CN-122024443-A - Unhooking MEMS and PPG health AI early warning integrated system for power grid operation safety belt

CN122024443ACN 122024443 ACN122024443 ACN 122024443ACN-122024443-A

Abstract

The invention relates to the technical field of safety monitoring of power grid operation, in particular to a safety belt unhooking MEMS and PPG health AI early warning integrated system for power grid operation, wherein a safety belt unhooking MEMS monitoring unit acquires hook displacement and stress data in real time; the system comprises a PPG health evaluation unit, an AI early warning unit, an environment self-adaptive preprocessing layer correction data processing unit, a convolutional neural network, a characteristic map, a pre-warning signal, a physical safety and health state collaborative monitoring and a worker safety, wherein the PPG health evaluation unit synchronously acquires heart rate and blood oxygen physiological parameters, abnormal characteristics are extracted through a pattern recognition algorithm, a correlation is established through a dual-channel neural network, a support vector machine classifier divides five-level health risks and outputs an evaluation result vector, the convolutional neural network extracts characteristics through channel convolution and three-level pooling, a characteristic map is generated through fusion, unhooking risks and health abnormal probabilities are calculated, and when any probability exceeds standard, an early warning signal containing abnormal codes, positioning information and risk levels is triggered, so that the physical safety and health state collaborative monitoring is realized, the comprehensiveness and the accuracy of early warning are improved, and the safety of operators is ensured.

Inventors

Tuo Furong
PU MING
WANG ZHANRONG
XU HONGSHUN
QI BINBIN
PENG JIAQI
XU HONGYANG
YE HAIPING
MA SHUNQING
Hai Yulian
LI ZHEN

Assignees

国网青海省电力公司海东供电公司
国网青海省电力公司

Dates

Publication Date: 20260512
Application Date: 20260205

Claims (10)

