A temperature collection, monitoring and alarm method for palm firefighting

Abstract

The invention discloses a temperature collection, monitoring and alarm method for palm-held firefighting, relates to the technical field of temperature monitoring for firefighting, and is used to solve the problem of inaccurate temperature monitoring prediction; the invention synchronously collects environmental factor data through temperature monitoring points, calculates the correlation between each factor and temperature after denoising, extracts multi-factor features to generate an output feature vector for temperature prediction, and constructs it as a time series matrix input into a long short-term memory network, extracts temperature trend features, predicts future trends and compares them with safety thresholds to determine whether to trigger an early warning; at the same time, through short-term, medium-term and long-term scale temperature trend analysis, fast Fourier transform, weighted moving average method and exponential smoothing method are used to identify temperature changes of different frequencies, set a three-level early warning mechanism and dynamically adjust the threshold; finally, based on early warning feedback data, a Q-learning algorithm is used to optimize the threshold and alarm strategy, form an adaptive learning closed loop, and improve the early fire warning capability.

Description

Technical Field

[0001] The present invention relates to the technical field of temperature monitoring for fire fighting, and more specifically, to a temperature collection, monitoring and alarm method for palm-held fire fighting. Background Art

[0002] With the increasing requirements for fire safety in modern industrial and living places, real-time and convenient temperature monitoring and alarm have become important requirements for ensuring safety. Traditional fire monitoring usually relies on large fixed equipment. This monitoring method has shortcomings in response speed, installation flexibility, portability, etc., and it is difficult to adapt to the changing on-site environment and temporary needs. Especially in some high-risk places, such as flammable storage areas, processing workshops and large equipment rooms, temperature changes are often early signs of fire. Mastering small temperature fluctuations is crucial for fire prevention;

[0003] The development of "Palm Fire" temperature collection and monitoring is aimed at meeting these needs and providing a portable, flexible and efficient solution. Based on the Internet of Things technology, Palm Fire uses highly sensitive temperature probes, data acquisition equipment and wireless transmission devices to upload on-site temperature data to the monitoring platform in real time. When the temperature exceeds the set safety threshold, the monitoring platform will automatically issue a variety of alarm signals to remind users to take necessary measures to prevent fire or other safety accidents.

[0004] Deficiencies in existing technologies: The current prediction algorithm for the temperature in the monitored area is relatively basic and is applicable when the temperature presents a simple linear trend. However, the temperature changes before a fire often have nonlinear characteristics. Affected by multiple factors (such as ambient temperature fluctuations, sudden heat sources, etc.), the temperature may change in a complex way. Such changes are difficult to accurately predict through linear models. In this case, it is impossible to fully capture the early signs of fire, which may lead to untimely fire warnings and reduce the safety and reliability of the monitored area. Summary of the invention

[0005] In order to overcome the above defects of the prior art, there is a solution as follows to solve the problem of inaccurate prediction of temperature acquisition in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A temperature collection, monitoring and alarm method for palm-held firefighting comprises the following steps:

[0008] Environmental factor data is collected synchronously at the temperature monitoring point, and the correlation between each environmental factor data and temperature is calculated after denoising, the multi-factor characteristics of temperature change are determined, and the output feature vector for temperature prediction is determined based on the factor fluctuations;

[0009] The output feature vector is constructed as a time series matrix and input into the long short-term memory network to extract the temperature trend features to predict the future temperature trend and compare it with the safety threshold to determine whether to trigger an early warning.

[0010] Analyze the temperature trends in the short, medium and long term, use fast Fourier transform, weighted moving average and exponential smoothing to identify temperature trends of different frequencies, set up a three-level early warning mechanism and dynamically adjust the threshold;

[0011] Record early warning feedback data, use Q-learning algorithm to optimize thresholds, and gradually adjust alarm strategies.

