CN117558116A - Fire control early warning system based on big data - Google Patents

Abstract

The invention provides a fire protection early warning system based on big data, which relates to the technical field of fire protection early warning and comprises the following components: the environment monitoring information loading module; the environment fluctuation curve construction module; a fluctuation distance information generation module; the first correlation analysis module is used for carrying out correlation analysis on the fluctuation distance time sequence information and the fluctuation time sequence to generate a first correlation coefficient; a wave gradient information generation module; the second correlation analysis module is used for carrying out correlation analysis on the fluctuation gradient time sequence information and the fluctuation time sequence to generate a second correlation coefficient; a fire trigger probability generation module; and the fire protection early warning signal generation module. The invention solves the technical problems that the traditional fire-fighting early warning is that the environment condition usually meets the threshold value and can give an alarm, the complex fluctuation of the environment parameter cannot be effectively treated, the comprehensive analysis of various information is lacking, the sensitivity is lower, and the accuracy, the instantaneity and the adaptability are poorer.

Description

Fire control early warning system based on big data

Technical Field

The invention relates to the technical field of fire protection early warning, in particular to a fire protection early warning system based on big data.

Background

The fire-fighting early-warning system is widely applied to multiple fields such as cities, industrial areas and forests, and the existing fire-fighting early-warning system has a plurality of problems, firstly, the traditional fire-fighting early-warning system is usually based on a fixed threshold value, is easily influenced by environmental changes, causes insufficient sensitivity, and is difficult to adapt to complex and changeable actual environments; secondly, due to lack of deep analysis on environmental fluctuation and comprehensive consideration on various factors, the accuracy of the traditional early warning system is relatively low; third, the conventional system is difficult to respond in time in a scene with high real-time requirement, and lacks real-time processing capability for large-scale and high-dimensional data.

Disclosure of Invention

The fire-fighting early warning system based on big data aims to solve the technical problems that the traditional fire-fighting early warning system is usually that environmental conditions meet threshold values and can warn, complex fluctuation of the environmental parameters cannot be effectively treated, comprehensive analysis on various information is lacking, sensitivity is low, and accuracy, instantaneity and adaptability are poor.

In view of the above, the present application provides a fire protection early warning system based on big data.

In a first aspect of the disclosure, a fire protection early warning method based on big data is provided, and the fire protection early warning method based on big data is applied to a fire protection early warning system based on big data, wherein the system is embedded in an internet platform, and the method comprises: when the environment real-time monitoring information of the first area does not meet the fire early warning threshold value, loading environment monitoring time sequence information; constructing an environment fluctuation time sequence curve according to the environment monitoring time sequence information; based on the fire early warning threshold value, carrying out distance calculation by combining the environmental fluctuation time sequence curve to generate fluctuation distance time sequence information; carrying out correlation analysis on the fluctuation distance time sequence information and the fluctuation time sequence to generate a first correlation coefficient; performing adjacent gradient analysis on the environmental fluctuation time sequence curve to generate fluctuation gradient time sequence information; performing correlation analysis on the fluctuation gradient time sequence information and the fluctuation time sequence to generate a second correlation coefficient; activating a trigger probability prediction channel, mapping the first correlation coefficient and the second correlation coefficient, and generating fire trigger probability; and when the fire trigger probability is greater than or equal to a trigger probability threshold, generating a fire early warning signal.

In another aspect of the disclosure, a fire protection early warning system based on big data is provided, the system is embedded in an internet platform, the system is used for the method, and the system includes: the environment monitoring information loading module is used for loading environment monitoring time sequence information when the environment real-time monitoring information of the first area does not meet the fire disaster early warning threshold value; the environment fluctuation curve construction module is used for constructing an environment fluctuation time sequence curve according to the environment monitoring time sequence information; the fluctuation distance information generation module is used for carrying out distance calculation by combining the environment fluctuation time sequence curve based on the fire disaster early warning threshold value to generate fluctuation distance time sequence information; the first correlation analysis module is used for carrying out correlation analysis on the fluctuation distance time sequence information and the fluctuation time sequence to generate a first correlation coefficient; the fluctuation gradient information generation module is used for carrying out adjacent gradient analysis on the environment fluctuation time sequence curve to generate fluctuation gradient time sequence information; the second correlation analysis module is used for carrying out correlation analysis on the fluctuation gradient time sequence information and the fluctuation time sequence to generate a second correlation coefficient; the fire trigger probability generation module is used for activating a trigger probability prediction channel, mapping the first correlation coefficient and the second correlation coefficient and generating fire trigger probability; the fire protection early warning signal generation module is used for generating a fire protection early warning signal when the fire triggering probability is greater than or equal to a triggering probability threshold value.

One or more technical solutions provided in the present application have at least the following technical effects or advantages:

when the environment real-time monitoring information of the first area does not meet the fire early warning threshold value, loading environment monitoring time sequence information, so that the sensitivity and accuracy of real-time monitoring are improved, and the system is ensured to be capable of timely sensing potential fire risks; an environment fluctuation time sequence curve is constructed according to the environment monitoring time sequence information, so that the sensing capability of the system to complex environment fluctuation is enhanced, and the system has higher adaptability; by combining with a fire early warning threshold value, calculating fluctuation distance time sequence information, performing correlation analysis of a fluctuation time sequence and performing adjacent gradient analysis of an environment fluctuation time sequence curve, comprehensive understanding of fluctuation characteristics is improved and accuracy of fire prediction is improved by introducing multidimensional analysis; activating a trigger probability prediction channel, mapping the first correlation coefficient and the second correlation coefficient to generate fire trigger probability, comprehensively considering a plurality of environment attributes, and improving the accuracy and the reliability of the fire trigger probability; when the fire trigger probability is greater than or equal to the trigger probability threshold, a fire-fighting early warning signal is generated, and the accuracy and the practicability of the signal are improved. In combination, the fire-fighting early-warning method based on big data remarkably improves the performance of the fire-fighting early-warning system in the aspects of accuracy, instantaneity and adaptability by introducing big data analysis and multi-level environmental parameter processing, so that the fire-fighting early-warning method based on big data is more suitable for complex and changeable environmental conditions.