1. The utility model provides a power grid operation safety belt unhook MEMS and healthy AI early warning integration system of PPG which characterized in that includes: The safety belt unhooking MEMS monitoring unit (1) collects displacement data and stress data of a safety belt hook connecting point in real time; The PPG health evaluation unit (2) receives the displacement data and the stress data, synchronously acquires heart rate and blood oxygen saturation physiological parameters acquired by the PPG sensor module, analyzes the variation trend of a fluctuation curve and the stress data of the displacement data through a pattern recognition algorithm, establishes a correlation model with the heart rate and the blood oxygen saturation physiological parameters respectively, and outputs an evaluation result vector containing risk level codes and risk probability intensity; The AI early warning unit (3) constructs a convolution neural network model comprising an input layer, a triple hidden layer and an output layer, the input layer respectively carries out time domain feature extraction on the displacement data and the stress data to form a displacement data vector and a stress data vector, the displacement data vector and the stress data vector and a health state assessment result vector form a three-dimensional input matrix, wherein the first hidden layer extracts spatial features of the three-dimensional input matrix through a preset convolution kernel, the second hidden layer is laminated with feature dimensions of the spatial features through a maximum pooling layer, the third hidden layer is based on the feature dimensions, a feature map is constructed through a full connection layer, and the output layer respectively calculates a safety belt unhook risk probability value and a health abnormal probability value according to the feature map, and triggers an early warning signal when any probability value exceeds a preset threshold.
2. The power grid operation safety belt unhooking MEMS and PPG health AI early warning integrated system according to claim 1 is characterized in that the fluctuation curve of displacement data and the change trend of stress data are analyzed through a pattern recognition algorithm, and the system specifically comprises the following steps: The method comprises the steps of carrying out nonlinear alignment on a fluctuation curve of displacement data and a standard operation displacement track of power grid operators, extracting displacement fluctuation characteristic quantity, carrying out probability cluster analysis on a change trend of stress data, identifying an abnormal fluctuation mode of the stress data, and establishing a joint analysis data set containing the displacement fluctuation characteristic quantity and the abnormal fluctuation mode.
3. The power grid operation safety belt unhooking MEMS and PPG health AI early warning integrated system is characterized in that based on a joint analysis data set, a state transition path of a displacement fluctuation feature quantity is decoded through a hidden Markov model, a displacement abnormal jump sequence is detected, meanwhile, a multi-scale component of an abnormal fluctuation mode is decomposed through a wavelet transformation technology, frequency domain energy distribution features of stress mutation points are extracted, and finally, the displacement abnormal jump sequence and the frequency domain energy distribution features are subjected to time synchronization association to generate coupling abnormal indexes of displacement data and stress data.
4. The grid operation safety belt unhooking MEMS and PPG health AI early warning integrated system according to claim 3 is characterized in that a coupling abnormality index is input into a two-channel neural network, wherein a first channel of the two-channel neural network learns a time sequence dependency relationship between a fluctuation curve and a heart rate through a long-short-period memory network, when continuous displacement abnormal jump is detected to be accompanied by heart rate rising slope exceeding a set range, a heart rate correlation factor is output, a second channel of the two-channel neural network is matched with spectrum similarity of frequency domain energy distribution characteristics and blood oxygen saturation physiological parameters through a convolution neural network, and when the spectrum similarity exceeds a preset spectrum similarity threshold, a blood oxygen correlation factor is output, and finally the heart rate correlation factor and the blood oxygen correlation factor are fused to generate a health correlation matrix.
5. The system for integrating unhook MEMS and PPG health AI early warning of a power grid operation safety belt according to claim 4, wherein the system is characterized in that a health association matrix is subjected to main component dimension reduction processing, health risk feature vectors are extracted, the health risk feature vectors are input into a support vector machine classifier, the support vector machine classifier calculates a health risk level decision boundary through a radial basis function kernel, five-level health risk states are divided according to distribution positions of the health risk feature vectors in a health risk level decision space, and an evaluation result vector comprising risk level codes and risk probability intensities is output.
6. The system for integrating unhook MEMS and PPG health AI warning of a power grid operation safety belt as set forth in claim 1, wherein an environmental adaptive preprocessing layer is added in front of an input layer of the convolutional neural network model, and the environmental adaptive preprocessing layer generates an environmental compensation coefficient by acquiring altitude, temperature and vibration intensity data in real time to adjust normalization parameters of displacement data and stress data.
7. The integrated system for unhook MEMS and PPG health AI warning of a power grid operation safety belt according to claim 6, wherein the time domain feature extraction extracts fluctuation extremum and change rate feature of displacement data in a short time window, track smoothness index in a long time window, peak-valley difference and first derivative zero crossing density of stress data, finally weighting and fusing the fluctuation extremum, change rate feature, track smoothness index, peak-valley difference and first derivative zero crossing density with environment compensation coefficient to form displacement data vector and stress data vector.
8. The grid operation safety belt unhooking MEMS and PPG health AI early warning integrated system according to claim 7 is characterized in that when the first hidden layer extracts the spatial characteristics of a three-dimensional input matrix through a preset convolution kernel, a multichannel self-adaptive convolution is adopted, the system comprises sliding a long-strip convolution kernel along a time dimension in a displacement data vector channel, capturing a displacement abnormal jump sequence in the displacement data vector, sliding a square convolution kernel along an amplitude dimension in a stress data vector channel, extracting the frequency domain energy distribution characteristics in the stress data vector, and focusing a health risk grade distribution mode in a health state evaluation result vector channel by using a dot convolution kernel; The maximum pooling layer of the second hidden layer adopts a three-level grid compression mechanism, wherein a first-level grid covers the whole working period to execute global maximum pooling, a long-term fluctuation track of a displacement data vector is captured, the second-level grid divides a typical working action period to execute regional maximum pooling, local mutation characteristics of a stress data vector are reserved, a third-level grid focuses on a risk grade conversion point of a health state assessment result vector, and integrity of health risk grade coding is maintained.
9. The grid operation safety belt unhooking MEMS and PPG health AI early warning integrated system of claim 8 is characterized in that a long-term fluctuation track, local mutation characteristics and health risk level codes output by a maximum pooling layer are input into a full-connection layer, a topological mapping relation between a displacement abnormal jump sequence and a health risk level is established through a graph convolution network, the heart rate correlation factor and the blood oxygen correlation factor are used as side weights, a gating circulation unit is used for fusing the local mutation characteristics with the coupling abnormal indexes to generate stress health interaction vectors, and the topological mapping relation and the stress health interaction vectors are subjected to weighted superposition to generate a characteristic map.
10. The grid operation safety belt unhooking MEMS and PPG health AI early warning integrated system according to claim 9 is characterized in that based on the characteristic map, the topological side weight of the displacement abnormal jump sequence and the health risk level is extracted, the topological side weight and the heart rate correlation factor are subjected to weighted aggregation to generate a displacement risk intensity coefficient, the stress health deviation degree is calculated by combining the stress health interaction vector and the coupling abnormal index and the health risk level decision boundary, the displacement risk intensity coefficient is finally input into a preset hyperbolic tangent function, a unhooking risk probability value is output, the stress health deviation degree is input into a preset logic function, and a health abnormal probability value is output; Triggering an early warning signal comprising the heart rate correlation factor anomaly code, the coupling anomaly index positioning information and the health risk level when the unhooking risk probability value exceeds the preset threshold or the health anomaly probability value exceeds the warning threshold of the health risk level decision boundary.

