Intelligent fire safety control method and system

Abstract

The invention relates to an intelligent fire safety control system which comprises a sensing layer, a transmission layer, a processing layer and an application layer. The sensing layer is responsible for collecting environmental factor data and response data of fire fighting equipment; the transmission layer is responsible for transmitting data acquired by the sensing layer to the processing layer; the big data processing center stores and processes the data uploaded by the transmission layer, the environmental factor compensation and filtering module is responsible for identifying the influence of environmental factors on fire fighting equipment, evaluating a nonlinear relationship between different features and the state of the fire fighting equipment through a decision function of a support vector machine, extracting features having significant influence on the fire fighting equipment, and sending the extracted features to the fire fighting equipment; and training a neural network model by using the extracted features, learning the relationship between the environmental factors and the fire-fighting equipment states, and inputting new environmental factor data into the trained neural network model to obtain fire-fighting equipment state data.

Description

Technical Field

[0001] The present invention relates to the technical field of fire safety, and particularly to an intelligent fire safety control method and system. Background Art

[0002] The Internet of Things technology can connect various fire-fighting devices to realize real-time collection and transmission of data, enabling the fire-fighting system to immediately grasp the latest dynamics of the fire situation. Through the analysis of multi-dimensional data such as historical fire data, environmental information, and building structures, big data technology can predict potential fire hazards, establish risk models, and thus conduct effective prevention and resource allocation. Intelligent fire protection is a technological revolution in the modern fire protection field. It uses advanced technologies such as the Internet of Things, big data, and artificial intelligence to empower traditional fire protection systems and build an efficient and intelligent fire prevention and disaster prevention system. The implementation of intelligent fire protection depends on the coordinated action of multiple advanced technologies, realizing intelligent management of the entire process of fire prediction, early warning, and disposal.

[0003] Due to the complexity and diversity in the industrial environment, fire-fighting devices may be affected by various interference factors. Traditional fire protection systems often have difficulty effectively identifying and compensating for these effects, resulting in frequent false alarms or missed alarms. Therefore, an intelligent fire safety control method and system are proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent fire safety control method and system.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions:

[0006] An intelligent fire safety control system includes a sensing layer, a transmission layer, a processing layer, and an application layer;

[0007] The sensing layer includes a sensor acquisition module and an intelligent control module. The sensor acquisition module deploys various sensors, such as smoke detectors, temperature sensors, flame detectors, etc., for real-time collection of environmental factor data and the response data of fire-fighting devices. The sensors can sense key indicators such as smoke concentration, temperature change, and the presence of flames in the environment, and convert this information into electrical signals for transmission. The intelligent control module is configured on fire-fighting devices, such as fire extinguishing bombs, sprinklers, etc., to implement the remote control and self-check functions of the devices. It is responsible for receiving the emergency plan control instructions from the application layer, remotely controlling the fire-fighting devices, such as starting sprinkling water, releasing fire extinguishing bombs, etc. At the same time, it self-checks the operating status of the devices to ensure that the devices are in good working condition;

[0008] The transmission layer includes a data communication module and a data integration and protocol conversion module. The data communication module uses communication methods such as LoRa, WiFi, and 5G to be responsible for transmitting the data collected by the sensing layer to the processing layer, establishing a communication link, and ensuring the accurate transmission and efficient processing of data. The data integration and protocol conversion module integrates the data from different sensors and devices, forms a unified data format, and performs protocol conversion to adapt to the data exchange requirements between different devices and systems;

[0009] The functional modules of the processing layer include a big data processing center, an environmental factor compensation and filtering module, a fire type recognition and alarm parameter adjustment module, and a fire risk prediction and assessment module. The big data processing center stores and processes the data uploaded by the transmission layer. The big data processing center can receive the data from the transmission layer, store and back up it, and at the same time, using big data analysis and artificial intelligence technologies, deeply mine and analyze the data, extract useful fire warning information and features. The environmental factor compensation and filtering module is responsible for identifying the impacts of environmental factors such as electromagnetic interference and temperature fluctuations on fire-fighting equipment, preprocessing the environmental factor data through a support vector machine, evaluating the non-linear relationship between different features and the state of fire-fighting equipment through the decision function of the support vector machine, extracting the features that have a significant impact on fire-fighting equipment, and using the extracted significant impact features and the response situation of fire-fighting equipment to train a neural network model. The neural network model learns the relationship between environmental factors and the state of fire-fighting equipment through multiple layers of neurons and weight adjustment. Input the new environmental factor data into the trained neural network model to obtain the sensor data and the state data of fire-fighting equipment after compensation or filtering. The fire type recognition and alarm parameter adjustment module is responsible for identifying the features of different fire types, adjusting the alarm parameters according to the identified features of the fire type, training and learning the fire data, establishing a fire type recognition model, and automatically adjusting the alarm parameters according to the recognition result to ensure the timeliness and accuracy of the alarm. The fire risk prediction and assessment module is responsible for predicting and assessing the fire risk, predicting the possible future fire risk according to the historical fire data and the current environmental information, and giving corresponding countermeasures and suggestions;

