency and the system safety are remarkably improved, and the technical bottlenecks of garage evacuation path solidification and resource scheduling lagging are effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of safety control of stereo garages, and in particular to a method for isolating and disposing hazardous sources in a lifting and transversely moving new energy stereo garage. Background Art

[0002] With the popularity of new energy vehicles, stereo garages are widely used due to their high space utilization rate. Among them, lifting and sliding stereo garages are famous for their "economical and convenient".

[0003] However, the risk of battery spontaneous combustion in new energy vehicles poses a severe challenge to garage safety. At present, the fire protection system of stereo garages mainly relies on fixed sprinkler fire extinguishing devices and gas explosion suppression systems.

[0004] For example, the Chinese patent document with publication number CN111734183A discloses a three-dimensional parking garage with a fire extinguishing function, which can prevent vehicles in the three-dimensional parking garage from igniting each other and reduce losses. At the same time, water is sprayed through the sprinkler head to cool and extinguish the fire. However, this method fails to fully utilize the space scheduling capacity inside the garage in the early stage of the fire, resulting in the inability to isolate the danger source in a timely and effective manner.

[0005] The Chinese patent document with publication number CN113668921A discloses a charging stereo garage and a supporting fire control method, which detects the temperature of each charging device through a temperature detection device, and judges whether the charging is abnormal based on the detected temperature value. When the temperature is abnormal, the corresponding vehicle loading plate of the abnormal charging device is moved to the fire truck position together with the vehicle for isolation, and the charging device is controlled to be powered off and stopped, and the dry powder spraying device is controlled to spray dry powder on the vehicle in the fire isolation space to extinguish the fire. However, the control method does not mention how to move the vehicle with abnormal temperature to the fire truck position in the most efficient way.

[0006] Therefore, how to quickly evacuate high-risk vehicles within a limited time and effectively dispatch resources for emergency response has become a technical problem that needs to be solved urgently. Summary of the invention

[0007] The present invention provides a method for isolating and disposing hazardous sources in a lifting and transversely moving new energy stereo garage, which can significantly improve the efficiency of hazardous source isolation and evacuation and system safety while ensuring the timeliness of disposal, and effectively solve the technical bottlenecks of solidified garage evacuation paths and delayed resource scheduling.

[0008] A method for isolating and disposing hazardous sources of a lifting and transversely moving new energy stereo garage comprises the following steps: (1) Construct a two-dimensional fire protection layout mapping model for a three-dimensional parking garage, collect vehicle temperature state parameters and mark dangerous vehicles, and establish an explosion time prediction constraint equation; (2) Define the state space and action space, design a multi-objective reward function that integrates the movement cost of dangerous vehicles and the dispatch cost of obstacle vehicles, train the path planning model based on the reinforcement learning algorithm, and calculate multiple evacuation path plans for dangerous vehicles; (3) Evaluate the difficulty of dispatching the obstructing vehicle, dynamically adjust the position of the obstructing vehicle based on the greedy algorithm to free up the fire passage, optimize the total dispatch time and meet the hard constraint of the explosion time; Select the path with the shortest transfer time from multiple evacuation paths as the optimal evacuation path, and calculate the moving order of the obstacle vehicles on the optimal evacuation path; (4) Based on the optimal evacuation path and the movement order of the obstacle vehicles on the optimal evacuation path, the adaptive generation of equipment control instructions is realized through the real-time state feedback mechanism to ensure that the dangerous vehicles are transferred to the fire truck position within the explosion time constraint.

[0009] Furthermore, in step (1), the two-dimensional fire layout mapping model maps the three-dimensional space of the stereo garage into a two-dimensional grid plane, and defines the parking space coordinate set as ; In the formula, Indicates The coordinates of the parking space, Indicates the total number of parking spaces; Get the temperature sensor Vehicle temperature of the vehicle ,like , it is marked as a dangerous vehicle, represents the critical temperature threshold; the explosion time prediction constraint equation is defined as: ; In the formula, represents the explosion time prediction constraint, is the material explosion coefficient.

[0010] Furthermore, in step (2), the state space is defined as ,in, are common vehicle coordinates; represents the coordinates of dangerous vehicles, represents the coordinates of the fire truck; the action space is limited to ,in Indicates the lateral direction. Indicates the lifting direction.

[0011] Furthermore, in step (2), a multi-objective reward function is designed that integrates the dangerous vehicle movement cost and the obstacle vehicle dispatch cost. The formula is: ; In the formula, A fixed reward for reaching the fire truck station. Only when a dangerous vehicle reaches the fire truck station , otherwise 0, , represents the weight coefficient, which is used to weigh the cost of dangerous vehicle movement and the cost of obstacle vehicle dispatch. is the actual moving step length of the dangerous vehicle, is the dispatching cost of the obstructed vehicles on the evacuation path, The calculation formula is: ; In the formula, It represents the obstacle density parameter, which is used to indicate the number of vehicles in the four directions of lifting, lowering and lateral movement of the obstacle vehicle.