The foregoing description is only an overview of the technical solutions of the present application, and may be implemented according to the content of the specification in order to make the technical means of the present application more clearly understood, and in order to make the above-mentioned and other objects, features and advantages of the present application more clearly understood, the following detailed description of the present application will be given.

Drawings

Fig. 1 is a schematic flow chart of a fire protection early warning method based on big data in the embodiment of the application.

Fig. 2 is a schematic structural diagram of a fire protection early warning system based on big data.

Reference numerals illustrate: the system comprises an environment monitoring information loading module 10, an environment fluctuation curve construction module 20, a fluctuation distance information generation module 30, a first correlation analysis module 40, a fluctuation gradient information generation module 50, a second correlation analysis module 60, a fire trigger probability generation module 70 and a fire protection early warning signal generation module 80.

Detailed Description

According to the fire protection early warning system based on the big data, the technical problems that the traditional fire protection early warning system is usually that environmental conditions meet threshold values and can give an alarm, complex fluctuation of the environmental parameters cannot be effectively caused, comprehensive analysis on various information is lacking, sensitivity is low, and accuracy, instantaneity and adaptability are poor are solved.

Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.

Example 1

As shown in fig. 1, an embodiment of the present application provides a fire protection early warning method based on big data, where the method includes:

and when the environment real-time monitoring information of the first area does not meet the fire early warning threshold value, loading environment monitoring time sequence information.

The fire-fighting early-warning method based on the big data is applied to a fire-fighting early-warning system based on the big data, and the system is embedded in an Internet platform.

The first area refers to a monitoring area, the first is of a special meaning of no sequence and order, a sensor for monitoring environmental information is arranged in the first area, the environmental parameters comprise a temperature sensor, a humidity sensor, a smoke sensor and the like, the environmental parameters are used for monitoring various parameters of the environment in real time, a threshold value for judging fire early warning is preset, the threshold value is a preset threshold value of various environmental parameters, and when the environmental monitoring information exceeds the threshold values, a fire alarm is triggered.

When the real-time monitored environmental information does not meet the fire early warning threshold value, the system further analyzes to judge whether the current environment has potential fire risks, and at the moment, time sequence information related to the current environment is acquired from the storage device, wherein the time sequence information comprises a series of environmental parameter data which change along with time, including temperature, humidity, gas concentration and the like, and the data are integrated to construct the environmental monitoring time sequence information.

And constructing an environment fluctuation time sequence curve according to the environment monitoring time sequence information.

And establishing a coordinate axis by taking time as a horizontal axis and taking environmental parameters as a vertical axis, inputting the acquired time sequence information into the coordinate axis, and performing dotting connection on the environmental parameters of each time node in the coordinate axis to construct an environmental fluctuation time sequence curve. Meanwhile, since the environmental parameters are monitored in real time, the environmental fluctuation time sequence curve needs to be updated regularly to reflect the latest monitoring data. By constructing an environment fluctuation time sequence curve, the current real-time environment change state can be more comprehensively described, and a foundation is provided for subsequent fire early warning analysis.

And based on the fire early warning threshold value, carrying out distance calculation by combining the environmental fluctuation time sequence curve, and generating fluctuation distance time sequence information.

And finding a numerical value corresponding to the fire early-warning threshold value on a vertical axis, drawing a straight line of the fire early-warning threshold value on the coordinate axis, comparing an environmental fluctuation time sequence curve with the straight line of the fire early-warning threshold value, calculating a difference value of the environmental fluctuation time sequence curve and the fire early-warning threshold value at each moment, taking the calculated fluctuation distance and a corresponding timestamp as fluctuation distances, forming fluctuation distance time sequence information, and reflecting the deviation degree of the environmental fluctuation relative to the threshold value.

And calculating the distance between the environment fluctuation and the threshold value, and generating fluctuation distance time sequence information to evaluate the abnormality degree of the current environment, thereby providing a basis for the subsequent fire early-warning probability calculation.

And carrying out correlation analysis on the fluctuation distance time sequence information and the fluctuation time sequence to generate a first correlation coefficient.

The fluctuation time series is a time corresponding to the fluctuation distance one by one. The fluctuation distance timing information is aligned with the fluctuation time series by the time stamp, ensuring that they are corresponding in time. And performing correlation analysis by using the aligned fluctuation distance time sequence information and the fluctuation time sequence, for example, adopting a Pearson correlation coefficient method, and calculating a first correlation coefficient by the correlation analysis, wherein the first correlation coefficient reflects the degree of correlation between the fluctuation distance and the fluctuation time, wherein the range of the correlation coefficient is between-1 and 1, positive values represent positive correlation, negative values represent negative correlation, and 0 represents no correlation.

By means of correlation analysis, the degree of correlation between the fluctuation distance and the fluctuation time is quantified, and a first correlation coefficient is generated, wherein the first correlation coefficient is used for predicting the fire trigger probability.

And carrying out adjacent gradient analysis on the environment fluctuation time sequence curve to generate fluctuation gradient time sequence information.

And (3) performing gradient calculation on the environment fluctuation time sequence curve, namely calculating the slope between adjacent data points, wherein the gradient can be obtained by differential operation, such as calculating fluctuation difference between adjacent time points, dividing the difference by the time difference, and obtaining the calculated slope. And forming time sequence data by the calculated gradient and the corresponding time stamp to form fluctuation gradient time sequence information, wherein the time sequence information reflects the change speed of environmental fluctuation, namely the fluctuation gradient.

And carrying out correlation analysis on the fluctuation gradient time sequence information and the fluctuation time sequence to generate a second correlation coefficient.

The second correlation coefficient is obtained by adopting the same method as the first correlation coefficient, and for brevity of the description, details are not repeated here.

And activating a trigger probability prediction channel, mapping the first correlation coefficient and the second correlation coefficient, and generating fire trigger probability.

A channel for fire trigger probability prediction is initiated, which may employ various prediction models, such as neural networks, logistic regression, etc., with the specific choice depending on the system design and performance requirements. Mapping the first correlation coefficient and the second correlation coefficient to an activated trigger probability prediction channel, wherein the mapping is performed through a pre-trained model, and the model outputs a trigger probability estimation related to the first correlation coefficient and the second correlation coefficient according to the values of the first correlation coefficient and the second correlation coefficient as a mapping result of the first correlation coefficient and the second correlation coefficient.