[0012] In a preferred embodiment, environmental factor data is collected synchronously at the temperature monitoring point, and the correlation between each environmental factor data and temperature is calculated after denoising, the multi-factor characteristics of temperature change are determined, and the output feature vector for temperature prediction is determined according to the factor fluctuations. The specific steps are as follows:

[0013] Set the data collection cycle to collect environmental factor data, including temperature, humidity, airflow intensity, equipment operation time, and environmental pressure, and store the environmental factor data in a multi-factor data set;

[0014] Apply Fourier denoising transform to remove high-frequency noise components in the data;

[0015] The mutual information analysis method is used to calculate the correlation between each factor in the environmental factor data and temperature;

[0016] Construct a multi-factor nonlinear mapping model, use a polynomial kernel function to construct a temperature-multi-factor relationship model for the selected multi-factor feature vector, and output the predicted temperature feature value;

[0017] For each factor, a weight that is adjusted over time is set as a dynamic weight function;

[0018] The predicted temperature eigenvalue is combined with the dynamic weight of the corresponding factor to obtain the output eigenvector.

[0019] In a preferred embodiment, the output feature vector is constructed as a time series matrix, input into a long short-term memory network, the temperature trend feature is extracted to predict the future temperature trend, and compared with the safety threshold to determine whether to trigger an early warning. The specific steps include:

[0020] Construct a time series matrix to determine the historical variation patterns of temperature and multiple factors;

[0021] Long short-term memory network is used for nonlinear trend modeling. The time series matrix is ​​input into the long short-term memory network to identify the nonlinear trend of temperature change.

[0022] LSTM retains or updates historical features through a gating mechanism, and the output hidden state is used as a feature for trend analysis;

[0023] After obtaining the hidden state, the hidden state is mapped to the predicted temperature trend through the fully connected layer to obtain the predicted temperature value;

[0024] The predicted temperature value is compared with the safety threshold to determine whether the temperature trend exceeds the warning range. When the predicted temperature exceeds the set safety threshold, a warning signal is automatically triggered to remind users of the risk of fire.

[0025] In a preferred embodiment, the temperature trends at short-term, medium-term and long-term scales are analyzed, and the temperature trends of different frequencies are identified by using fast Fourier transform, weighted moving average method and exponential smoothing method. The specific steps are as follows:

[0026] Define multiple time scale windows and divide the time scale into three windows: short-term, medium-term and long-term;

[0027] Conduct temperature trend analysis in each time scale window, and apply different trend analysis methods to each time scale window;

[0028] For short-term trend analysis, fast Fourier transform is used to identify the high-frequency fluctuation trend of temperature and extract the frequency characteristics of temperature changes in the short term;

[0029] Perform medium-term trend analysis and use the weighted moving average method to calculate the smoothed temperature trend in the medium-term window;

[0030] Long-term trend analysis was performed and exponential smoothing methods were applied to identify long-term temperature trends.

[0031] In a preferred embodiment, a three-level early warning mechanism is set up to dynamically adjust the threshold, including the following steps:

[0032] According to the temperature trend results at different time scales, a hierarchical early warning trigger mechanism is set up:

[0033] When the short-term trend is higher than the short-term threshold, a level 1 warning is triggered to alert users to the risk of rapid temperature rise in a short period of time;

[0034] When the medium-term trend exceeds the medium-term threshold, the second-level warning is triggered, which is used to monitor the stable rising trend of temperature and there is a fire hazard;

[0035] When the long-term trend exceeds the long-term threshold, a level 3 warning is triggered, indicating that the temperature has been rising slowly for a long time, and users need to pay attention to abnormal long-term temperature changes.

[0036] In a preferred embodiment, early warning feedback data is recorded, the alarm threshold is optimized using a Q-learning algorithm, and the alarm strategy is gradually adjusted, including the following steps:

[0037] Record the response after each warning is triggered and build a feedback data set;

[0038] The feedback data set contains the temperature characteristics, environmental conditions, and warning effectiveness information when the warning is triggered;

[0039] The Q-learning algorithm of reinforcement learning is applied to iteratively update the alarm strategy. The goal of strategy adjustment is to minimize the number of false alarms and missed alarms. The update formula is: , where s represents the combination of the current temperature feature and the environmental state, a represents the alarm strategy, and r is the reward value of the current strategy. is the learning rate, controlling the update step size, is the discount factor, which indicates the impact of future rewards on the current strategy;

[0040] After completing the adaptive adjustment of the early warning strategy, the alarm thresholds of different time scales are dynamically updated, and the strategy updated by Q-learning is mapped to the alarm threshold according to the characteristics of environmental factors in the feedback data set.