[0010] The application layer includes a fire alarm module and an emergency handling module. The fire alarm module receives the fire warning information from the processing layer and conveys the fire information to relevant personnel and institutions in a timely manner through methods such as audible and visual alarms and SMS notifications. The emergency handling module automatically or manually activates the corresponding emergency plan according to the fire type and severity, such as evacuating personnel, starting the sprinkler system, contacting the rescue agency, etc., to minimize the losses caused by the fire;

[0011] The environmental factor compensation and filtering module receives sensor data and device status data from the perception layer, as well as historical data from the processing layer, and identifies the impacts of environmental factors such as electromagnetic interference and temperature fluctuations on fire-fighting equipment. The environmental factor compensation and filtering module transmits the compensated or filtered sensor data and fire-fighting equipment status data to the big data processing center. The big data processing center transmits the analyzed fire warning information to the fire type identification and alarm parameter adjustment module for compensation or filtering. The fire type identification and alarm parameter adjustment module identifies the characteristics of different fire types and automatically adjusts the alarm parameters to adapt to different fire situations. The big data processing center transmits historical fire data and current environmental information to the fire risk prediction and assessment module.

[0012] The above technical solution further includes:

[0013] Further, the perception layer includes a sensor acquisition module and an intelligent control module. The sensor acquisition module includes a sensor unit and a signal processing unit. The sensor unit, which is the basic unit of the sensor acquisition module, includes a smoke detector, a temperature sensor, a flame detector, etc., and is responsible for sensing various physical quantities such as smoke concentration, temperature change, and the presence of flame in the environment, and converting the physical quantities into electrical signals. The signal processing unit is responsible for receiving the electrical signals output by the sensor unit and performing processing such as amplification, filtering, and conversion to ensure the accuracy and stability of the signals.

[0014] Further, the intelligent control module includes an intelligent control unit, a fire-fighting equipment monitoring unit, and an actuator unit. The intelligent control unit, which is the core part of the intelligent control module, is responsible for receiving the emergency plan control instructions from the application layer, remotely controlling the fire-fighting equipment, and performing a self-check function. The fire-fighting equipment monitoring unit is used to monitor the operating status of the fire-fighting equipment, such as parameters such as temperature, pressure, and flow rate. The actuator unit is responsible for the emergency plan control instructions from the application layer and performing specific emergency operations, such as starting a sprinkler, releasing a fire extinguishing bomb, etc.

[0015] Further, the data communication module includes a communication control unit, a data transmission unit, and a data verification unit. The communication control unit is responsible for establishing and managing communication links, including selecting communication methods, configuring communication parameters, monitoring communication status, etc. The data transmission unit is responsible for transmitting the data of the perception layer to the processing layer through the selected communication method. During the data transmission process, the data verification unit verifies the data to ensure the integrity and accuracy of the data. The data integration and protocol conversion module includes a data integration unit and a protocol conversion unit. The data integration unit performs integration processing on the data of the perception layer, including operations such as data cleaning, deduplication, and format conversion, to form a unified data format. The protocol conversion unit is responsible for implementing protocol conversion between different devices and systems, which includes parsing the protocol format of the original data and converting it into the protocol format required by the target device or system. The data integration unit sends the integrated data to the protocol conversion unit. The protocol conversion unit receives the integrated data and performs protocol conversion operations according to the requirements of the target device or system. The protocol conversion unit sends the converted data to the data transmission unit.

[0016] Further, the environmental factor compensation and filtering module includes a data preprocessing unit, a model training unit, and a data compensation and filtering unit. The data preprocessing unit uses a support vector machine to preprocess the environmental factor data, identify and extract the features that have a significant impact on the fire-fighting equipment. The data preprocessing unit preprocesses the response data of the fire-fighting equipment. The model training unit uses the extracted significant features and the response situation of the fire-fighting equipment to train a neural network model. The model training unit receives the significant feature set output by the data preprocessing unit and the response data of the fire-fighting equipment, and uses these data to train the neural network model so that it can learn the relationship between environmental factors and the state of the fire-fighting equipment. The data compensation and filtering unit inputs the new environmental factor data into the trained neural network model to obtain the fire-fighting equipment state data.

[0017] Further, the data preprocessing unit uses a support vector machine to preprocess the environmental factor data, identify and extract the features that have a significant impact on the fire-fighting equipment. The specific steps are as follows:

[0018] Data collection and cleaning:

[0019] Collect environmental factor data, such as temperature, humidity, electromagnetic interference intensity, etc., clean the data, and remove noise, outliers, and missing values;

[0020] Feature selection and extraction:

[0021] After data cleaning, support vector machines are used for feature selection and extraction. Support vector machines are a supervised learning algorithm that can find the optimal classification hyperplane by maximizing the classification margin. In the feature selection and extraction stage, we can use the weight vector of the support vector machine to evaluate the importance of each feature. The basic form of a support vector machine is a linear classifier, and the decision function of the support vector machine is f(x) = sign(w·x + b), where w is the weight vector, x is the input feature vector, and b is the bias term. Each component of the weight vector w represents the importance of the corresponding feature. Features with large weight vectors w contribute more to the classification decision and are regarded as features that have a significant impact on fire-fighting equipment. The calculation formula for the weight vector w is w = ∑α i *y i *x i , where α i is the Lagrange multiplier of the support vector, y i is the label of the support vector (+1 or -1), and x i is the feature vector of the support vector;

[0022] Feature scaling and standardization:

[0023] After feature selection and extraction, the features are scaled and standardized to ensure that different features are numerically comparable. By scaling the data to a specific range (such as 0 to 1) or standardizing it to data with zero mean and unit variance, the formula for feature scaling is

[0024] The formula for standardization is

[0025] where x is the original feature value, x min and x max are the minimum and maximum values of the feature respectively, and μ and σ are the mean and standard deviation of the feature respectively;

[0026] Constructing a feature vector:

[0027] After feature selection and extraction, scaling and standardization, the selected significant features are combined into a feature vector;

[0028] Furthermore, the model training unit trains a neural network model. The specific steps are as follows:

[0029] Data preparation:

[0030] Integrate the extracted significantly influential features and the response situation of the fire-fighting equipment into a data set. The data set usually includes input features (i.e., environmental factor features) and output labels (i.e., the response situation of the fire-fighting equipment). Divide the data set into a training set, a validation set, and a test set for model evaluation and optimization during the training process;

[0031] Neural network construction:

[0032] Select a multi-layer perceptron (MLP) as the neural network model. The multi-layer perceptron (MLP) is a feed-forward neural network composed of an input layer, a hidden layer, and an output layer. The input layer of the neural network receives input features, the hidden layer performs non-linear transformation on the input features, and the output layer generates prediction results. The calculation formula for the output of each layer is:

[0033] z (l) = W (l) a (l-1) + b (l)

[0034] a (l) = σ(z (l) )

[0035] where z (l) is the sum of the weighted input and bias of the l-th layer, W (l) is the weight matrix of the l-th layer, b (l) is the bias vector of the l-th layer, a (l-1) is the output (i.e., activation value) of the (l - 1)-th layer, and σ is the activation function (such as ReLU, sigmoid, or tanh);

[0036] Training parameter setting:

[0037] Set the training parameters of the neural network, including learning rate, batch size, number of iterations, optimization algorithm, etc.;

[0038] Model training:

[0039] Use the training set data to train the neural network. During the training process, calculate the prediction results through forward propagation, calculate the gradients through backward propagation, and update the weights and biases. During the backward propagation process, the update calculation formula for the weights is

[0040] where η is the learning rate, L is the loss function (such as cross-entropy loss, mean squared error, etc.), is the gradient of the loss function with respect to the weights;

[0041] Model evaluation and optimization:

[0042] Use the validation set data to evaluate the trained model, and optimize the model according to the evaluation results. Optimization includes adjusting the network architecture, adding regularization terms, changing the learning rate, etc.;

[0043] Model testing:

[0044] Use the test set data to test the optimized model to evaluate the generalization ability of the model.

[0045] A control method for an intelligent fire safety control system includes the following steps:

[0046] Data reception and preprocessing: The processing layer receives data from the transmission layer. The data includes fire warning information (such as smoke concentration, temperature, flame state, etc.) and device status information collected by the sensing layer;

[0047] Environmental factor compensation and filtering: The environmental factor compensation and filtering module preprocesses environmental factor data through a support vector machine, extracts significant features affecting fire-fighting equipment, uses the extracted significant features and the response of fire-fighting equipment to train a neural network model, and inputs new environmental factor data into the trained neural network model to obtain compensated or filtered sensor data and fire-fighting equipment status data;

[0048] Fire type identification and alarm parameter adjustment: The fire type identification and alarm parameter adjustment module extracts fire features from the preprocessed data, such as burning speed, smoke concentration change rate, flame color, etc., identifies the fire type according to the extracted features, and the fire type identification and alarm parameter adjustment module automatically adjusts alarm parameters, such as alarm thresholds, alarm response times, etc., to ensure the timeliness and accuracy of alarms;

[0049] Fire risk prediction and assessment: The big data processing center predicts and assesses fire risks to provide a data basis for decision-making support;

[0050] Data storage and report generation: The processing layer stores the processed data, identification results, alarm parameter adjustment records, and fire risk prediction results in the big data processing center and generates corresponding reports for management personnel to view.

[0051] The present invention has the following beneficial effects:

[0052] 1. In the present invention, an environmental factor compensation and filtering module is introduced. This module uses an advanced support vector machine algorithm to process the environmental factor data in the factory, extracts the features that significantly affect the fire-fighting equipment, and then trains a neural network model based on these features and the response data of the fire-fighting equipment. The neural network model can intelligently learn and identify the complex relationship between environmental factors and the state of fire-fighting equipment. The big data processing center uses the complex relationship between environmental factors and the state of fire-fighting equipment to eliminate the influence of interference factors on the fire-fighting equipment. As a result, the false alarm rate of the system is significantly reduced. Even in the face of a complex electromagnetic interference environment, it can maintain a high degree of accuracy and reliability, which not only greatly saves fire-fighting resources, but also avoids the panic and chaos of employees caused by false alarms, providing a strong guarantee for the normal production of the factory.