The mapping results of the first correlation coefficient and the second correlation coefficient are combined, for example, through weighted average or logic operation, to generate final fire triggering probability, and the probability value represents the possibility of fire occurrence in the current environment according to environment monitoring and correlation analysis.

And when the fire trigger probability is greater than or equal to a trigger probability threshold, generating a fire early warning signal.

The fire disaster triggering probability is compared with a preset triggering probability threshold value, the triggering probability threshold value is set according to the safety standard and the application scene of the system, the sensitivity and the accuracy of the system to the fire disaster risk are affected by the selection of the threshold value, if the triggering probability is larger than or equal to the triggering probability threshold value, the existence of the fire disaster risk is judged, the generation of a triggering early warning signal is carried out, the signal can be in various forms, such as an acoustic alarm, a visual prompt, a short message notification and the like, and corresponding alarm information is sent to a fire disaster management terminal so as to take necessary measures to prevent the potential fire disaster risk.

Further, activating a trigger probability prediction channel, mapping the first correlation coefficient and the second correlation coefficient, and generating a fire trigger probability, including:

The first correlation coefficient comprises a plurality of environment attribute first correlation coefficients, and the second correlation coefficient comprises a plurality of environment attribute second correlation coefficients, wherein the plurality of environment attribute first correlation coefficients and the plurality of environment attribute second correlation coefficients are in one-to-one correspondence.

The trigger probability prediction channel comprises an environment attribute prediction network and a trigger probability fitting network.

Traversing the first correlation coefficients of the environmental attributes and the second correlation coefficients of the environmental attributes, which are in one-to-one correspondence, and sequentially performing mapping through the environmental attribute prediction network to generate a plurality of environmental attribute prediction extremums, wherein the environmental attribute prediction extremums represent characteristic values closest to a threshold value in the predicted values.

And activating the trigger probability fitting network, counting the number ratio of the environmental attribute of which the plurality of environmental attribute prediction extremum values meet the fire early-warning threshold value, and setting the number ratio as the fire trigger probability.

For the first correlation coefficient, a plurality of environment attributes are included, corresponding correlation coefficients are calculated according to correlation analysis aiming at each environment attribute, a plurality of environment attribute first correlation coefficients are obtained, the coefficients reflect the correlation degree between different environment attributes and the fluctuation time sequence, all environment attribute first correlation coefficients form a set, and the set forms the first correlation coefficient. The second correlation coefficient is the same.

Wherein each of the first and second correlation coefficients corresponds one-to-one, which means that for the same environmental property, e.g. temperature property, the temperature property correlation coefficient of the first correlation coefficient corresponds to the temperature property correlation coefficient of the second correlation coefficient, and so on.

The triggering probability prediction channel comprises an environment attribute prediction network and a triggering probability fitting network, wherein the environment attribute prediction network is a neural network model which is obtained by training a first correlation coefficient calibration data set, a second correlation coefficient calibration data set, a constraint duration calibration data set and an environment attribute extremum monitoring data set of a preset environment attribute set based on a BP neural network; the triggering probability fitting network is a machine learning model for fitting triggering probability, and the purpose of the network is to predict extremum according to input environmental attribute, and finally determine the fire triggering probability by counting the quantity proportion of the environmental attribute meeting a fire early warning threshold value.

Traversing the first correlation coefficient and the second correlation coefficient of a plurality of environmental attributes, mapping each group of correlation coefficients in one-to-one correspondence through an environmental attribute prediction network, predicting the value of the environmental attribute in the future according to the correlation coefficient and the historical monitoring data by using the trained environmental attribute prediction network in the mapping process to obtain a corresponding environmental attribute prediction value, and determining the characteristic value closest to a fire early warning threshold value in the environmental attribute prediction value obtained by mapping each group as an environmental attribute prediction extremum, wherein the extremum reflects the environmental attribute information with the most predicted fire risk in the prediction value.

And activating a trigger probability fitting network by utilizing a plurality of existing environment attribute prediction extremums, comparing the environment attribute prediction extremums with a fire early-warning threshold value for each group of environment attribute prediction extremums, judging that the environment attribute prediction extremums are more than the fire early-warning threshold value and are met, counting the number of the environment attribute prediction extremums meeting the threshold value according to a comparison result, calculating the ratio of the number of the environment attribute prediction extremums, namely the ratio of the number of the environment attribute prediction extremums to the total environment attribute prediction extremums, and setting the ratio as the fire trigger probability.

Further, mapping is performed by traversing the plurality of environmental attribute first correlation coefficients and the plurality of environmental attribute second correlation coefficients, which are in one-to-one correspondence, sequentially through the environmental attribute prediction network, so as to generate a plurality of environmental attribute prediction extremums, which previously include:

based on the first region, traversing a plurality of environment attributes, and configuring a plurality of groups of environment attribute sensitive factors, wherein any group of environment attribute sensitive factors comprises natural sensitive factors and/or industrial sensitive factors.

For any set of environmental attribute sensitivity factors:

and based on the monitoring time information, carrying out duration analysis on the natural sensitivity factor, and generating a first duration.

And based on the Internet of things, performing duration analysis on the industrial sensitivity factors to generate a second duration.

Setting the minimum value of the first duration and/or the second duration as the characteristic duration of the associated environment attribute, and adding a plurality of characteristic durations.

And traversing the first correlation coefficients of the plurality of environmental attributes and the second correlation coefficients of the plurality of environmental attributes, which are in one-to-one correspondence, by taking the maximum characteristic duration of the plurality of characteristic durations as duration constraint, and sequentially performing mapping through the environmental attribute prediction network to generate the plurality of environmental attribute prediction extremums.

Several environmental attributes of the first area are traversed, and for each environmental attribute, a state table, such as solar radiation intensity, power output of an electrical device, etc., which may cause the respective environmental element to change, may be artificially configured according to previous knowledge and experience, as part of the sensitivity factors, to define which factors are regarded as sensitivity factors under specific conditions, forming several groups of environmental attribute sensitivity factors, which are states or events that may cause the environmental element to change, and each group may include one or more natural sensitivity factors and/or industrial sensitivity factors.

