A substation fire-fighting monitoring data collection method, device and system
By constructing weighting factors and mapping functions for strongly correlated sensors and dynamically adjusting the data completion method, the problem of missing data in substation fire monitoring data acquisition is solved, achieving data accuracy and completeness, and supporting efficient fire early warning and equipment status assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HENAN EPRI GAOKE GROUP CO LTD
- Filing Date
- 2026-04-28
- Publication Date
- 2026-07-24
AI Technical Summary
During the data acquisition process for substation fire monitoring, data loss or transmission interruption may occur due to electromagnetic environment, network fluctuations, and sensor failures. Existing methods for data filling have low accuracy and are prone to false alarms and missed alarms.
By analyzing the correlation between sensors and similar sensors, weighting factors and mapping functions for strongly correlated sensors are constructed. Combined with environmental periodic factors and data anomaly characteristics, the data imputation method is dynamically adjusted to ensure the accuracy and completeness of the data.
This improves the accuracy of data entry, ensures the authenticity and completeness of fire monitoring data, and provides high-quality data support for fire early warning and equipment status assessment.
Smart Images
Figure CN122448282A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of big data acquisition technology, specifically to a method, device, and system for acquiring fire monitoring data in substations. Background Technology
[0002] As a critical node in the power system, the fire safety of substations is of paramount importance. To ensure the safe operation of substations, a large number of sensors, such as temperature, smoke, combustible gas, and humidity sensors, are typically deployed to monitor fire-related parameters in real time. These sensors aggregate data through a communication network to a centralized fire data processing unit, generating massive amounts of monitoring data that provide a data foundation for subsequent fire early warning, equipment condition assessment, and fault diagnosis.
[0003] However, in practical applications, data loss or transmission interruptions inevitably occur during substation fire monitoring data acquisition due to complex electromagnetic environments, network fluctuations, momentary sensor malfunctions, or unstable communication links, resulting in incomplete data sequences. Existing methods for filling missing data, such as simple mean imputation, forward imputation, or interpolation using data from a single neighboring sensor, often lack accuracy. These methods ignore changes in equipment operating conditions (such as sudden load increases) and the periodicity of the data itself, easily leading to a "smoothing effect" that causes the interpolated data to deviate from the true trend. Especially at critical moments when anomalies such as fires occur, this deviation may cause the system to miss the optimal warning time or generate false alarms.
[0004] Therefore, how to provide a substation fire monitoring data acquisition method that can address the problem of missing data and accurately and reliably reproduce real monitoring data has become an urgent technical problem to be solved in this field. Summary of the Invention
[0005] To address the aforementioned technical problems, the purpose of this application is to provide a method, device, and system for acquiring fire monitoring data in substations. The specific technical solution adopted is as follows: In a first aspect, embodiments of this application provide a method for acquiring fire monitoring data in a substation, the method comprising the following steps: Real-time acquisition of monitoring data from various sensors at different equipment locations within the substation; When there are missing values in the monitoring data of each sensor, the correlation factor between each sensor and its corresponding sensors of the same type is obtained based on the correlation between the monitoring data of each sensor and its corresponding sensors of the same type within a preset time period before the missing value, and then the strongly correlated sensors of each sensor are obtained; based on the periodicity intensity of the data of each strongly correlated sensor, an environmental periodic factor is constructed, and combined with the correlation factor and the type of each sensor, a weighting factor of each strongly correlated sensor is constructed. Based on the data anomaly characteristics between each sensor and other sensors on the same device, a training sample set is selected, and then the mapping function between each sensor and the monitoring data of other sensors on the same device is obtained. Based on whether each sensor has a strongly correlated sensor and whether there are other sensors on its device, and in combination with the weighting factor and the mapping function, the data of each sensor at the missing time is acquired, and then the obtained data is filled into the missing time.
[0006] Preferably, the correlation factor between each sensor and its counterparts refers to the Pearson correlation coefficient between each sensor and its counterparts within a preset time period prior to the missing time.
[0007] Preferably, the strongly correlated sensors refer to sensors of the same type whose correlation factor with each sensor is greater than or equal to a preset correlation threshold.
[0008] Preferably, the method for constructing the environmental cycle factor is as follows: Statistically analyze the monitoring data sequences of each strongly correlated sensor within a preset time period before the missing time. The missing time of each sensor is recorded as the corresponding time in each historical day as the historical corresponding time. The monitoring data sequence of each strongly correlated sensor within a preset time period before the historical corresponding time in each historical day is statistically analyzed. The environmental cycle factor is negatively correlated with the DTW distance between the two monitoring data sequences mentioned above.