[0053] 2. In the present invention, the fire type recognition and alarm parameter adjustment module recognizes the characteristics of different fire types, such as wood fires, grease fires, etc. According to the recognized fire type, it automatically adjusts alarm parameters, such as alarm thresholds, response times, etc., to adapt to different fire situations, improving the pertinence and timeliness of fire early warning, ensuring the timeliness and accuracy of alarm information, realizing the intelligent adjustment of alarm parameters, and reducing the difficulty and complexity of manually setting parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a system block diagram of a smart fire safety control method and system proposed by the present invention;

[0055] Figure 2 It is a method step diagram of a smart fire safety control method and system proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0057] Please refer to Figure 1 - Figure 2 As shown, the present invention is a smart fire safety control system, including a sensing layer, a transmission layer, a processing layer, and an application layer;

[0058] The perception layer includes a sensor acquisition module and an intelligent control module. The sensor acquisition module deploys various sensors, such as smoke detectors, temperature sensors, flame detectors, etc., to collect environmental factor data and the response data of fire-fighting equipment in real time. The sensors can sense key indicators such as smoke concentration, temperature change, and the presence of flames in the environment, and convert this information into electrical signals for transmission. The intelligent control module is configured on fire-fighting equipment, such as fire extinguishing bombs, sprinklers, etc., to realize the remote control and self-check functions of the equipment. It is responsible for receiving the emergency plan control instructions from the application layer, remotely controlling the fire-fighting equipment, such as starting sprinkling water, releasing fire extinguishing bombs, etc. At the same time, it self-checks the operating status of the equipment to ensure that the equipment is in good working condition;

[0059] The transmission layer includes a data communication module and a data integration and protocol conversion module. The data communication module uses communication methods such as LoRa, WiFi, 5G, etc., to be responsible for transmitting the data collected by the perception layer to the processing layer, establishing a communication link to ensure the accurate transmission and efficient processing of data. The data integration and protocol conversion module integrates the data from different sensors and devices, forms a unified data format, and performs protocol conversion to meet the data exchange requirements between different devices and systems;

[0060] The processing layer function module includes a big data processing center, an environmental factor compensation and filtering module, a fire type identification and alarm parameter adjustment module, and a fire risk prediction and assessment module. The big data processing center stores and processes the data uploaded by the transmission layer. The big data processing center can receive the data from the transmission layer, store and back up it, and at the same time, use big data analysis and artificial intelligence technologies to deeply mine and analyze the data, extract useful fire warning information and features. The environmental factor compensation and filtering module is responsible for identifying the impacts of environmental factors such as electromagnetic interference and temperature fluctuations on fire-fighting equipment. It preprocesses the environmental factor data through a support vector machine, evaluates the non-linear relationship between different features and the state of fire-fighting equipment through the decision function of the support vector machine, extracts the features that have significant impacts on fire-fighting equipment, and uses the extracted significant impact features and the response of fire-fighting equipment to train a neural network model. The neural network model learns the relationship between environmental factors and the state of fire-fighting equipment through multiple layers of neurons and weight adjustment. Input the new environmental factor data into the trained neural network model to obtain the fire-fighting equipment state data. The fire type identification and alarm parameter adjustment module is responsible for identifying the features of different fire types and automatically adjusting the alarm parameters to adapt to different fire situations. It trains and learns the fire data to establish a fire type identification model, adjusts the alarm parameters according to the features of the identified fire type, trains and learns the fire data to establish a fire type identification model, and automatically adjusts the alarm parameters according to the identification result to ensure the timeliness and accuracy of the alarm. The fire risk prediction and assessment module is responsible for predicting and assessing the fire risk. According to the historical fire data and the current environmental information, it predicts the possible future fire risk and gives corresponding countermeasures and suggestions;

[0061] The application layer includes a fire alarm module and an emergency handling module. The fire alarm module receives the fire warning information from the processing layer and promptly conveys the fire information to relevant personnel and institutions through means such as audible and visual alarms and text message notifications. The emergency handling module automatically or manually activates the corresponding emergency plan according to the fire type and severity, such as evacuating personnel, starting the sprinkler system, contacting the rescue agency, etc., to minimize the losses caused by the fire;

[0062] The environmental factor compensation and filtering module receives sensor data and device status data from the perception layer, as well as historical data from the processing layer, identifies the impacts of environmental factors such as electromagnetic interference and temperature fluctuations on fire-fighting equipment, and the environmental factor compensation and filtering module transmits the compensated or filtered sensor data and fire-fighting equipment status data to the big data processing center. The big data processing center transmits the analyzed fire warning information to the fire type identification and alarm parameter adjustment module for compensation or filtering. The fire type identification and alarm parameter adjustment module identifies the characteristics of different fire types and automatically adjusts the alarm parameters to adapt to different fire situations. The big data processing center transmits historical fire data and current environmental information to the fire risk prediction and assessment module.