Wherein, for natural factors in environmental attributes, such as temperature, humidity, illumination, etc., natural sensitivity factors are configured, including sunrise and sunset time, seasonal variation, meteorological conditions, etc., for example, when the intensity of solar radiation reaches a certain value, the natural sensitivity factors are regarded as natural sensitivity factors; can also be used as a natural sensitivity factor when the temperature rises or falls suddenly.

Configuring an industrial sensitivity factor for industrial factors in environmental attributes, such as output power of electrical equipment, chemical production emission and the like, wherein the industrial sensitivity factor is considered as the industrial sensitivity factor when the output power of certain electrical equipment reaches a certain value; the industrial emission concentration can also be used as an industrial sensitivity factor when the industrial emission concentration exceeds a certain threshold.

For any set of environmental attribute sensitivity factors:

the monitoring time information, i.e. the time point information of recording the state of the natural sensitive factor, is obtained, which may be a time stamp recorded regularly by the monitoring system for identifying the time of the change of the environmental attribute. For each natural sensitive factor, determining its state in the monitoring time information, such as illumination intensity, analyzing its duration in the monitoring time information, where the duration represents the length of time the natural sensitive factor remains in a specific state, for example, simply calculating the continuous length of time of the state of the natural sensitive factor, or by applying a statistical method to analyze the time distribution of the state, generating a first duration, where the first duration reflects the duration of the natural sensitive factor in different states.

Data related to the industrial sensitive factors are collected through the Internet of things, and the data comprise control element information related to the industrial sensitive factors, such as adjustment parameters, working modes and the like of equipment. And performing duration analysis on the state of each industrial sensitive factor, wherein the duration analysis comprises detecting the moment when the state of the industrial sensitive factor changes from one state to the other state, calculating the time interval between the two moments, and generating a second duration as the duration, wherein the second duration reflects the duration of the industrial sensitive factor in different states.

Comparing the first duration with the second duration, selecting the minimum value, and setting the minimum value as the characteristic duration of the related environment attribute, wherein the characteristic duration reflects the shortest duration of the state of the environment attribute, namely, the related environment attribute maintains a certain state in the shortest duration. The obtained feature duration is added to a number of feature duration sets, including feature durations of other environmental attributes, to provide more comprehensive environmental information.

And using the maximum characteristic duration in the plurality of characteristic durations as a duration constraint, wherein the duration constraint is used for specifying a time extremum of the environment attribute prediction, so as to ensure that a predicted result is within a specific time. And traversing the first correlation coefficients of the environmental attributes and the second correlation coefficients of the environmental attributes, and for each group of values of the correlation coefficients of the environmental attributes, performing mapping through an environmental attribute prediction network, wherein the mapping comprises the steps of predicting the change trend of the environmental attributes in a duration constraint by using a neural network model, and generating a prediction extremum of the environmental attributes in the duration constraint, namely a plurality of environmental attribute prediction extremum, according to the mapping result.

Further, by using the maximum characteristic duration of the plurality of characteristic durations as a duration constraint, traversing the plurality of environmental attribute first correlation coefficients and the plurality of environmental attribute second correlation coefficients which are in one-to-one correspondence, and sequentially performing mapping through the environmental attribute prediction network, the generating the plurality of environmental attribute prediction extremum includes:

first attribute first correlation coefficients of the plurality of environmental attribute first correlation coefficients are extracted in one-to-one correspondence, and first attribute second correlation coefficients of the plurality of environmental attribute second correlation coefficients are extracted.

Wherein the first attribute first correlation coefficient and the first attribute second correlation coefficient belong to pearson correlation coefficients.

When the first attribute first correlation coefficient does not belong to the weak correlation coefficient interval, the environment attribute prediction network is activated, mapping is conducted on the maximum characteristic duration, the first attribute first correlation coefficient and the first attribute second correlation coefficient, a first attribute prediction extremum is generated, and the plurality of environment attribute prediction extremums are added.

For each group of environment attribute first correlation coefficients, randomly extracting the first correlation coefficients of the first attributes from the correlation coefficient data; for each set of environmental attribute second correlation coefficients, randomly extracting the second correlation coefficients of its first attribute from the correlation coefficient data.

The first attribute first correlation coefficient and the first attribute second correlation coefficient belong to pearson correlation coefficients, and the pearson correlation coefficient (Pearson correlation coefficient) is a statistical index for measuring linear correlation between two variables, and the value of the pearson correlation coefficient is between-1 and 1, wherein when the correlation coefficient is 1, the first attribute first correlation coefficient and the first attribute second correlation coefficient represent complete positive correlation, namely the two variables are in linear positive correlation; when the correlation coefficient is-1, the complete negative correlation is represented, namely, the two variables are in a linear negative correlation relationship; when the correlation coefficient is 0, no linear correlation is indicated.

Since the abscissa is time and the time unit is fixed, the intensity of the pearson correlation coefficient can reflect the magnitude of the change gradient in the context of this environmental attribute, and in the time series data, the pearson correlation coefficient is used to characterize the linear relationship between the time series variables.

And judging the first correlation coefficient of the first attribute, and determining whether the first correlation coefficient belongs to a weak correlation coefficient section, wherein the section is a threshold range which is defined in advance according to specific requirements. If the first correlation coefficient of the first attribute does not belong to the weak correlation coefficient interval, the environmental attribute prediction network is activated, which means that it is determined that a stronger correlation exists, and further analysis using the prediction network is required.

Mapping, by the environmental attribute prediction network, the maximum feature duration, the first attribute first correlation coefficient, and the first attribute second correlation coefficient, including inputting the values into a neural network model, and generating a predicted extremum for the first attribute based on the mapping result, the predicted extremum representing an extreme case of a change in the first attribute over a given time horizon. And adding the generated first attribute prediction extremum to a plurality of environment attribute prediction extremum sets for subsequent analysis and calculation.

Further, when the first attribute first correlation coefficient does not belong to the weak correlation coefficient interval, the environmental attribute prediction network is activated, mapping is performed on the maximum characteristic duration, the first attribute first correlation coefficient and the first attribute second correlation coefficient, a first attribute prediction extremum is generated, and the plurality of environmental attribute prediction extremums are added, before the mapping, the method includes:

and collecting a first correlation coefficient calibration data set, a second correlation coefficient calibration data set, a constraint duration calibration data set and an environment attribute extremum monitoring data set of a preset environment attribute set.