[0009] Preferably, the method for constructing the weighting factors of each strongly correlated sensor is as follows: When each sensor is an environmental parameter sensor, the weighting factor of each strongly correlated sensor is positively correlated with the correlation factor and the environmental periodic factor. When each sensor is a burst parameter sensor, the weight factor of each strongly correlated sensor is positively correlated with the correlation factor.
[0010] Preferably, the method for obtaining the training sample set is as follows: Two data points collected at the same time by each sensor and any other sensor on the same device are recorded as a data point pair; The mean Euclidean distance between each data point pair of each sensor and other sensors on the same device and its two adjacent data point pairs is calculated. All data point pairs whose normalized mean value is less than or equal to a preset interference threshold are denoted as the training sample set between each sensor and other sensors on the same device.
[0011] Preferably, the method for obtaining the mapping function is as follows: The training sample set is used as input for the least squares method, and a multinomial regression function is selected to output the mapping function between each sensor and other sensors on the same device.
[0012] Preferably, the specific process of acquiring data from each sensor at the missing time is as follows: When there are strongly correlated sensors among the various sensors, the data for each sensor at the missing time is calculated according to the following expression: In the formula, This represents the data from the q-th sensor at the missing time. Let q be the data collected by the c-th strongly correlated sensor of the q-th sensor at the missing time. Let be the weighting factor of the c-th strongly correlated sensor for the q-th sensor; Let q be the number of strongly correlated sensors with the q-th sensor; Otherwise, if other sensors are present in the device where each sensor is located, the data for each sensor at the missing time is calculated according to the following expression: In the formula, This represents the data from the q-th sensor at the missing time. This represents the total number of other sensors on the device containing the q-th sensor; The mapped data value is obtained by using the mapping function of the data value of the vth sensor among the other sensors on the device where the qth sensor is located at the missing time. Otherwise, calculate the data for each sensor at the missing time using the following expression: In the formula, This represents the data from the q-th sensor at the missing time. This represents the data from the previous sampling time of the q-th sensor at the missing time. The mean of the first-order differences of all monitoring data within the nearest neighbor window of the missing time of the q-th sensor, wherein the nearest neighbor window of the missing time refers to the time window within a preset duration before the missing time; This is a preset adjustment factor, and its value range is [0.7, 1].
[0013] Secondly, embodiments of this application provide a substation fire monitoring data acquisition device, which includes: a data acquisition module, a strongly correlated sensor discrimination module, a same-device data mapping module, and a data filling module.
[0014] The data acquisition module is used to collect monitoring data from various sensors at various equipment locations within the substation in real time. The strongly correlated sensor discrimination module is used to identify strongly correlated sensors based on the degree of correlation between each sensor and sensors of the same type, and to obtain the weighting factor of each strongly correlated sensor for each sensor in combination with the type of each sensor. The same-device data mapping module is used to obtain the mapping function based on the mapping relationship between each sensor and the non-abnormal monitoring data of other sensors on the same device; The data imputation module is used to impute missing data at missing times based on the weighting factor and the mapping function.
[0015] Thirdly, embodiments of this application also provide a substation fire monitoring data acquisition system, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the above-described substation fire monitoring data acquisition methods.
[0016] As can be seen from the above embodiments, the substation fire monitoring data acquisition method, device, and system provided in this application have at least the following beneficial effects: This application addresses the problem that traditional methods of data imputation based on smoothing can easily lead to imputed data deviating from the true trend of data changes, resulting in false alarms and missed alarms. First, it analyzes the correlation between each sensor and its counterparts before the data loss point to identify strongly correlated sensors. Based on the data periodicity and sensor type of these strongly correlated sensors, a weighting factor is constructed for each sensor, enabling the assessment of the importance of each strongly correlated sensor in interpolating its missing data. Next, by analyzing the mapping relationship between each sensor and other sensors on the same device for non-abnormal data points, a mapping function is constructed to map the missing data of each sensor, providing a reference for calculating missing data. Based on the presence of strongly correlated sensors and other sensors on the same device, and in conjunction with the weighting factor and mapping function, the data of each sensor at the time of the missing value is acquired according to different scenarios. This allows the calculation method for missing values to be dynamically adjusted according to the actual situation, making the calculated missing data more reflective of the true data state, improving the accuracy of missing value imputation, ensuring the integrity and authenticity of fire monitoring data, and providing high-quality data support for subsequent fire early warning and status assessment. Attached Figure Description
[0017] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 A flowchart illustrating the steps of a substation fire monitoring data acquisition method according to one embodiment of this application; Figure 2 This is a schematic diagram of the structure of a substation fire monitoring data acquisition device provided in one embodiment of this application. Detailed Implementation
[0019] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a substation fire monitoring data acquisition method, device, and system proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0020] Unless otherwise specified and limited, terms such as “comprising,” “including,” or any other variations thereof are intended to cover a non-exclusive inclusion, such that a circuit structure, article, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the article or device that includes said element. Furthermore, the term “and / or” as used herein includes any and all combinations of one or more of the associated listed items. All technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.