[0063] In one embodiment, for the above-mentioned sensor acquisition module, the perception layer includes a sensor acquisition module and an intelligent control module. The sensor acquisition module includes a sensor unit and a signal processing unit. The sensor unit, which is the basic unit of the sensor acquisition module, includes a smoke detector, a temperature sensor, a flame detector, etc., and is responsible for sensing various physical quantities such as smoke concentration, temperature change, and the presence of flames in the environment, and converting the physical quantities into electrical signals. The signal processing unit is responsible for receiving the electrical signals output by the sensor unit and performing processing such as amplification, filtering, and conversion to ensure the accuracy and stability of the signals.

[0064] In one embodiment, for the above-mentioned intelligent control module, the intelligent control module includes an intelligent control unit, a fire-fighting equipment monitoring unit, and an actuator unit. The intelligent control unit, which is the core part of the intelligent control module, is responsible for receiving the emergency plan control instructions from the application layer, remotely controlling the fire-fighting equipment, and performing a self-check function. The fire-fighting equipment monitoring unit is used to monitor the operating status of the fire-fighting equipment, such as parameters such as temperature, pressure, and flow rate. The actuator unit is responsible for the emergency plan control instructions from the application layer and performing specific emergency operations, such as starting a sprinkler, releasing a fire extinguishing bomb, etc.

[0065] In one embodiment, for the above data communication module, the data communication module includes a communication control unit, a data transmission unit, and a data verification unit. The communication control unit is responsible for establishing and managing communication links, including selecting communication methods, configuring communication parameters, monitoring communication status, etc. The data transmission unit is responsible for transmitting the data from the sensing layer to the processing layer through the selected communication method. During the data transmission process, the data verification unit verifies the data to ensure the integrity and accuracy of the data. The data integration and protocol conversion module includes a data integration unit and a protocol conversion unit. The data integration unit performs integration processing on the data from the sensing layer, including operations such as data cleaning, deduplication, and format conversion, to form a unified data format. The protocol conversion unit is responsible for implementing protocol conversion between different devices and systems, which includes parsing the protocol format of the original data and converting it into the protocol format required by the target device or system. The data integration unit sends the integrated data to the protocol conversion unit. The protocol conversion unit receives the integrated data and performs protocol conversion operations according to the requirements of the target device or system. The protocol conversion unit sends the converted data to the data transmission unit.

[0066] In one embodiment, for the above environmental factor compensation and filtering module, the environmental factor compensation and filtering module includes a data preprocessing unit, a model training unit, and a data compensation and filtering unit. The data preprocessing unit uses a support vector machine to preprocess the environmental factor data, identify and extract the features that have a significant impact on the fire-fighting equipment. The data preprocessing unit preprocesses the response data of the fire-fighting equipment. The model training unit uses the extracted significant features and the response situation of the fire-fighting equipment to train a neural network model. The model training unit receives the significant feature set and the response data of the fire-fighting equipment output by the data preprocessing unit, and uses these data to train the neural network model so that it can learn the relationship between the environmental factors and the state of the fire-fighting equipment. The data compensation and filtering unit inputs the new environmental factor data into the trained neural network model to obtain the state data of the fire-fighting equipment.

[0067] In one embodiment, for the above data preprocessing unit, the data preprocessing unit uses a support vector machine to preprocess the environmental factor data, identify and extract the features that have a significant impact on the fire-fighting equipment. The specific steps are as follows:

[0068] Data collection and cleaning:

[0069] Collect environmental factor data, such as temperature, humidity, electromagnetic interference intensity, etc., clean the data, and remove noise, outliers, and missing values;

[0070] Suppose we have a dataset containing temperature, humidity, and electromagnetic interference intensity, where some data points may be abnormal due to sensor failures or transmission errors. We need to use data cleaning techniques such as mean filling, median filling, or interpolation to handle these abnormal data points;

[0071] Feature Selection and Extraction:

[0072] After data cleaning, support vector machines are used for feature selection and extraction. Support vector machines are a supervised learning algorithm that can find the optimal classification hyperplane by maximizing the classification margin. In the feature selection and extraction stage, we can use the weight vector of the support vector machine to evaluate the importance of each feature. The basic form of a support vector machine is a linear classifier, and its decision function is f(x) = sign(w·x + b), where w is the weight vector, x is the input feature vector, and b is the bias term. Each component of the weight vector w represents the importance of the corresponding feature. Features with a large weight vector w contribute more to the classification decision and are considered features that have a significant impact on fire-fighting equipment. The calculation formula for the weight vector w is w = ∑α i *y i *x i , where α i is the Lagrange multiplier of the support vector, y i is the label of the support vector (+1 or -1), and x i is the feature vector of the support vector;

[0073] Suppose we have a dataset with 10 features. By training a support vector machine model, we can obtain the weight of each feature. Then, we can sort the features according to the absolute value of the weight and select the features with larger absolute values of the weight as significant features;