And extracting first correlation coefficient calibration data, second correlation coefficient calibration data, constraint duration calibration data and environment attribute extremum monitoring data which are in one-to-one correspondence from the first correlation coefficient calibration data set, the second correlation coefficient calibration data set, the constraint duration calibration data set and the environment attribute extremum monitoring data set.

And analyzing the environmental attribute extremum prediction record data of the BP neural network on the first correlation coefficient calibration data, the second correlation coefficient calibration data and the constraint duration calibration data by the environmental attribute extremum monitoring data to obtain environmental attribute extremum prediction deviation.

And when the predicted deviation of the extreme value of the environmental attribute is smaller than a preset deviation threshold, directly calling the first correlation coefficient calibration data set, the second correlation coefficient calibration data set, the constraint duration calibration data set and the extreme value monitoring data set to execute training.

And when the predicted deviation of the extreme value of the environmental attribute for continuous preset times is smaller than the preset deviation threshold value, generating the predicted network of the environmental attribute.

The method comprises the steps of collecting relevant data recorded under the condition of known environmental attribute change, wherein the relevant data comprises a first correlation coefficient calibration data set, a second correlation number calibration data set, a constraint duration calibration data set and an environmental attribute extremum monitoring data set, the collected data set covers environmental attribute change under different conditions so as to ensure that an environmental attribute prediction network can work effectively under various conditions, the calibration data set is used for training and verifying the environmental attribute prediction network, the relation between the first correlation coefficient and the second correlation number and the environmental attribute change can be accurately captured, and the environmental attribute extremum monitoring data set is used for actually monitoring the change condition of the environmental attribute so as to verify the accuracy and applicability of the prediction.

And carrying out sample extraction on the basis of preset environmental attributes for each sample data set, and ensuring that the first correlation coefficient, the second correlation coefficient, the constraint duration and the environmental attribute extremum monitoring data of each sample have a one-to-one correspondence in the extraction process.

The method comprises the steps of using environment attribute extremum prediction record data of first correlation coefficient calibration data, second correlation number calibration data and constraint duration calibration data as training data of a neural network, wherein the record data comprises actual values of environment attributes and predicted values of the neural network.

A BP neural network is constructed, the BP neural network is a feedforward neural network for supervised learning, training is carried out through a back propagation algorithm, the network receives first correlation coefficient calibration data, second correlation coefficient calibration data and constraint duration calibration data as input, and an environment attribute prediction extremum is taken as output. Training the BP neural network by using training data, and adjusting the weight and deviation of the network by using a back propagation algorithm to accurately predict the extreme value of the environmental attribute, and predicting the extreme value of the environmental attribute by using the trained neural network for each sample to obtain the predicted value of the environmental attribute corresponding to the neural network.

And comparing the environment attribute value predicted by the neural network with the actual environment attribute value, and calculating the deviation between the environment attribute value and the actual environment attribute value, wherein the deviation can indicate the accuracy and the error magnitude of the prediction of the neural network, and the smaller deviation indicates that the prediction of the neural network is more accurate.

And (3) judging the condition, checking whether the calculated predicted deviation of the extreme value of the environmental attribute is smaller than a preset deviation threshold, wherein the preset deviation threshold is a value predefined according to the system requirement and the performance standard and is used for judging whether the predicted performance of the neural network is good enough.

If the predicted deviation of the extreme value of the environmental attribute is smaller than a preset deviation threshold value, training operation is performed, namely a first correlation coefficient calibration data set, a second correlation coefficient calibration data set, a constraint duration calibration data set and an extreme value monitoring data set are called, and then the BP neural network is trained by using the data sets, wherein the aim of training is to further optimize the neural network by using more data, and the prediction accuracy of the neural network on the environmental attribute is improved.

Maintaining a counter, recording the number of times of continuously meeting the condition, and increasing the counter if the predicted deviation of the extreme value of the environmental attribute of the continuous preset number of times is smaller than a preset deviation threshold value; otherwise, the counter is cleared. The preset times are defined according to the system requirements and performance standards, and the preset times are introduced to ensure that the neural network can keep good performance in a plurality of continuous training processes, so that misjudgment caused by randomness or abnormality of a certain training process can be reduced. When the predicted deviation of the extreme value of the environmental attribute of the continuous preset times is smaller than the preset deviation threshold value, the neural network is good in continuous training for several times, and at the moment, the predicted network of the environmental attribute is generated, which means that the network has good predicted performance under the current condition.

Further, the method further comprises the following steps: and when the predicted deviation of the extreme value of the environmental attribute is greater than or equal to the preset deviation threshold, performing super-parameter adjustment on the BP neural network, and then calling the first correlation coefficient calibration data set, the second correlation coefficient calibration data set, the constraint duration calibration data set and the extreme value monitoring data set to execute training.

If the predicted deviation of the extreme value of the environmental attribute is greater than or equal to a preset deviation threshold, performing super-parameter adjustment operation, including adjusting super-parameters such as learning rate, layer number, neuron number and the like of the neural network, and enabling the super-parameters of the neural network to be better adapted to the characteristics of the current environment and data by adjusting the super-parameters of the neural network so as to improve the prediction performance. And after the super-parameters are adjusted, the first correlation coefficient calibration data set, the second correlation coefficient calibration data set, the constraint duration calibration data set and the environment attribute extremum monitoring data set are re-called, and the BP neural network is retrained, so that the network is adapted to the adjusted super-parameters, and the extremum of the environment attribute is predicted more accurately.

Through super parameter adjustment and retraining, the structure and performance of the neural network are optimized to cope with the change of environmental attributes and improve the prediction accuracy.

Further, when the fire trigger probability is greater than or equal to a trigger probability threshold, generating a fire warning signal, including:

first area basic information is obtained, wherein the first area basic information comprises topography height information, inflammable and explosive quantity information and wind power level information.

And activating a fire disaster level response coordinate system, and analyzing the relief height information, the inflammable and explosive quantity information and the wind power level information to generate a fire disaster early warning level.

And generating the fire early-warning signal according to the fire early-warning level and sending the fire early-warning signal to a fire management terminal.