[0021] The following description, in conjunction with the accompanying drawings, details a specific scheme for a substation fire monitoring data acquisition method, device, and system provided in this application.
[0022] Please see Figure 1 The diagram illustrates a flowchart of a substation fire monitoring data acquisition method according to an embodiment of this application. The method includes the following steps: Step 1: Collect monitoring data from various sensors at each equipment location in the substation in real time.
[0023] The substation fire protection data central processing unit acquires various monitoring data from various equipment locations within the substation in real time. This monitoring data includes ambient temperature, humidity, combustible gas concentration, and smoke concentration. These data are collected by sensors deployed throughout the substation, such as temperature sensors, humidity sensors, combustible gas sensors, and smoke sensors. In this embodiment, temperature and humidity sensors are designated as environmental parameter sensors, while combustible gas and smoke sensors are designated as emergency parameter sensors. These sensors are deployed at various equipment locations within the substation, with different sensor types installed at each location. For example, for oil-filled equipment such as transformers, the focus is on monitoring temperature and combustible gas concentration; for cable linings, the focus is on monitoring smoke concentration and temperature; and for battery rooms, the focus is on monitoring combustible gas concentration, temperature, and humidity. Different types of sensors are installed at different equipment locations. The monitoring data collected by each sensor is transmitted to the fire protection data central processing unit in real time via a communication network. In this embodiment, smoke sensors, combustible gas sensors, humidity sensors, and temperature sensors collect data synchronously, with a sampling period of 5 seconds, meaning data is collected every 5 seconds.
[0024] To avoid the impact of dimensional inconsistencies on subsequent analysis and calculations, the minimum-maximum normalization method is used to normalize the monitoring data collected by each sensor to eliminate the influence of dimensions. The minimum-maximum normalization method is a well-known technique, and its specific process will not be elaborated further.
[0025] At this point, the initial collection of fire monitoring data for the substation can be completed.
[0026] Step 2: When there are missing values in the monitoring data of each sensor, based on the correlation between each sensor and its corresponding sensors of the same type within a preset time period before the missing value, obtain the correlation factor between each sensor and its corresponding sensors of the same type, and then obtain the strongly correlated sensors of each sensor; based on the periodicity intensity of the data of each strongly correlated sensor, construct an environmental periodic factor, and combine the correlation factor and the type of each sensor to construct the weight factor of each strongly correlated sensor of each sensor.
[0027] During the data acquisition process for fire monitoring in substations, temporary data gaps may occur due to network fluctuations or momentary sensor malfunctions. Traditional methods typically involve simply filling in the data with the mean or using forward normal values. However, because monitoring data can change abruptly under normal circumstances during a fire, using traditional methods to fill in the data can lead to errors, potentially causing the system to miss the optimal early warning window.
[0028] In order to fill in the missing values in the monitoring data collected by each sensor, it is first necessary to determine whether there are missing values in the data collected by each sensor. Taking the q-th sensor as an example, we monitor in real time whether there are missing values in the monitoring data collected by the sensor. If there is no corresponding monitoring data for the q-th sensor at the current sampling time, the current sampling time is recorded as the missing time of the q-th sensor.