[0041] The technical effects and advantages of the temperature collection, monitoring and alarm method for palm firefighting of the present invention are as follows:

[0042] The present invention synchronously collects environmental factor data through temperature monitoring points, calculates the correlation between each factor and temperature after denoising, extracts multi-factor features and generates output feature vectors for temperature prediction, then constructs the feature vector as a time series matrix and inputs it into a long short-term memory network to extract temperature trend features, predict future temperature trends, and compare with safety thresholds to determine whether to trigger an early warning. At the same time, the temperature trends at short-term, medium-term and long-term scales are analyzed, and fast Fourier transform, weighted moving average and exponential smoothing methods are used to identify temperature changes of different frequencies. A three-level early warning mechanism is set to dynamically adjust the alarm threshold to adapt to environmental changes. Finally, by recording early warning feedback data and using the Q-learning algorithm to continuously optimize the threshold and alarm strategy, an adaptive learning closed-loop mechanism is formed, so that the temperature collection monitoring alarm for palm firefighting has efficient fire early warning capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 The present invention is a flowchart of a temperature collection, monitoring and alarm method for palm-held firefighting. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] In order to achieve the above objectives, Figure 1 A schematic diagram of the structure of a temperature collection, monitoring and alarm method for palm firefighting of the present invention is provided, which specifically comprises the following steps:

[0046] Environmental factor data is collected synchronously at the temperature monitoring point, and the correlation between each environmental factor data and temperature is calculated after denoising, the multi-factor characteristics of temperature change are determined, and the output feature vector for temperature prediction is determined based on the factor fluctuations;

[0047] The output feature vector is constructed as a time series matrix and input into the long short-term memory network to extract the temperature trend features to predict the future temperature trend and compare it with the safety threshold to determine whether to trigger an early warning.

[0048] Analyze the temperature trends in the short, medium and long term, use fast Fourier transform, weighted moving average and exponential smoothing to identify temperature trends of different frequencies, set up a three-level early warning mechanism and dynamically adjust the threshold;

[0049] Record early warning feedback data, use Q-learning algorithm to optimize thresholds, and gradually adjust alarm strategies.

[0050] Step 1: In the temperature collection and monitoring for palm firefighting, the influence of various environmental factors should be considered to identify the temperature change trend before the fire. That is, the key factors are selected through the multi-factor temperature feature recognition process and the temperature factor relationship model is established to determine the multi-factor characteristics of temperature change and improve the accuracy of subsequent nonlinear analysis and early warning. The specific steps are as follows:

[0051] During the temperature monitoring process, data of various environmental factors at each monitoring point need to be collected synchronously to ensure the comprehensiveness and accuracy of the temperature change characteristics. First, set the data collection cycle , and record the collected information such as temperature T, humidity H, airflow intensity F, equipment operation time U, and environmental pressure P in a multi-factor data set In, it is expressed as follows: ,in, Represents the multi-factor feature vector at time t;

[0052] In order to ensure data quality, Fourier denoising transform is used to remove high-frequency noise components in the data. Convert to the frequency domain, suppress the noise part whose frequency exceeds the set threshold, and then use inverse Fourier transform Restore smoothed time domain data , the processing process is: ,in, is the frequency window function, which filters out the noise components. After this step, the denoised smooth data is obtained. , to ensure the accuracy of subsequent feature analysis;

[0053] To screen the correlation features, in the environmental factor data, the relationship between temperature and various factors may have different degrees of correlation. In order to extract the factors that have the greatest impact on temperature change, the mutual information analysis method is used to measure the correlation between each factor and temperature. Quantified the temperature T and the factor The correlation between the parameters (such as humidity, airflow, etc.) is calculated as follows: ,in, is the temperature T and the factor The joint probability distribution at time t is, and are their respective marginal probability distributions;

[0054] Mutual Information Value The larger the factor, the The system sorts the mutual information values ​​of each factor and selects several factors with the highest mutual information values ​​(such as humidity and airflow intensity) as the main features to form a simplified feature vector ;

[0055] Construct a multi-factor nonlinear mapping model and select the multi-factor feature vector It is further mapped to a high-dimensional feature space to capture the nonlinear relationship between temperature and various factors. To achieve this mapping, a polynomial kernel function is used. , a relationship model between temperature and multiple factors is constructed in high-dimensional space. This relationship model can be expressed as: , where are the kernel model parameters, determined according to the optimization algorithm, is a polynomial kernel function between eigenvectors, which is used to determine the nonlinear relationship between multiple factors and temperature in high-dimensional space. c and are the constant and order of the kernel function, which regulate the complexity of the nonlinear mapping.