Description

Unhooking MEMS and PPG health AI early warning integrated system for power grid operation safety belt Technical Field The invention relates to the technical field of power grid operation safety monitoring, in particular to a power grid operation safety belt unhooking MEMS and PPG health AI early warning integrated system. Background The safety monitoring of power grid operation is an important technology, the safety monitoring is particularly applied to safety protection links of power grid high-altitude operation personnel, the core is that comprehensive risk early warning is achieved by associating safety belt states with personnel health data, the safety protection precision and comprehensive core requirements of the power grid high-altitude operation are adapted, when the power grid operation personnel high-altitude operation is carried out, the unhooking risk can be directly reflected by the displacement and stress variation of the safety belt hooks, meanwhile, the abnormal heart rate and blood oxygen physiological parameters can be caused by operation strength and environmental factors, due to the potential association of the two types of data, the abnormal displacement or the stress mutation can cause personnel physiological stress, the physiological abnormality can also influence the operation stability to further aggravate unhooking the unhooking risk, the cooperative influence of the two is not effectively integrated, so that the single-dimension monitoring is difficult to comprehensively prejudge potential safety hazards, and the critical risk early warning can be omitted, and in order to solve the technical problem, therefore, the power grid operation safety belt unhooking MEMS and PPG health AI early warning integrated system is provided. Disclosure of Invention The invention aims to provide an off-hook MEMS and PPG health AI early warning integrated system for a safety belt for power grid operation, so as to solve the problems in the background technology. In order to achieve the above purpose, the utility model provides a power grid operation safety belt unhooking MEMS and healthy AI early warning integrated system of PPG, includes: the safety belt unhooking MEMS monitoring unit acquires displacement data and stress data of a safety belt hook connecting point in real time; The PPG health evaluation unit receives the displacement data and the stress data, synchronously acquires heart rate and blood oxygen saturation physiological parameters acquired by the PPG sensor module, analyzes the variation trend of a fluctuation curve and the stress data of the displacement data through a pattern recognition algorithm, establishes a correlation model with the heart rate and the blood oxygen saturation physiological parameters respectively, and outputs an evaluation result vector containing risk level codes and risk probability intensity; The AI early warning unit builds a convolution neural network model comprising an input layer, a triple hiding layer and an output layer, wherein the input layer respectively carries out time domain feature extraction on the displacement data and the stress data to form a displacement data vector and a stress data vector, the displacement data vector and the stress data vector and a health state evaluation result vector form a three-dimensional input matrix, the first hiding layer extracts spatial features of the three-dimensional input matrix through a preset convolution kernel, the second hiding layer compresses feature dimensions of the spatial features through a maximum pooling layer, the third hiding layer builds a feature map through a full connection layer based on the feature dimensions, and the output layer respectively calculates a safety belt unhook risk probability value and a health abnormal probability value according to the feature map, and triggers an early warning signal when any probability value exceeds a preset threshold. Compared with the prior art, the invention has the beneficial effects that: According to the invention, displacement and stress data are acquired in real time through the safety belt unhooking MEMS monitoring unit, the PPG health assessment unit utilizes a pattern recognition algorithm to mine abnormal characteristics, a dual-channel neural network is combined to establish association with heart rate and blood oxygen, a support vector machine classifier is used for dividing five-level health risks, a standardized assessment result is output, an environment self-adaptive preprocessing layer is additionally arranged in the AI early warning unit, the influence of environment interference on the data is corrected, a characteristic map is constructed through the split-channel convolution, three-level pooling and characteristic fusion of the triple hidden layer, unhooking risk and health abnormality probability are calculated, when any probability value exceeds standard, early warning signals containing abnormal codes, positioning inform