[0029] Furthermore, data collected by sensors of the same type deployed in substations typically exhibit similar fluctuation trends. Therefore, the correlation between data collected by sensors of the same type at different locations within the same enclosed space can be analyzed, laying the foundation for subsequent interpolation of missing points based on data collected by sensors of the same type. Taking the q-th sensor as an example, when a missing value is detected in the q-th sensor, the time window within a preset duration (set to 1 hour in this embodiment) before the missing time of the q-th sensor is recorded as the nearest neighbor window of the missing time of the q-th sensor; the Pearson correlation coefficient between the monitoring data sequences of the q-th sensor and its s-th similar sensor within the nearest neighbor window of the missing time is calculated and recorded as the correlation factor between the q-th sensor and its s-th similar sensor; the method for obtaining the Pearson correlation coefficient is a well-known method and will not be elaborated here. During the calculation of the Pearson correlation coefficient, when the denominator is 0, a preset minimum positive number (0.01 in this embodiment) is added to the denominator as a parameter adjustment factor to prevent the calculation from crashing due to a denominator of 0.
[0030] This application sets a preset correlation threshold (0.6 in this embodiment); when the correlation factor between the q-th sensor and its s-th sensor of the same type is greater than or equal to the preset correlation threshold, the s-th sensor of the same type is recorded as a strongly correlated sensor of the q-th sensor.
[0031] Furthermore, when there are strongly correlated sensors among the various sensors, for certain types of strongly correlated sensors, such as temperature and humidity sensors, the changing trends of the monitoring data they collect usually exhibit periodicity with changes in day and night or equipment operating conditions. If the data collected by the strongly correlated sensors is periodic, then the more accurately the data from these strongly correlated sensors reflects the true changing trend of the data, the higher the accuracy when using the data from these strongly correlated sensors for interpolation.
[0032] Based on the above characteristics, taking the i-th strongly correlated sensor of the q-th sensor as an example, the corresponding time in the historical M (M is a preset number of historical days, which is 10 days in this embodiment) corresponding to the missing time of the q-th sensor is recorded as the historical corresponding time; the monitoring data sequence of the nearest window of the i-th strongly correlated sensor of the q-th sensor in each historical day is statistically analyzed; if the obtained monitoring data sequence is more similar to the monitoring data sequence of the i-th strongly correlated sensor in the nearest window of the missing time, the periodicity of the monitoring data of the strongly correlated sensor is stronger, and it is more valuable for reference when interpolating the missing data of the q-th sensor.
[0033] In a preferred embodiment, an environmental periodicity factor is constructed based on the data periodicity intensity of each strongly correlated sensor to characterize the data periodicity intensity of each strongly correlated sensor. The method for constructing the environmental periodicity factor is as follows: statistically analyze the monitoring data sequences of each strongly correlated sensor within a preset time period prior to the missing time; record the corresponding times of the missing time of each sensor in historical days as historical corresponding times, and statistically analyze the monitoring data sequences of each strongly correlated sensor within a preset time period prior to the historical corresponding times in historical days; the environmental periodicity factor is negatively correlated with the DTW distance between the two monitoring data sequences. This negative correlation means that the dependent variable decreases (increases) as the independent variable increases (decreases).
[0034] In this embodiment, the environmental periodic factor of the i-th strongly correlated sensor of the q-th sensor is denoted as... Its specific expression is: In the formula, is the environmental periodicity factor of the i-th strongly correlated sensor of the q-th sensor; M is the preset number of historical days, which is 10 in this embodiment. When the number of historical days is less than 10, the actual number of historical days is used for calculation. Let be the monitoring data sequence of the i-th strongly correlated sensor of the q-th sensor at the missing time window; It is the monitoring data sequence of the nearest neighbor window of the i-th strongly correlated sensor of the q-th sensor at the corresponding time in the history of the m-th day; This is the DTW distance function, used to calculate the DTW distance between input data. This is a preset constant used to prevent the denominator from being 0; in this embodiment, it is set to 1. As a normalization function, this embodiment adopts the minimum-maximum normalization method, and the minimum and maximum values are determined based on the corresponding parameter set of the corresponding historical time in the previous M days of the sensor.
[0035] The larger the environmental periodicity factor, the more periodic the data collected by the i-th strongly correlated sensor at the missing time in the historical corresponding time of the q-th sensor is. Therefore, the data of the i-th strongly correlated sensor at the missing time is more valuable for the q-th sensor.
[0036] It should be noted that the calculation of the environmental periodicity factor only applies to environmental parameter sensors such as temperature sensors and humidity sensors that are affected by the continuous periodic changes in the diurnal environment. The environmental periodicity factor is not calculated for sensors with sudden parameters such as smoke concentration and combustible gas concentration.