The method comprises the steps of obtaining the relief height information of a first area through a Geographic Information System (GIS) or other map data sources, wherein the relief height information comprises a topographic map, elevation data and the like and is used for describing the relief characteristics of the area; acquiring the quantity information of inflammable and explosive substances in the first area through a fire safety monitoring system or related data collecting equipment, wherein the information relates to the types and the quantities of the inflammable and explosive substances such as chemicals, fuels and the like stored in the area; and acquiring wind level information of a first area through a meteorological monitoring station or a meteorological data source, wherein the wind level information comprises current meteorological data such as wind speed, wind direction and the like and is used for evaluating the strength of wind power. And integrating the acquired relief height information, the inflammable and explosive quantity information and the wind power level information into a basic information set of the first area.

A predefined fire level response coordinate system is activated, the coordinate system being a three-dimensional coordinate system, wherein each zone has a pre-calibrated fire pre-alarm level, and the location of each coordinate point in three-dimensional space is related to the fire pre-alarm level for that zone.

And taking the acquired terrain height information, the inflammable and explosive quantity information and the wind power level information as inputs to form a three-dimensional coordinate point (x, y and z), wherein x represents the terrain height, y represents the inflammable and explosive quantity and z represents the wind power level. Comparing and analyzing the input three-dimensional coordinate points with the coordinate points in the activated fire level response coordinate system, including calculating the distance between the coordinate points to determine the position of the input points in the coordinate system, positioning the input three-dimensional coordinate points in the fire level response coordinate system according to the analysis result, and obtaining the fire early warning level corresponding to the position.

According to the calculated fire early-warning level, corresponding fire early-warning signals are generated, and according to different fire early-warning levels, different management priorities are pre-configured according to the severity and the emergency degree of the fire, for example, high-level fire early-warning needs more urgent and important management response so as to ensure effective utilization of resources and proper emergency response. The generated fire early warning signal is sent to the fire management terminal and can be completed in a network communication mode, a short message mode, a telephone mode and the like, so that relevant personnel can be ensured to receive the fire early warning signal in time and can take necessary actions.

By generating the fire early warning signal and sending the fire early warning signal to the fire management terminal, the response and management of the fire early warning level are realized, so that proper measures are ensured to be taken to cope with the potential fire risk.

In summary, the fire protection early warning method based on big data provided by the embodiment of the application has the following technical effects:

1. when the environment real-time monitoring information of the first area does not meet the fire early warning threshold value, the environment monitoring time sequence information is loaded, so that the sensitivity and accuracy of real-time monitoring are improved, and the system is ensured to timely sense the potential fire risk.

2. The environmental fluctuation time sequence curve is constructed according to the environmental monitoring time sequence information, so that the sensing capability of the system to complex environmental fluctuation is enhanced, and the system has higher adaptability.

3. By combining with a fire early warning threshold value, calculation of fluctuation distance time sequence information, correlation analysis of fluctuation time sequence and adjacent gradient analysis of an environment fluctuation time sequence curve are carried out, comprehensive understanding of fluctuation characteristics is improved by introducing multidimensional analysis, and accuracy of fire prediction is improved.

4. And activating a trigger probability prediction channel, mapping the first correlation coefficient and the second correlation coefficient to generate fire trigger probability, comprehensively considering a plurality of environment attributes, and improving the accuracy and the reliability of the fire trigger probability.

5. When the fire trigger probability is greater than or equal to the trigger probability threshold, a fire-fighting early warning signal is generated, and the accuracy and the practicability of the signal are improved.

In combination, the fire-fighting early-warning method based on big data remarkably improves the performance of the fire-fighting early-warning system in the aspects of accuracy, instantaneity and adaptability by introducing big data analysis and multi-level environmental parameter processing, so that the fire-fighting early-warning method based on big data is more suitable for complex and changeable environmental conditions.

Example two

Based on the same inventive concept as the fire protection early warning method based on big data in the foregoing embodiment, as shown in fig. 2, the present application provides a fire protection early warning system based on big data, where the system is embedded in an internet platform, and the system includes:

the environment monitoring information loading module 10 is configured to load environment monitoring time sequence information when the environment real-time monitoring information of the first area does not meet the fire early warning threshold value.

The environment fluctuation curve construction module 20 is used for constructing an environment fluctuation time sequence curve according to the environment monitoring time sequence information by the environment fluctuation curve construction module 20.

The fluctuation distance information generation module 30 is configured to perform distance calculation in combination with the environmental fluctuation time sequence curve based on the fire disaster early warning threshold value, and generate fluctuation distance time sequence information.

The first correlation analysis module 40 is configured to perform correlation analysis on the fluctuation distance time sequence information and the fluctuation time sequence, and generate a first correlation coefficient.

The fluctuation gradient information generation module 50 is used for carrying out adjacent gradient analysis on the environment fluctuation time sequence curve to generate fluctuation gradient time sequence information.

The second correlation analysis module 60 is configured to perform correlation analysis on the fluctuation gradient timing information and the fluctuation time sequence, and generate a second correlation coefficient.

And the fire trigger probability generation module 70 is configured to activate a trigger probability prediction channel, map the first correlation coefficient and the second correlation coefficient, and generate a fire trigger probability.

The fire protection early warning signal generation module 80 is configured to generate a fire protection early warning signal when the fire trigger probability is greater than or equal to a trigger probability threshold.

Further, the fire trigger probability generation module includes:

the correlation coefficient description module is used for enabling the first correlation coefficient to comprise a plurality of environment attribute first correlation coefficients, and enabling the second correlation coefficient to comprise a plurality of environment attribute second correlation coefficients, wherein the plurality of environment attribute first correlation coefficients and the plurality of environment attribute second correlation coefficients are in one-to-one correspondence.

And the trigger probability prediction channel description module is used for the trigger probability prediction channel to comprise an environment attribute prediction network and a trigger probability fitting network.

The environment attribute prediction extremum generation module is used for traversing the plurality of environment attribute first correlation coefficients and the plurality of environment attribute second correlation coefficients which are in one-to-one correspondence, and mapping is performed through the environment attribute prediction network in sequence to generate a plurality of environment attribute prediction extremums, wherein the environment attribute prediction extremum represents a characteristic value closest to a threshold value in the prediction values.

And the fitting network activation module is used for activating the triggering probability fitting network, counting the number ratio of the environmental attribute of which the plurality of environmental attribute prediction extremum values meet the fire early warning threshold value, and setting the number ratio as the fire triggering probability.