[0012] Furthermore, in step (2), the path planning model is trained based on the deep reinforcement learning algorithm, and the training loss function is defined as: ; In the formula, represents the loss value, Indicates in status Take action against dangerous vehicles Rewards received after and the next state , is the discount factor, Indicates that in the next state Next, all actions Maximum value, Indicates the current status and actions of value, and They are the parameters of the Q network and target network of the deep reinforcement learning algorithm respectively; The present invention adopts the DQN (Deep Q Network) algorithm for training. The DQN model includes a Q network, a target network and an experience replay unit. The Q network is an agent that is trained to generate the best state-action value. Q is a Q-table (a value table), which is the expectation of the benefit that can be obtained by taking action a in the state s at a certain moment. The environment will feedback the corresponding reward r according to the agent's action, so the action that can obtain the maximum benefit is selected according to the Q value; the role of the experience replay unit is to interact with the environment and generate data to train the Q network; the target network is used to calculate the maximum reward value that can be obtained by estimating the next state corresponding to the action. The target network and the Q network are exactly the same at the beginning. The experience replay method is used to select samples, and the transfer samples obtained in each iteration are Stored in the experience pool, each time taking out part of the data for training, after a certain round of training, the parameters of the Q network will be synchronized to the target network, and the action selection is adopted Greedy strategy: In the early stage of training, actions are randomly selected with a higher probability to explore more state space. As training progresses, the probability of random selection is gradually reduced, and actions are selected more based on the current Q value.

[0013] Furthermore, the specific process of step (3) is as follows: Extract multiple evacuation paths output by step (2) , and filter the obstacle vehicle collection , calculate the distance from the obstacle vehicle to the dangerous vehicle and arrange them in ascending order, and calculate the transfer time of the dangerous vehicle under each evacuation path based on the principle of clearing the obstacle vehicle closest to the dangerous vehicle first; The evacuation path with the shortest transfer time is selected as the optimal evacuation path, and the moving order of the obstacle vehicles on the optimal evacuation path is calculated by the greedy algorithm.

[0014] Furthermore, the transfer time of dangerous vehicles under each evacuation path is calculated using the formula: ; In the formula, is the moving distance of the obstruction dispatching vehicle, is the device speed constant, is the dispatching cost of the obstructed vehicles on the evacuation path, is the evacuation time cost of the dangerous vehicle path. The evacuation process must meet the following explosion time constraints: ; In the formula, represents the transfer time of dangerous vehicles under each evacuation path, represents the explosion time prediction constraint.

[0015] Furthermore, in step (4), the adaptive generation of device control instructions is achieved through a real-time state feedback mechanism, specifically: The optimal evacuation path and the movement order of the obstacle vehicles on the optimal evacuation path are parsed into a sequence of equipment control instructions. After completing a single instruction, the real-time status matrix is ​​updated to monitor the movement status of the dangerous vehicles until they reach the fire truck position.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention achieves efficient hazard source isolation and vehicle evacuation by constructing a two-dimensional fire protection layout mapping model of a stereo garage, combining a deep reinforcement learning algorithm and a greedy scheduling optimization strategy. The hazard source isolation disposal method proposed by the present invention not only improves the safety of the garage, but also optimizes the resource scheduling efficiency, overcoming the shortcomings of the prior art in terms of evacuation path solidification and scheduling lag. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flow chart of a method for isolating and disposing hazardous sources in a lifting and transversely moving new energy stereo garage.

[0018] Figure 2 It is a schematic diagram of two-dimensional fire protection layout mapping in an embodiment of the present invention.

[0019] Figure 3 Schematic diagram of multiple evacuation paths calculated by the path planning model in an embodiment of the present invention.

[0020] Figure 4 This is the DQN network architecture of the path planning model in the embodiment of the present invention.

[0021] Figure 5 This is a scheduling flow chart of the greedy algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be pointed out that the embodiments described below are intended to facilitate the understanding of the present invention and do not have any limiting effect on the present invention.

[0023] like Figure 1 As shown, a method for isolating and disposing hazardous sources of a lifting and transversely moving new energy stereo garage comprises the following steps: Step S1: construct a two-dimensional fire protection layout mapping model of a three-dimensional parking garage, and define a parking space coordinate set as: ; Fire truck position coordinates , vehicles and fire trucks layout see Figure 2 , set critical temperature threshold for dangerous vehicles .

[0024] The vehicle is collected through the temperature sensor The temperature exceeds the state parameter, marking it as a dangerous vehicle, and obtaining the material explosion coefficient of the battery by querying the vehicle model , vehicle temperature Establish the explosion time prediction constraint equation: ; In the formula, is the material explosion coefficient.

