Method for analyzing terminal position distribution based on public network base station data

Abstract

The application relates to the technical field of big data analysis, and provides a method for analyzing terminal position distribution based on public network base station data. The method comprises the following steps: obtaining air quality data of a target region based on public network base station data, calculating a pollution index mean value and fluctuation degree; constructing a pollution topology graph and a fluctuation feature graph according to the pollution index mean value and the fluctuation degree; obtaining weather change and people flow density prediction information and adjusting the pollution topology graph and the fluctuation feature graph, and outputting a corrected pollution distribution topology graph and a corrected index fluctuation feature graph; and optimizing a sensor layout scheme and adjusting a position according to the corrected graph. The application solves the technical problem that, in the prior art, the PM2.5 sensor layout position lacks dynamic and accurate layout, leading to insufficient air quality monitoring precision, and achieves the technical effect that the public network base station data is dynamically analyzed by using big data analysis technology, the sensor layout position is accurately optimized, and the air quality monitoring precision is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data analysis, in particular to a method for analyzing terminal position distribution based on public network base station data. BACKGROUND

[0002] With the acceleration of urbanization and the continuous expansion of industrial activities, air pollution problems have gradually emerged, among which PM2.5 pollution poses a significant threat to the health of residents. PM2.5 refers to particulate matter in the air with a diameter of less than or equal to 2.5 microns, also known as fine particulate matter, which is usually generated by motor vehicle exhaust, industrial emissions, coal-fired power generation, construction dust, biomass burning, etc. The current common PM2.5 monitoring method is mostly based on fixed-point sensor layout, collecting air pollution data through limited monitoring points. This approach is difficult to adapt dynamically when facing changes in urban population density, weather factors, and sudden pollution incidents, resulting in insufficient precision and real-time monitoring data. In addition, existing technologies often overlook the potential application value of public network base station data in terminal position analysis and crowd activity monitoring, making the sensor layout scheme lack the ability to respond dynamically to changes in terminal position based on big data analysis, and unable to mine the correlation between crowd migration patterns and pollution exposure risk through massive communication signal data, further reducing the spatial coverage and accuracy of air quality monitoring. Therefore, there is an urgent need for a method that can combine public network base station data, analyze terminal position changes in real time through big data analysis technology, and accurately guide the dynamic layout of PM2.5 sensors to improve the precision and reliability of air quality monitoring. SUMMARY

[0003] The present application provides a method for analyzing terminal position distribution based on public network base station data, which aims to solve the technical problem of insufficient precision in air quality monitoring due to the lack of dynamic and accurate layout of PM2.5 sensor placement in existing technologies.

[0004] The present application provides a method for analyzing terminal position distribution based on public network base station data, which includes: based on public network base station data, obtaining air quality monitoring data of a target area in a historical time zone, and calculating a plurality of air pollution index means and a plurality of index fluctuation degrees; fitting and constructing a pollution distribution topology graph and an index fluctuation feature graph according to the plurality of air pollution index means and the plurality of index fluctuation degrees; obtaining predicted weather change information and predicted population density distribution of the target area in a preset future period, adjusting the pollution distribution topology graph and the index fluctuation feature graph, and outputting a corrected pollution distribution topology graph and a corrected index fluctuation feature graph; optimizing the sensor placement scheme according to the corrected pollution distribution topology graph and the corrected index fluctuation feature graph, and obtaining an adaptive placement scheme to adjust the sensor placement position in the preset future period.

[0005] One or more technical solutions provided in the application have at least the following technical effects or advantages:

[0006] The method for analyzing terminal position distribution based on public network base station data collects historical air quality data of a target area and calculates the average value and fluctuation degree of the air pollution index. Then, a distribution topology and fluctuation feature map of the pollution situation are established based on the data. Next, the original map is corrected and optimized by obtaining weather change prediction and human flow density prediction information of a future period. Finally, a more reasonable PM2.5 sensor layout scheme is determined based on the optimized map, and the sensor position is adjusted in the predetermined future period to improve the air quality monitoring accuracy.