[0074] Feature Scaling and Standardization:

[0075] After feature selection and extraction, the features are scaled and standardized to ensure that different features are numerically comparable. By scaling the data to a specific range (such as 0 to 1) or standardizing it to data with zero mean and unit variance, the formula for feature scaling is

[0076] The formula for standardization is

[0077] where x is the original feature value, x min and x max are the minimum and maximum values of the feature respectively, and μ and σ are the mean and standard deviation of the feature;

[0078] Suppose we have a temperature feature with an eigenvalue range from 0 to 100, which we can scale to the range from 0 to 1. Similarly, for other features, we can perform similar scaling or normalization;

[0079] Construct a feature vector:

[0080] After feature selection and extraction, scaling and normalization, combine the selected significant features into a feature vector;

[0081] Suppose we have selected temperature, humidity, and electromagnetic interference intensity as significant features and have scaled and normalized them. Then, we can combine these features into a three-dimensional feature vector, such as [xtemperature, xhumidity, xEMI].

[0082] In one embodiment, for the above model training unit, the model training unit trains a neural network model, and the specific steps are as follows:

[0083] Data preparation:

[0084] Integrate the extracted significant features and the response of the fire protection equipment into a data set. The data set usually includes input features (i.e., environmental factor features) and output labels (i.e., the response of the fire protection equipment), and divide the data set into a training set, a validation set, and a test set for model evaluation and optimization during the training process;

[0085] Suppose we have a data set, where the input features include temperature, humidity, and electromagnetic interference intensity, and the output label is the startup status of the fire protection equipment (0 means not started, 1 means started). We divide the data set into a 70% training set, a 15% validation set, and a 15% test set;

[0086] Neural network construction:

[0087] Select a multi-layer perceptron (MLP) as the neural network model. The multi-layer perceptron (MLP) is a feedforward neural network composed of an input layer, a hidden layer, and an output layer. The input layer of the neural network receives input features, the hidden layer performs a non-linear transformation on the input features, and the output layer generates a prediction result. The calculation formula for the output of each layer is:

[0088] z (l) =W (l) a (l-1) +b (l)

[0089] a (l) =σ(z (l) )

[0090] where, z(l) is the sum of the weighted input and the bias of the l-th layer, W (l) is the weight matrix of the l-th layer, b (l) is the bias vector of the l-th layer, a (l-1) is the output (i.e., activation value) of the (l - 1)-th layer, and σ is the activation function (such as ReLU, sigmoid, or tanh);

[0091] Assume that the input layer has 3 neurons (corresponding to 3 input features), the hidden layer can have several neurons (determined according to experience or experiments), and the output layer has 1 neuron (corresponding to the activation state of the fire-fighting equipment);

[0092] Training parameter setting:

[0093] Set the training parameters of the neural network, including the learning rate, batch size, number of iterations, optimization algorithm, etc.;

[0094] We can choose a learning rate of 0.01, a batch size of 32, 1000 iterations, and an optimization algorithm of Stochastic Gradient Descent (SGD) or Adam;

[0095] Model training:

[0096] Use the training set data to train the neural network. During the training process, calculate the prediction results through forward propagation, calculate the gradients through backpropagation, and update the weights and biases. During the backpropagation process, the update calculation formula for the weights is

[0097] where η is the learning rate, L is the loss function (such as cross-entropy loss, mean squared error, etc.), is the gradient of the loss function with respect to the weights;

[0098] During the training process, use the cross-entropy loss function to evaluate the difference between the prediction results and the true labels, and use the Adam optimization algorithm to update the weights and biases;

[0099] Model evaluation and optimization:

[0100] Use the validation set data to evaluate the trained model, and optimize the model according to the evaluation results. Optimization includes adjusting the network architecture, adding regularization terms, changing the learning rate, etc.;

[0101] If the accuracy on the validation set is not high, try to increase the number of hidden layers or neurons, or use techniques such as dropout to prevent overfitting;

[0102] Model testing:

[0103] Use the test set data to test the optimized model to evaluate the generalization ability of the model;

[0104] If the accuracy rate on the test set meets the requirements, then we can consider this model to be effective and can be used in practice.

[0105] A control method for an intelligent fire safety control system, comprising the following steps:

[0106] Data reception and preprocessing: The processing layer receives data from the transport layer, and the data includes fire warning information (such as smoke concentration, temperature, flame state, etc.) and device status information collected by the sensing layer;

[0107] Suppose the transport layer sends a data record containing a smoke concentration of 500 ppm, a temperature of 25 °C, and no flame state;

[0108] Environmental factor compensation and filtering: The environmental factor compensation and filtering module preprocesses environmental factor data through a support vector machine, extracts significant features affecting fire-fighting equipment, uses the extracted significant features and the response of fire-fighting equipment to train a neural network model, and inputs new environmental factor data into the trained neural network model to obtain compensated or filtered sensor data and fire-fighting equipment status data;