[0025] Step S2: Build a path planning model based on the deep Q network (DQN), calculate the optimal evacuation path for dangerous vehicles, and define the state space ,in is the coordinate of an ordinary vehicle, which is an obstacle vehicle on the evacuation path; the action space is limited to , lateral displacement ( direction) or lift ( Taking into account the movement cost of dangerous vehicles and the dispatch cost of obstacle vehicles, the multi-objective reward function is defined as: ; In the formula, A fixed reward for reaching the fire truck station. Only when a dangerous vehicle reaches the fire truck station , otherwise 0, , represents the weight coefficient, which is used to weigh the cost of dangerous vehicle movement and the cost of obstacle vehicle dispatch. is the actual moving step length of the dangerous vehicle, The cost of scheduling vehicles with obstacles on the evacuation path. In order to estimate the scheduling time of vehicles with obstacles, set the obstacle density parameter , used to indicate the number of vehicles around the obstacle vehicle (in four directions of lifting, lifting and lateral movement), and to evaluate the difficulty of vehicle dispatching, then The calculation method is: .

[0026] like Figure 4 As shown, the DQN network structure (i.e., deep reinforcement learning algorithm) is used to train the model, and the loss function is defined as: ; In the formula, represents the loss value, Indicates in status Take action against dangerous vehicles Rewards received after and the next state , is the discount factor, Indicates that in the next state Next, all actions Maximum value, Indicates the current status and actions of value, and are the parameters of the Q network and the target network respectively.

[0027] The DQN model used includes Q network, target network and experience replay unit. Q network is an agent trained to generate the best state-action value. Q is Q-table (value table), which is the expected benefit of taking action a in state s at a certain moment. The environment will feedback the corresponding reward r according to the agent's action. Therefore, the action that can obtain the maximum benefit is selected according to the Q value. The role of the experience replay unit is to interact with the environment and generate data to train the Q network. The target network is used to calculate the maximum reward value that can be obtained by estimating the next state corresponding to the action. The target network and the Q network are exactly the same at the beginning. The experience replay method is used to select samples and the transfer samples obtained in each iteration are Stored in the experience pool, each time taking out part of the data for training, after a certain round of training, the parameters of the Q network will be synchronized to the target network, and the action selection is adopted Greedy strategy: In the early stage of training, actions are randomly selected with a higher probability to explore more state space. As training progresses, the probability of random selection is gradually reduced, and actions are selected more based on the current Q value.

[0028] like Figure 3 As shown, the path planning model trained in step S2 is used to preliminarily calculate the top 3 optimal evacuation paths.

[0029] Step S3, using a greedy algorithm to dynamically adjust the position of the obstructing vehicle to free up the fire passage. In order to select the optimal evacuation path, it is necessary to calculate the scheduling time of the obstructing vehicle on the evacuation path and extract the multiple evacuation paths output in the S2 path planning step. , and filter the obstacle vehicle collection The strategy of the greedy algorithm is to sort by the priority of the moving cost, give priority to clearing the obstacle vehicle with the greatest impact (that is, the obstacle vehicle closest to the dangerous vehicle is the priority clearing target), and calculate the number of obstacles on the path. The time to transfer dangerous vehicles in the following situations: ; In the formula, To dispatch the moving distance of the blocking vehicle, is the total cost of dispatching obstacle vehicles, which is used to adjust the impact of dispatch difficulty on the total time of dangerous vehicle transfer. is the evacuation time cost of dangerous vehicle paths, is the equipment speed constant, selected from a certain model of lifting and transverse moving stereo garage.

[0030] ; The evacuation process must meet the explosion time constraints: ; The scheduling process of the greedy algorithm is as follows Figure 5 As shown, the calculation is To obtain the most evacuated path, the greedy algorithm is used to calculate that on path 3, the vehicles are scheduled in the order of the movement of the obstacle vehicles {(2, 3), (3, 1), (5, 1)}.

[0031] Step S4: parse the S2 path planning and S3 scheduling optimization results into a device control instruction sequence ,The instruction format is {device ID, action type, target coordinates, timestamp}; after completing a single instruction, the real-time status matrix is ​​updated , monitor the movement status of dangerous vehicles until they arrive at the fire truck position.