[0007] The above description is only a summary of the technical solutions of the application. In order to more clearly understand the technical means of the application, the application can be implemented according to the content of the specification, and in order to make the above and other purposes, features and advantages of the application more obvious and easy to understand, the following specific embodiments of the application are described. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0009] Figure 1 The flowchart of the method for analyzing terminal position distribution based on public network base station data in an embodiment.

[0010] Figure 2 The flowchart of calculating the average value and index fluctuation degree of the air pollution index of the method for analyzing terminal position distribution based on public network base station data in an embodiment. DETAILED DESCRIPTION

[0011] The embodiments of the application provide a method for analyzing terminal position distribution based on public network base station data, which solves the technical problem of lack of dynamic and accurate layout of PM2.5 sensor layout in the prior art, resulting in insufficient air quality monitoring accuracy.

[0012] The technical solutions in the embodiments of the application will be described clearly and completely in conjunction with the drawings in the embodiments of the application. Obviously, the described embodiments are only some of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0013] It is to be understood that the terms "including", "comprising", "having" and "with" are open-ended and are used as equivalents that do not exclude additional steps or elements, for example, process, method, system, product, or server including a series of steps or units not expressly listed, but which can include other steps or modules that are not expressly listed or inherent to such process, method, product, or device.

[0014] As shown in the embodiments, the present application provides a method for analyzing terminal position distribution based on public network base station data, which comprises: Figure 1 As shown in the embodiments, the present application provides a method for analyzing terminal position distribution based on public network base station data, which comprises:

[0015] Based on the public network base station data, the air quality monitoring data of the target area in the historical time zone is obtained, and a plurality of air pollution index mean values and a plurality of index fluctuation degrees are calculated.

[0016] In the embodiments of the present application, first, the historical air quality monitoring data in the target area is obtained from the public network base station data, and a plurality of time series of air pollution index records are formed. Subsequently, the pollution index time series corresponding to each position point is subjected to mathematical statistical processing, and the numerical mean value is calculated, which represents the average pollution level of the point in the period. Then, the fluctuation analysis is performed on each time series, and the index fluctuation degree is obtained, which is used to reflect the stability and change range of the pollution index in the region. After all the time series are calculated, a plurality of pollution index mean values and a plurality of index fluctuation degrees are formed, which are used for subsequent pollution feature spectrum construction.

[0017] Further, as shown in the embodiments, the present application provides a method for analyzing terminal position distribution based on public network base station data, which comprises: Figure 2 As shown in the embodiments, the present application provides a method for analyzing terminal position distribution based on public network base station data, which comprises:

[0018] Based on the public network base station data, the air quality monitoring data of the target area in the historical time zone is obtained, and a plurality of air pollution index mean values and a plurality of index fluctuation degrees are calculated, which comprises:

[0019] Preferably, first, the air quality monitoring data of the target area in the historical time zone is located and extracted from the public network base station data according to the geographic identification, and the historical air quality monitoring data is collected based on a plurality of PM2.5 sensors arranged at different positions in the region through an initial sensor arrangement scheme, and each air pollution index sequence in the historical air quality monitoring data corresponds to a specific sensor point. Subsequently, mean calculation is performed on each pollution index sequence to obtain the air pollution index mean of each pollution index sequence, which is used to reflect the average pollution level of each point in the selected time period. In addition, fluctuation analysis is also performed on each air pollution index sequence, the standard deviation is calculated, and the standard deviation is divided by the mean to obtain the corresponding index fluctuation degree, which is used to describe the stability or change intensity of the pollution data at the point. Finally, the air pollution index mean and the index fluctuation degree of all monitoring points are output, providing basic input for constructing the pollution atlas and optimizing the arrangement scheme.

[0020] According to the plurality of air pollution index means and the plurality of index fluctuation degrees, a pollution distribution topology map and an index fluctuation characteristic map are fitted and constructed.