[0056] Through this nonlinear mapping model, the complex relationship between temperature, humidity and airflow can be captured, so that the temperature trend characteristics can be more accurately displayed under the influence of environmental changes. Finally, the output predicted temperature characteristic value It will be used as data input for the next step to support subsequent temperature monitoring and alarm judgment;

[0057] In the multi-factor temperature feature recognition, the impact of different factors on temperature changes may change dynamically over time. In order to enhance the adaptability of temperature monitoring, a dynamic weight calculation mechanism is introduced to set a weight for each environmental factor that is adjusted over time. The dynamic weight function The purpose is to adjust the influence of each factor on temperature prediction in real time according to the degree of deviation of the factor at the current moment, and define the dynamic weight function as: ,in, represents the value of the i-th factor at time t, Yes Factor The benchmark value is determined based on the median or typical value of the factor in historical data to ensure that the normal state of the factor is reflected. Indicates the fluctuation range of the factor, which is used to measure the standard deviation or dynamic range of the factor to capture the fluctuation range under normal circumstances;

[0058] It should be noted that the dynamic weight function The deviation degree of the factors is adjusted through the form of exponential function. When the value of a factor deviates greatly from the baseline value, the weight will automatically decrease to avoid the impact of abnormal fluctuations of the factor on temperature prediction; on the contrary, when the factor value is close to the baseline value, the weight remains high to normally reflect its impact on temperature changes. The introduction of this weight mechanism can effectively deal with the disturbance of temperature data caused by sudden or drastic environmental changes, ensuring that temperature prediction is more stable and flexible.

[0059] After the dynamic weight calculation is completed, the temperature characteristic value will be predicted With dynamic weights for each factor Combine to generate the final weighted output feature vector , the feature vector integrates the weight information of multiple factors and the temperature prediction value, providing reliable input data for the next step of nonlinear analysis and prediction model. The construction process is as follows: ,in, is the predicted temperature characteristic value at time t, combining the nonlinear relationship of multiple factors. The dynamic weight of each factor at time t makes the temperature prediction result in the output feature vector adaptable and stable under the influence of multiple factors;

[0060] The output feature vector constructed by weighting , can effectively integrate dynamically changing environmental factors into the temperature prediction model. Factors with different weights can automatically adjust their influence during system operation, making the temperature prediction results sensitive to complex environmental changes and outputting feature vectors It is directly passed to the subsequent analysis module to support more accurate early warning and response mechanisms for fires, thereby completing the construction of dynamic multi-factor temperature characteristics and providing basic data with adaptability and stability for subsequent analysis.

[0061] Step 2: Perform nonlinear temperature trend analysis. After completing the multi-factor temperature feature identification, the output feature vector As input, nonlinear temperature trend analysis is further performed. This step uses nonlinear analysis methods to identify complex patterns of temperature changes in order to accurately capture abnormal temperature trends in the early stages of a fire and reduce the risk of false alarms and missed alarms. The specific steps are as follows:

[0062] In nonlinear trend analysis, we first construct a time series matrix of multi-factor inputs of the time series, expressed as: , where n represents the length of the time window, such as, is the eigenvector of the previous i moments, by introducing the time series matrix , determine the historical change patterns of temperature and multiple factors, provide stable input data for nonlinear trend identification, and construct a time series matrix , it is possible to observe the historical changes of temperature and multiple factors from multiple time steps, identify the patterns and trends of temperature changes, rather than being limited to data at a single moment, thereby improving the accuracy of nonlinear analysis;

[0063] Long short-term memory network (LSTM) is used for nonlinear trend modeling to transform the time series matrix Input into the long short-term memory network (LSTM), taking advantage of its ability to memorize long-term and short-term data features, to identify the nonlinear trend of temperature change. The calculation process of the LSTM network is: ,in, is the hidden state at the current moment, representing the potential characteristics of the temperature trend; , are the input weight matrix and the hidden state weight matrix respectively, b is the bias term, is the activation function (such as the hyperbolic tangent function);

[0064] LSTM selectively retains or updates historical features through gating mechanisms (forget gate, input gate, and output gate) to avoid short-term fluctuations in nonlinear trends affecting trend prediction. Hidden states will be used as key features for trend analysis and further for identifying and predicting temperature trends. The LSTM network captures the historical characteristics of temperature and multiple factors through a recursive layer structure and can retain long-term characteristic patterns in nonlinear trends, thereby accurately identifying complex trends in temperature changes.