[0037] Furthermore, when the type of the q-th sensor is an environmental parameter sensor, the larger the correlation factor between the q-th sensor and its i-th strongly correlated sensor, and the greater the periodicity of the data collected by the i-th strongly correlated sensor, the greater the reference significance of the data value corresponding to the i-th strongly correlated sensor when interpolating the missing data of the q-th sensor.
[0038] Based on the above analysis, and considering the environmental periodic factors of each strongly correlated sensor, along with the correlation factors and the type of each sensor, weighting factors for each strongly correlated sensor are constructed to characterize their importance in data interpolation of missing data. The method for constructing these weighting factors is as follows: when each sensor is an environmental parameter sensor, the weighting factors of each strongly correlated sensor are positively correlated with both the correlation factors and the environmental periodic factors; when each sensor is a burst parameter sensor, the weighting factors of each strongly correlated sensor are positively correlated with the correlation factors. This positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases).
[0039] In this embodiment, the weight factor of the i-th strongly correlated sensor of the q-th sensor is denoted as... Its specific expression is: In the formula, Let be the weighting factor of the i-th strongly correlated sensor for the q-th sensor; Let be the environmental periodicity factor of the i-th strongly correlated sensor of the q-th sensor; Let be the correlation factor between the q-th sensor and its i-th strongly correlated sensor; For the normalization function, this embodiment uses the minimum-maximum normalization method. The minimum and maximum values are determined based on the corresponding parameter sets of all strongly correlated sensors of the q-th sensor. Specifically, when the q-th sensor has only one strongly correlated sensor, let... .
[0040] in, The purpose of adding 1 to the term and dividing by 2 is to normalize the relevant factors with values in the range [-1, 1] to [0, 1]; the resulting The larger the value, the more important the data from the i-th strongly correlated sensor of the q-th sensor is for filling in the missing data of the q-th sensor.
[0041] Step 3: Based on the data anomaly characteristics between each sensor and other sensors on the same device, select a training sample set, and then obtain the mapping function between the monitoring data of each sensor and other sensors on the same device.
[0042] Furthermore, within the entire substation, a certain sensor may lack strongly correlated sensors. Therefore, this sensor cannot use the data characteristics of its strongly correlated sensors for data interpolation. For such sensors, this application performs data interpolation by analyzing the data value mapping characteristics between each sensor and other sensors on the same equipment. Specifically, the logic is as follows: data collected by different types of sensors on the same equipment will change with the operation of the equipment. For example, when the operating load of a certain equipment increases, the temperature sensor deployed on that equipment will collect a higher temperature, while the humidity sensor will collect a lower humidity. Since the data collected by both the temperature and humidity sensors are affected by the operating state of the equipment, there is a certain mapping function relationship between the data collected by the temperature and humidity sensors.
[0043] Based on the above characteristics, taking the q-th sensor as an example, when there are no strongly correlated sensors for the q-th sensor within the substation, the two data points collected simultaneously by the q-th sensor and any other sensor on the same equipment are recorded as a single data point pair. To enhance the accuracy of the subsequent least squares fitting algorithm, it is necessary to first eliminate data point pairs between each sensor and other sensors on the same equipment that are affected by noise.
[0044] Specifically, the Euclidean distance between each data point pair between the q-th sensor and the h-th sensor on the same device, and between each pair of data points and their two adjacent data point pairs, is calculated. The average of the obtained Euclidean distances is recorded as the interference coefficient of each data point pair between the q-th sensor and the h-th sensor on the same device, which is used to characterize the degree of noise interference to each data point pair. The larger the interference coefficient, the greater the difference between the value of the data point pair and its adjacent data point pairs, indicating a greater degree of anomaly in the data point pair, and therefore, it should not be included in the calculation of the mapping function between the corresponding two sensors.
[0045] Calculate the normalized value of the interference coefficient of all data point pairs within the nearest neighbor window at the missing time between the q-th sensor and the h-th sensor on the same device; set a preset interference threshold (0.6 in this embodiment), and record the data point pairs whose normalized interference coefficient is less than or equal to the preset interference threshold as the training sample set between the q-th sensor and the h-th sensor on the same device; use the training sample set as the input of the least squares method, select a multinomial regression function, and output the mapping function between the q-th sensor and the h-th sensor on the same device. The least squares method is a well-known technique, and its specific process will not be elaborated further.