[0109] Suppose the system identifies that there is strong electromagnetic interference in the current environment, and this interference may cause false alarms of smoke detectors. The processing layer will compensate the smoke concentration data according to historical data and machine learning models to reduce the false alarm rate caused by electromagnetic interference;

[0110] For another example, the system detects large temperature fluctuations, but the smoke concentration and flame state are normal. The processing layer may temporarily ignore the temperature fluctuations to avoid missed alarms caused thereby;

[0111] Fire type identification and alarm parameter adjustment: The fire type identification and alarm parameter adjustment module extracts fire characteristics from the preprocessed data, such as combustion speed, smoke concentration change rate, flame color, etc., identifies the fire type according to the extracted characteristics, and the fire type identification and alarm parameter adjustment module automatically adjusts alarm parameters, such as alarm threshold, alarm response time, etc., to ensure the timeliness and accuracy of alarms;

[0112] Suppose the system identifies that the current fire type is a rapidly burning wood fire. The fire type identification and alarm parameter adjustment module will automatically adjust the alarm threshold according to the characteristics of wood fires (such as rapid increase in smoke concentration, yellowish flame color, etc.) so that the system can give an alarm in time when the smoke concentration reaches a relatively low level;

[0113] For another example, when it is recognized that the current fire type is a slowly burning grease fire, the fire type recognition and alarm parameter adjustment module will extend the alarm response time to avoid false alarms caused by fluctuations in the smoke concentration generated by the burning of grease;

[0114] Fire risk prediction and assessment: The big data processing center predicts and assesses the fire risk, providing a data basis for decision-making support;

[0115] Based on historical data and current fire characteristics, predict the possible development trend and spread range of the fire in the next period of time, providing accurate fire information and positioning for emergency rescue agencies and personnel;

[0116] Data storage and report generation: The processing layer stores the processed data, recognition results, alarm parameter adjustment records, and fire risk prediction results in the big data processing center and generates corresponding reports for managers to view;

[0117] The system generates a detailed report containing current fire warning information, environmental factor compensation and filtering results, fire type recognition and alarm parameter adjustment records, and fire risk prediction results for managers to make decisions and respond emergently.

[0118] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A smart fire safety control system, characterized in that: Includes perception layer, transmission layer, processing layer and application layer; The perception layer is responsible for collecting environmental factor data and fire equipment response data; The transmission layer is responsible for transmitting the data collected by the perception layer to the processing layer; The processing layer includes a big data processing center, an environmental factor compensation and filtering module, a fire type identification and alarm parameter adjustment module, and a fire risk prediction and assessment module. The big data processing center stores and processes the data uploaded by the transmission layer. The environmental factor compensation and filtering module is responsible for identifying the impact of environmental factors on fire-fighting equipment, preprocessing environmental factor data through a support vector machine, evaluating the nonlinear relationship between different features and the state of fire-fighting equipment through the decision function of the support vector machine, extracting significant features that affect fire-fighting equipment, and using the extracted significant features and the response of fire-fighting equipment to train a neural network model. The neural network model learns the relationship between environmental factors and the state of fire-fighting equipment through multiple layers of neurons and weight adjustment, and inputs new environmental factor data into the trained neural network model to obtain compensated or filtered sensor data and fire-fighting equipment state data. The fire type identification and alarm parameter adjustment module is responsible for identifying the characteristics of different fire types, and adjusting the alarm parameters according to the characteristics of the identified fire types. The fire risk prediction and assessment module is responsible for predicting and assessing fire risks; The application layer includes a fire alarm module and an emergency processing module. The fire alarm module receives fire warning information from the processing layer and conveys the fire information to relevant personnel and institutions. The emergency processing module automatically activates the corresponding emergency plan according to the type and severity of the fire.

2. The intelligent fire safety control system according to claim 1, characterized in that: The perception layer includes a sensor acquisition module and an intelligent control module. The sensor acquisition module includes a sensor unit and a signal processing unit. The sensor unit is responsible for sensing environmental factor data in the environment and response data of fire-fighting equipment, and converting physical quantities into electrical signals. The signal processing unit is responsible for receiving the electrical signals output by the sensor unit and performing data processing.

3. The intelligent fire safety control system according to claim 2, characterized in that: The intelligent control module includes an intelligent control unit, a fire equipment monitoring unit and an actuator unit. The intelligent control unit is responsible for receiving the emergency plan control instructions of the application layer, remotely controlling the fire equipment, and performing self-test functions. The fire equipment monitoring unit is used to monitor the operating status of the fire equipment. The actuator unit is responsible for the emergency plan control instructions of the application layer and performs specific emergency operations.

4. The intelligent fire safety control system according to claim 1, characterized in that: The data communication module includes a communication control unit, a data transmission unit and a data verification unit. The communication control unit is responsible for establishing and managing communication links. The data transmission unit is responsible for transmitting the data of the perception layer to the processing layer through the selected communication method. During the data transmission process, the data verification unit verifies the data. The data integration and protocol conversion module includes a data integration unit and a protocol conversion unit. The data integration unit integrates and processes the data of the perception layer. The protocol conversion unit is responsible for realizing protocol conversion between different devices and systems. The data integration unit sends the integrated data to the protocol conversion unit. The protocol conversion unit receives the integrated data and performs protocol conversion operations according to the controlled device. The protocol conversion unit sends the converted data to the data transmission unit.