Further, the environment attribute prediction extremum generation module includes:

the sensitive factor configuration module is used for traversing a plurality of environment attributes based on the first area and configuring a plurality of groups of environment attribute sensitive factors, wherein any group of environment attribute sensitive factors comprise natural sensitive factors and/or industrial sensitive factors.

The sensitivity factor analysis module is used for aiming at any group of environment attribute sensitivity factors:

And based on the monitoring time information, carrying out duration analysis on the natural sensitivity factor, and generating a first duration.

And based on the Internet of things, performing duration analysis on the industrial sensitivity factors to generate a second duration.

Setting the minimum value of the first duration and/or the second duration as the characteristic duration of the associated environment attribute, and adding a plurality of characteristic durations.

And traversing the first correlation coefficients of the plurality of environmental attributes and the second correlation coefficients of the plurality of environmental attributes, which are in one-to-one correspondence, by taking the maximum characteristic duration of the plurality of characteristic durations as duration constraint, and sequentially performing mapping through the environmental attribute prediction network to generate the plurality of environmental attribute prediction extremums.

Further, the sensitivity factor analysis module further includes:

the correlation coefficient extraction module is used for extracting first correlation coefficients of first attributes of the plurality of environmental attribute first correlation coefficients in one-to-one correspondence, and first correlation coefficients of second attributes of the plurality of environmental attribute second correlation coefficients.

And the coefficient description module is used for obtaining a first attribute first correlation coefficient and a first attribute second correlation coefficient, wherein the first attribute first correlation coefficient and the first attribute second correlation coefficient belong to a pearson correlation coefficient.

And the mapping module is used for activating the environment attribute prediction network when the first attribute first correlation coefficient does not belong to the weak correlation coefficient interval, mapping the maximum characteristic duration, generating a first attribute prediction extremum, and adding the first attribute prediction extremum into the plurality of environment attribute prediction extremums.

Further, the mapping module includes:

the data set acquisition module is used for acquiring a first correlation coefficient calibration data set, a second correlation coefficient calibration data set, a constraint duration calibration data set and an environment attribute extremum monitoring data set of a preset environment attribute set.

The calibration data extraction module is used for extracting first correlation coefficient calibration data, second correlation coefficient calibration data, constraint duration calibration data and environment attribute extremum monitoring data which are in one-to-one correspondence from the first correlation coefficient calibration data set, the second correlation coefficient calibration data set, the constraint duration calibration data set and the environment attribute extremum monitoring data set.

And the prediction deviation acquisition module is used for analyzing the environment attribute extremum prediction record data of the BP neural network on the first correlation coefficient calibration data, the second correlation coefficient calibration data and the constraint duration calibration data through the environment attribute extremum monitoring data to acquire the environment attribute extremum prediction deviation.

And the training module is used for directly calling the first correlation coefficient calibration data set, the second correlation coefficient calibration data set, the constraint duration calibration data set and the environment attribute extremum monitoring data set to execute training when the environment attribute extremum prediction deviation is smaller than a preset deviation threshold.

And the prediction network generation module is used for generating the environment attribute prediction network when the environment attribute extremum prediction deviation of the continuous preset times is smaller than the preset deviation threshold value.

Further, the device further comprises a super-parameter adjusting module, which is used for performing super-parameter adjustment on the BP neural network when the predicted deviation of the extreme value of the environmental attribute is greater than or equal to the preset deviation threshold, and then retrieving the first correlation coefficient calibration data set, the second correlation coefficient calibration data set, the constraint duration calibration data set and the extreme value monitoring data set to perform training.

Further, the fire protection early warning signal generation module includes:

the base information acquisition module is used for acquiring first area base information, wherein the first area base information comprises terrain height information, inflammable and explosive quantity information and wind power level information.

The information analysis module is used for activating a fire disaster level response coordinate system, and analyzing the relief height information, the inflammable and explosive quantity information and the wind power level information to generate a fire disaster early warning level.

And the early warning signal generation module is used for generating the fire protection early warning signal according to the fire disaster early warning level and sending the fire protection early warning signal to the fire disaster management terminal.

In the present disclosure, through the foregoing detailed description of a fire protection early warning method based on big data, it is clear to those skilled in the art that a fire protection early warning system based on big data in this embodiment is described relatively simply for the device disclosed in the embodiment, and relevant places refer to the description of the method section.

The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (8)

1. The utility model provides a fire control early warning system based on big data, its characterized in that, a fire control early warning system based on big data embedded in internet platform, a fire control early warning system based on big data include:

the environment monitoring information loading module is used for loading environment monitoring time sequence information when the environment real-time monitoring information of the first area does not meet the fire disaster early warning threshold value;

the environment fluctuation curve construction module is used for constructing an environment fluctuation time sequence curve according to the environment monitoring time sequence information;

the fluctuation distance information generation module is used for carrying out distance calculation by combining the environment fluctuation time sequence curve based on the fire disaster early warning threshold value to generate fluctuation distance time sequence information;

the first correlation analysis module is used for carrying out correlation analysis on the fluctuation distance time sequence information and the fluctuation time sequence to generate a first correlation coefficient;

the fluctuation gradient information generation module is used for carrying out adjacent gradient analysis on the environment fluctuation time sequence curve to generate fluctuation gradient time sequence information;

The second correlation analysis module is used for carrying out correlation analysis on the fluctuation gradient time sequence information and the fluctuation time sequence to generate a second correlation coefficient;

the fire trigger probability generation module is used for activating a trigger probability prediction channel, mapping the first correlation coefficient and the second correlation coefficient and generating fire trigger probability;

the fire protection early warning signal generation module is used for generating a fire protection early warning signal when the fire triggering probability is greater than or equal to a triggering probability threshold value.