[0065] In the hidden state Finally, the hidden state is mapped to the predicted temperature trend through the fully connected layer to obtain the predicted temperature value : ,in, and are the weight and bias parameters of the fully connected layer respectively;

[0066] Will With safety threshold Compare and judge whether the temperature trend exceeds the warning range. When the set safety range is exceeded, an early warning signal is automatically triggered to remind users of possible fire risks;

[0067] By predicting future temperature trends, the temperature monitoring function can issue timely warnings before abnormal temperature changes exceed the safe range, ensuring the real-time and accuracy of temperature monitoring, and effectively improving the sensitivity of early fire monitoring. Through nonlinear temperature trend analysis, the temperature anomaly characteristics in the early stage of a fire can be extracted from the complex changes in temperature and environmental factors, thereby achieving more sensitive fire warnings.

[0068] Step 3: Construct a multi-time scale intelligent early warning trigger mechanism. After completing the nonlinear temperature trend analysis, the future temperature change trend is effectively predicted. In order to further improve the accuracy and flexibility of the early warning, a multi-time scale intelligent early warning trigger mechanism is introduced. Through trend analysis of different time scales and hierarchical alarm strategies, it adapts to the diversity of temperature changes and provides refined early warnings at different change speeds in the early stage of a fire. The specific steps are as follows:

[0069] Define multiple time scale windows and divide the time scale into short-term, medium-term and long-term windows to capture temperature change trends at different frequencies. Set the length of each time window to:

[0070] Short-term window : Overwrite the past Temperature changes over time periods, used to identify rapid temperature rise trends within a short period of time;

[0071] Mid-term window : Overwrite the past Time period, focusing on the medium-term volatility of temperature, capturing trends that rise slowly or fluctuate less frequently;

[0072] Long-term window : Covering longer Time period, tracking long-term temperature change patterns, suitable for identifying slow trend warming;

[0073] The length of each time window , , It needs to be set according to the temperature change characteristics of the actual application scenario to ensure that different time scales can effectively cover short-, medium- and long-term temperature changes;

[0074] Temperature trend analysis is performed in each time window. For each time window, different trend analysis methods are applied to extract the temperature trend characteristics at a specific time scale:

[0075] To conduct short-term trend analysis, fast Fourier transform (FFT) is used to identify the high-frequency fluctuation trend of temperature and extract the frequency characteristics of temperature change in the short term. The short-term temperature change rate at time t is defined as: , where F represents the Fourier transform operation and the input is a short-term time window Temperature series within , by analyzing high-frequency components, an alarm can be generated when the temperature rises rapidly;

[0076] Perform medium-term trend analysis and use the weighted moving average (WMA) to calculate the medium-term window Smooth temperature change trend, eliminate the interference of high-frequency noise, and the temperature value of the medium-term trend for: ,in, Assigned to past temperature data The weight of makes the more recent temperature data have a greater impact on the trend. This method can capture the medium-frequency trend of temperature and is suitable for slowly changing temperature rise scenarios.

[0077] Long-term trend analysis, applying exponential smoothing methods to identify long-term temperature trends, defining long-term trends for: ,in, To smooth the coefficients and control the impact of new data on the trend, long-term trend analysis can track the slow rise pattern of temperature and is suitable for identifying potential fire risks;

[0078] Temperature trend results at different time scales , set up a hierarchical early warning trigger mechanism:

[0079] Level 1 warning (short-term rapid temperature rise): When the short-term trend Above short-term threshold When the temperature rises rapidly, a level 1 warning is triggered to remind users of the risk of rapid temperature rise in a short period of time, which is suitable for potential emergency fire situations;

[0080] Level 2 warning (mid-term trend of warming): When the mid-term trend Exceeding the mid-term threshold When the temperature rises steadily, a second-level warning is triggered to monitor the stable rising trend of the temperature, which may be a potential fire hazard;