[0032] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for isolating and treating hazardous sources in a lifting and transversely moving new energy stereo garage, characterized in that: The following steps are involved: (1) Construct a two-dimensional fire protection layout mapping model for a three-dimensional parking garage, collect vehicle temperature state parameters and mark dangerous vehicles, and establish an explosion time prediction constraint equation; (2) Define the state space and action space, design a multi-objective reward function that integrates the movement cost of dangerous vehicles and the dispatch cost of obstacle vehicles, train the path planning model based on the deep reinforcement learning algorithm, and calculate multiple evacuation path plans for dangerous vehicles; (3) Evaluate the difficulty of dispatching the obstructing vehicle, dynamically adjust the position of the obstructing vehicle based on the greedy algorithm to free up the fire passage, optimize the total dispatch time and meet the hard constraint of the explosion time; Select the path with the shortest transfer time from multiple evacuation paths as the optimal evacuation path, and calculate the moving order of the obstacle vehicles on the optimal evacuation path; (4) Based on the optimal evacuation path and the movement order of the obstacle vehicles on the optimal evacuation path, the adaptive generation of equipment control instructions is realized through the real-time state feedback mechanism to ensure that the dangerous vehicles are transferred to the fire truck position within the explosion time constraint.

2. The method for isolating and treating hazardous sources of a lifting and transversely moving new energy stereo garage according to claim 1 is characterized in that: In step (1), the two-dimensional fire layout mapping model maps the three-dimensional space of the stereo garage into a two-dimensional grid plane, and defines the parking space coordinate set as ； In the formula, Indicates The coordinates of the parking space, Indicates the total number of parking spaces; Get the temperature sensor Vehicle temperature of the vehicle ,like , it is marked as a dangerous vehicle. represents the critical temperature threshold; the explosion time prediction constraint equation is defined as: ； In the formula, represents the explosion time prediction constraint, is the material explosion coefficient.

3. The method for isolating and treating hazardous sources in a lifting and transversely moving new energy stereo garage according to claim 1 is characterized in that: In step (2), define the state space ,in, are common vehicle coordinates; represents the coordinates of dangerous vehicles, represents the fire truck position coordinates; the action space is limited to ,in Indicates the lateral direction. Indicates the lifting direction.

4. The method for isolating and treating hazardous sources in a lifting and transversely moving new energy stereo garage according to claim 3 is characterized in that: In step (2), a multi-objective reward function is designed that integrates the dangerous vehicle movement cost and the obstacle vehicle dispatch cost. The formula is: ； In the formula, A fixed reward for dangerous vehicles arriving at the fire truck position. , otherwise 0, , represents the weight coefficient, which is used to weigh the cost of dangerous vehicle movement and the cost of obstacle vehicle dispatch. is the actual moving step length of the dangerous vehicle, is the dispatching cost of the obstructed vehicles on the evacuation path, The calculation formula is: ； In the formula, It represents the obstacle density parameter, which is used to indicate the number of vehicles in the four directions of lifting, lowering and lateral movement of the obstacle vehicle.

5. The method for isolating and treating hazardous sources in a lifting and transversely moving new energy stereo garage according to claim 4 is characterized in that: In step (2), the path planning model is trained based on the deep reinforcement learning algorithm, and the training loss function is defined as: ； In the formula, represents the loss value, Indicates in status Take action against dangerous vehicles Rewards received after and the next state , is the discount factor, Indicates that in the next state Next, all actions Maximum value, Indicates the current status and actions of value, and are the parameters of the Q network and target network of the deep reinforcement learning algorithm respectively.

6. The method for isolating and treating hazardous sources in a lifting and transversely moving new energy stereo garage according to claim 1 is characterized in that: The specific process of step (3) is as follows: Extract multiple evacuation paths output by step (2) , and filter the obstacle vehicle collection , calculate the distance from the obstacle vehicle to the dangerous vehicle and arrange them in ascending order, and calculate the transfer time of the dangerous vehicle under each evacuation path based on the principle of clearing the obstacle vehicle closest to the dangerous vehicle first; The evacuation path with the shortest transfer time is selected as the optimal evacuation path, and the moving order of the obstacle vehicles on the optimal evacuation path is calculated by the greedy algorithm.

7. The method for isolating and treating hazardous sources in a lifting and transversely moving new energy stereo garage according to claim 6 is characterized in that: Calculate the transfer time of dangerous vehicles under each evacuation path, the formula is: ； In the formula, is the moving distance of the obstruction dispatching vehicle, is the device speed constant, is the dispatching cost of the obstructed vehicles on the evacuation path, is the evacuation time cost of the dangerous vehicle path. The evacuation process must meet the following explosion time constraints: ； In the formula, represents the transfer time of dangerous vehicles under each evacuation path, represents the explosion time prediction constraint.

8. The method for isolating and treating hazardous sources in a lifting and transversely moving new energy stereo garage according to claim 6 is characterized in that: In step (4), the adaptive generation of device control instructions is achieved through a real-time state feedback mechanism, specifically: The optimal evacuation path and the movement order of the obstacle vehicles on the optimal evacuation path are parsed into a sequence of equipment control instructions. After completing a single instruction, the real-time status matrix is ​​updated to monitor the movement status of the dangerous vehicles until they reach the fire truck position.

Images