[0046] Step 4: Based on whether each sensor has a strongly correlated sensor and whether there are other sensors on its device, and in combination with the weighting factor and the mapping function, acquire the data of each sensor at the missing time, and then fill the missing time with the obtained data.
[0047] Furthermore, depending on whether each sensor has a strongly correlated sensor and whether there are other sensors in the device, the values of each sensor at the missing time are calculated for different cases. The specific process is as follows: When the q-th sensor has a strongly correlated sensor, the data of the q-th sensor at the missing time is calculated according to the following expression: In the formula, This represents the data from the q-th sensor at the missing time. Let q be the data collected by the c-th strongly correlated sensor of the q-th sensor at the missing time. Let be the weighting factor of the c-th strongly correlated sensor for the q-th sensor; Let q be the number of strongly correlated sensors of the q-th sensor.
[0048] Otherwise, if other sensors exist in the device containing the q-th sensor, the data of the q-th sensor at the missing time is calculated according to the following expression: In the formula, Let V represent the data of the q-th sensor at the missing time; V represents the total number of other sensors on the device containing the q-th sensor. The mapped data value is obtained by using the mapping function of the vth sensor among the other sensors on the device where the qth sensor is located at the missing time.
[0049] Otherwise, the monitoring data prior to the missing time of the q-th sensor is used for smooth filling. Specifically, the data of the q-th sensor at the missing time is calculated according to the following expression: In the formula, This represents the data from the q-th sensor at the missing time. This represents the data from the previous sampling time of the q-th sensor at the missing time. It is the mean of the first-order differences of all monitoring data within the nearest neighbor window at the missing time of the q-th sensor; The preset adjustment factor has a value range of [0.7, 1], and is set to 0.9 in this embodiment.
[0050] Similarly, when missing data is detected in each sensor, the data of each sensor at the time of the missing data is calculated using the method described above.
[0051] The above steps can be used to obtain data from each sensor at the missing time, fill in the missing data at the calculated time, and complete the filling of missing data in the substation fire monitoring data, thereby obtaining complete monitoring data from each sensor.
[0052] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of a substation fire monitoring data acquisition device provided in an embodiment of this application. In this embodiment, the terminal includes units used to execute the steps in an embodiment corresponding to a substation fire monitoring data acquisition method. See also... Figure 2 The fire monitoring data acquisition device includes: a data acquisition module, a strongly correlated sensor discrimination module, a data mapping module for the same equipment, and a data filling module.
[0053] The data acquisition module is used to collect monitoring data from various sensors at various equipment locations within the substation in real time. The strongly correlated sensor discrimination module is used to identify strongly correlated sensors based on the degree of correlation between each sensor and sensors of the same type, and to obtain the weighting factor of each strongly correlated sensor for each sensor in combination with the type of each sensor. The same-device data mapping module is used to obtain the mapping function based on the mapping relationship between each sensor and the non-abnormal monitoring data of other sensors on the same device; The data imputation module is used to impute missing data at missing times based on the weighting factor and the mapping function.
[0054] Based on the same inventive concept as the above method, this application also provides a substation fire monitoring data acquisition system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described substation fire monitoring data acquisition methods.
[0055] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
[0056] It should be noted that, unless otherwise specified and limited, terms such as “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a circuit structure, article, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such article or device. Without further limitations, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the article or device that includes said element. Furthermore, the term “and / or” as used herein includes any and all combinations of one or more of the associated listed items.
[0057] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.
[0058] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.
Claims
1. A method for acquiring fire monitoring data in a substation, characterized in that, The method includes the following steps: Real-time acquisition of monitoring data from various sensors at different equipment locations within the substation; When there are missing values in the monitoring data of each sensor, the correlation factor between each sensor and its corresponding sensors of the same type is obtained based on the correlation between the monitoring data of each sensor and its corresponding sensors of the same type within a preset time period before the missing value, and then the strongly correlated sensors of each sensor are obtained; based on the periodicity intensity of the data of each strongly correlated sensor, an environmental periodic factor is constructed, and combined with the correlation factor and the type of each sensor, a weighting factor of each strongly correlated sensor is constructed. Based on the data anomaly characteristics between each sensor and other sensors on the same device, a training sample set is selected, and then the mapping function between each sensor and the monitoring data of other sensors on the same device is obtained. Based on whether each sensor has a strongly correlated sensor and whether there are other sensors on its device, and in combination with the weighting factor and the mapping function, the data of each sensor at the missing time is acquired, and then the obtained data is filled into the missing time.