[0081] Level 3 warning (long-term slow temperature rise): When the long-term trend Exceeding long-term threshold When the temperature rises slowly over a long period of time, the third-level warning is triggered, indicating that attention should be paid to possible abnormal temperature changes;

[0082] Further dynamically adjust the threshold to adapt to environmental changes, dynamically adjust the alarm threshold of each time scale according to the real-time changes of the environment, and automatically increase or decrease the threshold if certain factors in the environment (such as humidity, airflow, etc.) change significantly , , , thereby improving the adaptability of early warning, the dynamic threshold adjustment strategy can be defined as: ,in, is the initial threshold of each time scale, is the adjustment factor, It is a function of environmental factors, indicating the degree of environmental fluctuation. By dynamically adjusting the warning threshold, the system can more flexibly respond to temperature change trends under different environmental conditions.

[0083] Through the multi-time scale intelligent early warning trigger mechanism, it is possible to identify trend changes in temperature at different time scales and provide a hierarchical early warning strategy, thereby effectively responding to temperature changes at different rates in the early stages of a fire.

[0084] Step 4: Adaptive learning and dynamic optimization of early warning strategies. After the multi-time scale intelligent early warning trigger mechanism is completed, the early warning strategy can achieve hierarchical response to temperature trends at different time scales. In order to further improve the adaptability and response accuracy of temperature monitoring in actual environments, this step introduces adaptive learning and dynamic optimization mechanisms. By analyzing environmental factors and the actual situation of early warning triggers in real time, the early warning strategy is dynamically optimized so that it can continuously adjust the early warning parameters and thresholds as the environment and temperature characteristics change, thereby providing more accurate fire early warning services. The specific steps are as follows:

[0085] The actual response after each warning is triggered (such as the effectiveness of the warning, the record of false alarms or missed alarms) is recorded, and a feedback data set F is constructed. The feedback data set contains the temperature characteristics, environmental conditions and warning effectiveness information when the warning is triggered, which is defined as: ,in, is the temperature characteristic value at different time scales, The status of environmental factors, including real-time environmental data such as humidity and airflow, Represents the warning feedback result (for example, 0 represents a false alarm, 1 represents a valid alarm); by continuously accumulating feedback data sets, the actual performance of the current warning strategy under different environmental conditions is obtained;

[0086] Based on the analysis results of the feedback data set, the early warning strategy is adaptively adjusted using the reinforcement learning algorithm to optimize the alarm thresholds and strategies in different environments. The early warning strategy optimization is regarded as a reward-penalty problem. The Q-learning algorithm of reinforcement learning is applied to iteratively update the alarm strategy to improve the effectiveness of the early warning strategy. The goal of each strategy adjustment is to minimize the number of false alarms and missed alarms. The update formula is defined as: , where s represents the current state (a combination of temperature characteristics and environmental state), a represents the alarm strategy (such as the increase or decrease of the alarm threshold), and r is the reward value of the current strategy, which is set according to the accuracy of the early warning feedback (for example, the reward value for a valid alarm is positive, and the reward value for a false alarm is negative). is the learning rate, controlling the update step size, is the discount factor, which indicates the impact of future rewards on the current strategy;

[0087] Through repeated iterations, the optimal alarm strategy is gradually learned to ensure that the alarm sensitivity and accuracy under different environmental conditions remain optimal;

[0088] After completing the adaptive adjustment of the early warning strategy, the alarm thresholds of different time scales are dynamically updated to adapt to environmental changes. According to the characteristics of environmental factors in the feedback data set F, the optimal strategy obtained by Q-learning is mapped to the alarm threshold adjustment formula: , where k represents the time scale, which can be short-term, medium-term or long-term. is the threshold adjustment value in the current environment, generated by the Q-learning algorithm, and the new threshold Applied to the next warning trigger to adapt to the impact of environmental changes in real time;

[0089] After each alarm is triggered, the new feedback information is incorporated into the feedback data set, and the adaptive adjustment of the early warning strategy is re-executed to form a closed-loop mechanism of continuous optimization. The optimized strategy is fed back to the temperature monitoring in each round, and the alarm strategy and threshold are updated, so that the early warning strategy is continuously optimized with the dynamic changes of the environment and temperature mode. Through the closed-loop optimization mechanism, the accuracy and adaptability of the early warning strategy can be maintained in long-term operation.