2. The fire protection early warning system based on big data of claim 1, wherein the fire trigger probability generation module comprises:

the correlation coefficient description module is used for enabling the first correlation coefficient to comprise a plurality of environment attribute first correlation coefficients, and enabling the second correlation coefficient to comprise a plurality of environment attribute second correlation coefficients, wherein the plurality of environment attribute first correlation coefficients and the plurality of environment attribute second correlation coefficients are in one-to-one correspondence;

the trigger probability prediction channel description module is used for the trigger probability prediction channel to comprise an environment attribute prediction network and a trigger probability fitting network;

The environment attribute prediction extremum generation module is used for traversing the plurality of environment attribute first correlation coefficients and the plurality of environment attribute second correlation coefficients which are in one-to-one correspondence, and mapping is performed through the environment attribute prediction network in sequence to generate a plurality of environment attribute prediction extremums, wherein the environment attribute prediction extremum represents a characteristic value closest to a threshold value in the prediction values;

and the fitting network activation module is used for activating the triggering probability fitting network, counting the number ratio of the environmental attribute of which the plurality of environmental attribute prediction extremum values meet the fire early warning threshold value, and setting the number ratio as the fire triggering probability.

3. The fire protection pre-warning system based on big data of claim 2, wherein the environmental attribute prediction extremum generation module comprises:

the sensitive factor configuration module is used for traversing a plurality of environment attributes based on the first area and configuring a plurality of groups of environment attribute sensitive factors, wherein any group of environment attribute sensitive factors comprise natural sensitive factors and/or industrial sensitive factors;

the sensitivity factor analysis module is used for aiming at any group of environment attribute sensitivity factors:

based on monitoring time information, carrying out duration analysis on the natural sensitive factors to generate a first duration;

Based on the Internet of things, performing duration analysis on the industrial sensitive factors to generate a second duration;

setting the minimum value of the first duration and/or the second duration as the characteristic duration of the associated environment attribute, and adding a plurality of characteristic durations;

and traversing the first correlation coefficients of the plurality of environmental attributes and the second correlation coefficients of the plurality of environmental attributes, which are in one-to-one correspondence, by taking the maximum characteristic duration of the plurality of characteristic durations as duration constraint, and sequentially performing mapping through the environmental attribute prediction network to generate the plurality of environmental attribute prediction extremums.

4. The fire protection pre-warning system based on big data of claim 3, wherein the sensitivity factor analysis module further comprises:

the correlation coefficient extraction module is used for extracting first attribute first correlation coefficients of the plurality of environmental attribute first correlation coefficients in one-to-one correspondence, and first attribute second correlation coefficients of the plurality of environmental attribute second correlation coefficients;

the coefficient description module is used for obtaining a first attribute first correlation coefficient and a first attribute second correlation coefficient of the first attribute, wherein the first attribute first correlation coefficient and the first attribute second correlation coefficient belong to a pearson correlation coefficient;

And the mapping module is used for activating the environment attribute prediction network when the first attribute first correlation coefficient does not belong to the weak correlation coefficient interval, mapping the maximum characteristic duration, generating a first attribute prediction extremum, and adding the first attribute prediction extremum into the plurality of environment attribute prediction extremums.

5. The fire protection pre-warning system based on big data of claim 4, wherein the mapping module comprises:

the data set acquisition module is used for acquiring a first correlation coefficient calibration data set, a second correlation coefficient calibration data set, a constraint duration calibration data set and an environment attribute extremum monitoring data set of a preset environment attribute set;

the calibration data extraction module is used for extracting first correlation coefficient calibration data, second correlation coefficient calibration data, constraint duration calibration data and environment attribute extremum monitoring data which are in one-to-one correspondence from the first correlation coefficient calibration data set, the second correlation coefficient calibration data set, the constraint duration calibration data set and the environment attribute extremum monitoring data set;

the prediction deviation obtaining module is used for analyzing the environment attribute extremum prediction record data of the BP neural network on the first correlation coefficient calibration data, the second correlation coefficient calibration data and the constraint duration calibration data through the environment attribute extremum monitoring data to obtain environment attribute extremum prediction deviation;

The training module is used for directly calling the first correlation coefficient calibration data set, the second correlation coefficient calibration data set, the constraint duration calibration data set and the environment attribute extremum monitoring data set to execute training when the environment attribute extremum prediction deviation is smaller than a preset deviation threshold;

and the prediction network generation module is used for generating the environment attribute prediction network when the environment attribute extremum prediction deviation of the continuous preset times is smaller than the preset deviation threshold value.

6. The fire protection early warning system based on big data according to claim 5, further comprising a super-parameter adjustment module, configured to perform super-parameter adjustment on the BP neural network when the predicted deviation of the extreme value of the environmental attribute is greater than or equal to the preset deviation threshold, and then retrieve the first correlation coefficient calibration data set, the second correlation coefficient calibration data set, the constraint duration calibration data set, and the extreme value monitoring data set to perform training.

7. The fire protection early warning system based on big data of claim 1, wherein the fire protection early warning signal generation module comprises:

the base information acquisition module is used for acquiring first area base information, wherein the first area base information comprises terrain height information, inflammable and explosive quantity information and wind power level information;

The information analysis module is used for activating a fire disaster level response coordinate system, and analyzing the relief height information, the inflammable and explosive quantity information and the wind power level information to generate a fire disaster early warning level;

and the early warning signal generation module is used for generating the fire protection early warning signal according to the fire disaster early warning level and sending the fire protection early warning signal to the fire disaster management terminal.

8. The big data-based fire protection early warning method is characterized by being used for implementing the big data-based fire protection early warning system according to any one of claims 1-7, wherein the big data-based fire protection early warning system is embedded in an internet platform, and the big data-based fire protection early warning method comprises the following steps:

when the environment real-time monitoring information of the first area does not meet the fire early warning threshold value, loading environment monitoring time sequence information;

constructing an environment fluctuation time sequence curve according to the environment monitoring time sequence information;

based on the fire early warning threshold value, carrying out distance calculation by combining the environmental fluctuation time sequence curve to generate fluctuation distance time sequence information;

carrying out correlation analysis on the fluctuation distance time sequence information and the fluctuation time sequence to generate a first correlation coefficient;

Performing adjacent gradient analysis on the environmental fluctuation time sequence curve to generate fluctuation gradient time sequence information;

performing correlation analysis on the fluctuation gradient time sequence information and the fluctuation time sequence to generate a second correlation coefficient;

activating a trigger probability prediction channel, mapping the first correlation coefficient and the second correlation coefficient, and generating fire trigger probability;

and when the fire trigger probability is greater than or equal to a trigger probability threshold, generating a fire early warning signal.