2. The substation fire monitoring data acquisition method as described in claim 1, characterized in that, The correlation factor between each sensor and its counterparts refers to the Pearson correlation coefficient between each sensor and its counterparts' monitoring data within a preset time period prior to the missing time.
3. The substation fire monitoring data acquisition method as described in claim 1, characterized in that, The strongly correlated sensors refer to the same type of sensors whose correlation factor with each sensor is greater than or equal to a preset correlation threshold.
4. The substation fire monitoring data acquisition method as described in claim 1, characterized in that, The method for constructing the environmental periodic factor is as follows: Statistically analyze the monitoring data sequences of each strongly correlated sensor within a preset time period before the missing time. The missing time of each sensor is recorded as the corresponding time in each historical day as the historical corresponding time. The monitoring data sequence of each strongly correlated sensor within a preset time period before the historical corresponding time in each historical day is statistically analyzed. The environmental cycle factor is negatively correlated with the DTW distance between the two monitoring data sequences mentioned above.
5. The substation fire monitoring data acquisition method as described in claim 1, characterized in that, The method for constructing the weighting factors of each strongly correlated sensor is as follows: When each sensor is an environmental parameter sensor, the weighting factor of each strongly correlated sensor is positively correlated with the correlation factor and the environmental periodic factor. When each sensor is a burst parameter sensor, the weight factor of each strongly correlated sensor is positively correlated with the correlation factor.
6. The substation fire monitoring data acquisition method as described in claim 1, characterized in that, The method for obtaining the training sample set is as follows: Two data points collected at the same time by each sensor and any other sensor on the same device are recorded as a data point pair; The mean Euclidean distance between each data point pair of each sensor and other sensors on the same device and its two adjacent data point pairs is calculated. All data point pairs whose normalized mean value is less than or equal to a preset interference threshold are denoted as the training sample set between each sensor and other sensors on the same device.
7. The substation fire monitoring data acquisition method as described in claim 1, characterized in that, The method for obtaining the mapping function is as follows: The training sample set is used as input for the least squares method, and a multinomial regression function is selected to output the mapping function between each sensor and other sensors on the same device.
8. The substation fire monitoring data acquisition method as described in claim 1, characterized in that, The specific process for acquiring data from each sensor at the missing time is as follows: When there are strongly correlated sensors among the various sensors, the data for each sensor at the missing time is calculated according to the following expression: In the formula, This represents the data from the q-th sensor at the missing time. Let q be the data collected by the c-th strongly correlated sensor of the q-th sensor at the missing time. Let be the weighting factor of the c-th strongly correlated sensor for the q-th sensor; Let q be the number of strongly correlated sensors with the q-th sensor; Otherwise, if other sensors are present in the device where each sensor is located, the data for each sensor at the missing time is calculated according to the following expression: In the formula, This represents the data from the q-th sensor at the missing time. This represents the total number of other sensors on the device containing the q-th sensor; The mapped data value is obtained by using the mapping function of the data value of the vth sensor among the other sensors on the device where the qth sensor is located at the missing time. Otherwise, calculate the data for each sensor at the missing time using the following expression: In the formula, This represents the data from the q-th sensor at the missing time. This represents the data from the previous sampling time of the q-th sensor at the missing time. The mean of the first-order differences of all monitoring data within the nearest neighbor window of the missing time of the q-th sensor, wherein the nearest neighbor window of the missing time refers to the time window within a preset duration before the missing time; This is a preset adjustment factor, and its value range is [0.7, 1].
9. A substation fire monitoring data acquisition device, characterized in that, The substation fire monitoring data acquisition method as described in any one of claims 1-8, wherein the fire monitoring data acquisition device comprises: The data acquisition module is used to collect monitoring data from various sensors at various equipment locations within the substation in real time. The strongly correlated sensor discrimination module is used to identify strongly correlated sensors based on the degree of correlation between each sensor and sensors of the same type, and to obtain the weighting factor of each strongly correlated sensor for each sensor in combination with the type of each sensor. The same-device data mapping module is used to obtain the mapping function based on the mapping relationship between each sensor and the non-abnormal monitoring data of other sensors on the same device; The data imputation module is used to impute missing data at missing times based on the weighting factor and the mapping function.
10. A substation fire monitoring data acquisition system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements a substation fire monitoring data acquisition method as described in any one of claims 1-8.