[0090] The multi-time scale early warning mechanism transmits the real-time feedback of the alarm trigger to the adaptive learning module. Through the reinforcement learning algorithm, step 4 dynamically adjusts the alarm thresholds and strategies at each time scale according to the feedback information of the actual early warning. It can automatically optimize the alarm strategy according to the changes in the environment and temperature characteristics, and realize the continuous improvement and dynamic adaptation of the strategy.

[0091] The adaptive learning mechanism enables temperature monitoring to continuously optimize warning parameters and strategies based on historical feedback, reduce false alarms and missed alarms, and thus achieve more accurate temperature monitoring and fire warnings in actual use. While dynamically adjusting the alarm threshold, it adapts to different environmental change characteristics, effectively improving the sensitivity and accuracy of the warning in different scenarios.

[0092] After completing adaptive learning and strategy optimization, a closed-loop feedback mechanism is formed. The data of each alarm feedback will be transmitted back to the relevant temperature detection, and the alarm strategy will be continuously updated. The warning parameters and thresholds will be more in line with environmental requirements and temperature trends in each round of feedback and adjustment, forming a complete cycle of self-learning and intelligent improvement.

[0093] After receiving temperature monitoring information and alarm data, the palm-held fire-fighting equipment will perform data analysis and status updates, and display them to the user through an interface. The equipment usually displays the data visually along a timeline to help users understand temperature trends and changes in real time. At the same time, for alarm signals, the equipment will trigger various forms of alarm prompts, such as vibration, sound, or push notifications, so that users can immediately detect temperature abnormalities and take action. When the user confirms the alarm or performs equipment calibration operations on the device, the palm-held fire-fighting equipment will send control instructions back to the monitoring temperature system via the wireless network. The monitoring temperature system will perform corresponding operations based on the received instructions, such as resetting the alarm status, adjusting the alarm threshold, or updating the data collection frequency, to enhance the interactivity and flexibility of temperature monitoring.

[0094] It should be noted that the threshold information in this embodiment is pre-set by professionals and will not be explained in detail here. Some parameter English letters in the embodiments have the same situation, but different meanings are explained when used, which will not be explained one by one here.

[0095] The present invention synchronously collects environmental factor data through temperature monitoring points, calculates the correlation between each factor and temperature after denoising, extracts multi-factor features and generates output feature vectors for temperature prediction, then constructs the feature vector as a time series matrix and inputs it into a long short-term memory network to extract temperature trend features, predict future temperature trends, and compare with safety thresholds to determine whether to trigger an early warning. At the same time, the temperature trends at short-term, medium-term and long-term scales are analyzed, and fast Fourier transform, weighted moving average and exponential smoothing methods are used to identify temperature changes of different frequencies. A three-level early warning mechanism is set to dynamically adjust the alarm threshold to adapt to environmental changes. Finally, by recording early warning feedback data and using the Q-learning algorithm to continuously optimize the threshold and alarm strategy, an adaptive learning closed-loop mechanism is formed, so that the temperature collection monitoring alarm for palm firefighting has efficient fire early warning capabilities.