5. The intelligent fire safety control system according to claim 1, characterized in that: The environmental factor compensation and filtering module includes a data preprocessing unit, a model training unit and a data compensation and filtering unit. The data preprocessing unit uses a support vector machine to preprocess the environmental factor data, identify and extract features that have a significant impact on fire-fighting equipment, and preprocesses the response data of the fire-fighting equipment. The model training unit uses the extracted significant features and the response of the fire-fighting equipment to train a neural network model. The model training unit receives the significant feature set output by the data preprocessing unit and the response data of the fire-fighting equipment, and uses these data to train the neural network model so that it can learn the relationship between environmental factors and the status of fire-fighting equipment. The data compensation and filtering unit inputs new environmental factor data into the trained neural network model to obtain fire-fighting equipment status data.

6. The intelligent fire safety control system according to claim 5, characterized in that: The data preprocessing unit preprocesses the environmental factor data using a support vector machine to identify and extract features that have a significant impact on firefighting equipment. The specific steps are: Data collection and cleaning: Collect environmental factor data, clean the data, and remove noise, outliers, and missing values; Feature selection and extraction: After data cleaning, support vector machine is used for feature selection and extraction. The basic form of support vector machine is a linear classifier. The decision function of support vector machine is f(x)=sign(w·x+b), where w is the weight vector, x is the input feature vector, b is the bias term, and each component of the weight vector w represents the importance of the corresponding feature. Features with large weight vector w contribute greatly to the classification decision and are regarded as features with significant impact on fire equipment. The calculation formula of weight vector w is w=∑α i *y i *x i , where α i is the Lagrange multiplier of the support vector, y i is the label of the support vector (+1 or -1), x i is the eigenvector of the support vector; Feature Scaling and Normalization: After feature selection and extraction, the features are scaled and standardized by scaling the data to a specific range or standardizing it to data with zero mean and unit variance. The formula for feature scaling is The standardized formula is Among them, x is the original eigenvalue, x min and x max are the minimum and maximum values ​​of the feature, μ and σ are the mean and standard deviation of the feature, respectively; Construct the feature vector: After feature selection and extraction, scaling and normalization, the selected salient features are combined into a feature vector.

7. The intelligent fire safety control system according to claim 5, characterized in that: The model training unit trains a neural network model, specifically in the following steps: Data preparation: Integrate the extracted significant influencing features and the response of the fire-fighting equipment into a data set, wherein the data set includes input features and output labels, wherein the extracted significant influencing features are input features, and the response of the fire-fighting equipment is output labels, and the data set is divided into a training set, a validation set, and a test set; Neural network construction: A multilayer perceptron is selected as a neural network model. The multilayer perceptron is a feedforward neural network consisting of an input layer, a hidden layer, and an output layer. The input layer of the neural network receives input features, the hidden layer performs nonlinear transformation on the input features, and the output layer generates prediction results. The calculation formula for the output of each layer is: z (l) =W (l) a (l-1) +b (l) a (l) =σ(z (l) ) Among them, z (l) is the sum of the weighted input and bias of the lth layer, W (l) is the weight matrix of the lth layer, b (l) is the bias vector of the lth layer, a (l-1) is the output of the l-1th layer (i.e., the activation value), and σ is the activation function; Training parameter settings: Set the training parameters of the neural network; Model training: The neural network is trained using the training set data. During the training process, the prediction results are calculated through forward propagation, and the gradients are calculated and the weights and biases are updated through back propagation. During the back propagation process, the weight update calculation formula is Where η is the learning rate, L is the loss function, is the gradient of the loss function with respect to the weight; Model evaluation and optimization: Use the validation set data to evaluate the trained model and optimize the model based on the evaluation results Model testing: Use the test set data to test the optimized model.

8. The control method of a smart fire safety control system according to claim 1, characterized in that: The following steps are involved: Data reception and preprocessing: The processing layer receives data from the transmission layer, including fire warning information and equipment status information collected by the perception layer; Environmental factor compensation and filtering: The environmental factor compensation and filtering module pre-processes the environmental factor data through the support vector machine, extracts the significant features that affect the fire-fighting equipment, and trains a neural network model using the extracted significant features and the response of the fire-fighting equipment. The new environmental factor data is input into the trained neural network model to obtain the compensated or filtered sensor data and fire-fighting equipment status data; Fire type identification and alarm parameter adjustment: The fire type identification and alarm parameter adjustment module extracts fire features from the preprocessed data, identifies the fire type based on the extracted features, and automatically adjusts the alarm parameters based on the identified fire type; Fire risk prediction and assessment: The big data processing center predicts and assesses fire risks, providing a data basis for decision support; Data storage and report generation: The processing layer stores the processed data, identification results, alarm parameter adjustment records, and fire risk prediction results in the big data processing center, and generates corresponding reports for management personnel to review.

Images