[0096] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0097] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0098] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0099] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0100] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0101] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A temperature collection, monitoring and alarm method for palm-held firefighting, characterized in that: The steps include: Environmental factor data is collected synchronously at the temperature monitoring point, and the correlation between each environmental factor data and temperature is calculated after denoising, the multi-factor characteristics of temperature change are determined, and the output feature vector for temperature prediction is determined based on the factor fluctuations; The output feature vector is constructed as a time series matrix and input into the long short-term memory network to extract the temperature trend features to predict the future temperature trend and compare it with the safety threshold to determine whether to trigger an early warning. Analyze the temperature trends in the short, medium and long term, use fast Fourier transform, weighted moving average and exponential smoothing to identify temperature trends of different frequencies, set up a three-level early warning mechanism and dynamically adjust the threshold; Record early warning feedback data, use Q-learning algorithm to optimize alarm thresholds, and gradually adjust alarm strategies; The temperature trends at short-term, medium-term and long-term scales were analyzed, and the temperature trends at different frequencies were identified using fast Fourier transform, weighted moving average and exponential smoothing methods. The specific steps are as follows: Define multiple time scale windows and divide the time scale into three windows: short-term, medium-term and long-term; Conduct temperature trend analysis in each time scale window, and apply different trend analysis methods to each time scale window; For short-term trend analysis, fast Fourier transform is used to identify the high-frequency fluctuation trend of temperature and extract the frequency characteristics of temperature changes in the short term; Perform medium-term trend analysis and use the weighted moving average method to calculate the smoothed temperature trend in the medium-term window; Conduct long-term trend analysis and apply exponential smoothing methods to identify long-term temperature trends; A three-level early warning mechanism is set up to dynamically adjust the threshold, including the following steps: According to the temperature trend results at different time scales, a hierarchical early warning trigger mechanism is set up: When the short-term trend is higher than the short-term threshold, a level 1 warning is triggered to alert users to the risk of rapid temperature rise in a short period of time; When the medium-term trend exceeds the medium-term threshold, the second-level warning is triggered, which is used to monitor the stable rising trend of temperature and there is a fire hazard; When the long-term trend exceeds the long-term threshold, a level 3 warning is triggered, indicating that the temperature has been rising slowly for a long time, and users need to pay attention to abnormal long-term temperature changes.

2. A palm-mounted firefighting temperature collection, monitoring and alarm method according to claim 1, characterized in that: Environmental factor data is collected synchronously at the temperature monitoring point, and the correlation between each environmental factor data and temperature is calculated after denoising, the multi-factor characteristics of temperature change are determined, and the output feature vector for temperature prediction is determined according to the factor fluctuations. The specific steps are as follows: Set the data collection cycle to collect environmental factor data, including temperature, humidity, airflow intensity, equipment operation time, and environmental pressure, and store the environmental factor data in a multi-factor data set; Apply Fourier denoising transform to remove high-frequency noise components in the data; The mutual information analysis method is used to calculate the correlation between each factor in the environmental factor data and temperature; Construct a multi-factor nonlinear mapping model, use a polynomial kernel function to construct a temperature-multi-factor relationship model for the selected multi-factor feature vector, and output the predicted temperature feature value; For each factor, a weight that is adjusted over time is set as a dynamic weight function; The predicted temperature eigenvalue is combined with the dynamic weight of the corresponding factor to obtain the output eigenvector.

3. The temperature collection, monitoring and alarm method for palm firefighting according to claim 2 is characterized in that: The output feature vector is constructed as a time series matrix and input into the long short-term memory network to extract the temperature trend features to predict the future temperature trend and compare it with the safety threshold to determine whether to trigger an early warning. The specific steps include: Construct a time series matrix to determine the historical variation patterns of temperature and multiple factors; Long short-term memory network is used for nonlinear trend modeling. The time series matrix is ​​input into the long short-term memory network to identify the nonlinear trend of temperature change. The long short-term memory network retains or updates historical features through a gating mechanism, and the hidden state of the output is used as a feature for trend analysis; After obtaining the hidden state, the hidden state is mapped to the predicted temperature trend through the fully connected layer to obtain the predicted temperature value; The predicted temperature value is compared with the safety threshold to determine whether the temperature trend exceeds the warning range. When the predicted temperature exceeds the set safety threshold, a warning signal is automatically triggered to remind users of the risk of fire.

4. The temperature collection, monitoring and alarm method for palm firefighting according to claim 3 is characterized in that: Record the early warning feedback data, use the Q-learning algorithm to optimize the alarm threshold, and gradually adjust the alarm strategy, including the following steps: Record the response after each warning is triggered and build a feedback data set; The feedback data set contains the temperature characteristics, environmental conditions, and warning effectiveness information when the warning is triggered; The Q-learning algorithm of reinforcement learning is applied to iteratively update the alarm strategy. The goal of strategy adjustment is to minimize the number of false alarms and missed alarms. The update formula is: , where s represents the combination of the current temperature feature and the environmental state, a represents the alarm strategy, and r is the reward value of the current strategy. is the learning rate, controlling the update step size, is the discount factor, which indicates the impact of the reward on the current strategy; After completing the adaptive adjustment of the early warning strategy, the alarm thresholds of different time scales are dynamically updated, and the strategy updated by Q-learning is mapped to the alarm threshold according to the characteristics of environmental factors in the feedback data set.

Images