[0021] In one embodiment, first, the geographic position coordinates of each PM2.5 sensor in the initial arrangement scheme are matched with the corresponding air pollution index mean and index fluctuation degree to construct an air pollution index mean array and a fluctuation degree array, and each data in the array corresponds to a respective sensor point. Subsequently, a pre-trained pollution distribution fitter and a fluctuation characteristic fitter are called, the air pollution index mean array is input into the pollution distribution fitter, and the index fluctuation degree array is input into the fluctuation characteristic fitter. In the fitter, based on the existing pollution evolution law and geographical space characteristics, the pollution degree and fluctuation of each point are continuously spatially interpolated and topologically modeled to output a pollution distribution topology map and an index fluctuation characteristic map covering the entire target region, wherein the pollution distribution topology map is used to describe the diffusion structure of the pollution concentration in space, and the index fluctuation characteristic map is used to reflect the sensitivity and uncertainty of the pollution change in the region. The two atlas can comprehensively reflect the spatial distribution trend of the regional pollution and the fluctuation risk distribution, providing quantifiable spatial basis for subsequent dynamic adjustment of factors such as weather and human flow, thereby significantly improving the scientificity and pertinence of the sensor arrangement scheme.

[0022] Further, the application provides that according to the plurality of air pollution index means and the plurality of index fluctuation degrees, a pollution distribution topology map and an index fluctuation characteristic map are fitted and constructed, comprising:

[0023] According to the position coordinates of the plurality of sensors in the initial sensor arrangement, the plurality of air pollution index averages and the plurality of index fluctuation degrees are respectively arranged to construct an air pollution index average array and an index fluctuation degree array; a pollution distribution fitter and an index fluctuation fitter are pre-trained; and the pollution distribution fitter and the index fluctuation fitter are used to perform fitting according to the air pollution index average array and the index fluctuation degree array to output a pollution distribution topology map and an index fluctuation feature map.

[0024] Optionally, first, according to the spatial position coordinates of each PM2.5 sensor in the initial sensor arrangement, all sensors are numbered in the order of geographical distribution or grid number, and the air pollution index average corresponding to each sensor is sequentially arranged to form an air pollution index average array. Similarly, the index fluctuation degrees corresponding to each sensor point are sequentially arranged to form an index fluctuation degree array, which together reflect the pollution level and fluctuation characteristics in space. Subsequently, the pollution distribution fitter and the index fluctuation fitter pre-trained based on historical environmental monitoring records are called. These two fitters usually use a generative adversarial network (GAN) or other deep learning structure and can learn and restore the pollution evolution pattern and fluctuation trend in continuous space from the input pollution data array. Then, the constructed air pollution index average array is input into the pollution distribution fitter to perform fitting interpolation of the pollution level in the spatial dimension, and a pollution distribution topology map in the continuous spatial range is output. At the same time, the index fluctuation degree array is input into the index fluctuation fitter to extract the spatial pattern of pollution fluctuation at each location, and an index fluctuation feature map is output. The obtained pollution distribution topology map and index fluctuation feature map can present the spatial aggregation area of pollution concentration and its fluctuation sensitive area from a macroscopic perspective, effectively supporting subsequent dynamic environmental factor correction and sensor layout optimization, and improving the response capability and decision accuracy of the overall monitoring to pollution evolution.

[0025] Further, the application provides a pre-trained pollution distribution fitter and index fluctuation fitter, comprising:

[0026] According to the historical environmental monitoring records of the target area, a sample air pollution index average array set, a sample pollution distribution topology map set, a sample index fluctuation degree array set, and a sample index fluctuation feature map set are obtained; the generator and the discriminator of the generative adversarial network are trained using the sample air pollution index average array set and the sample pollution distribution topology map set until the network converges to obtain a pollution distribution fitter; and the generator and the discriminator of the generative adversarial network are trained using the sample index fluctuation degree array set and the sample index fluctuation feature map set until the network converges to obtain an index fluctuation fitter.

[0027] Optionally, first, the environmental monitoring records of the target area in multiple historical periods are collected, and PM2.5 sensor layout data corresponding to each historical period is extracted. For each period, according to the pollution index monitoring sequence of each sensor point, the pollution index mean and index fluctuation are calculated respectively, and arranged in order of sensor location coordinates to construct a sample air pollution index mean array and a sample index fluctuation array. At the same time, according to the spatial pollution distribution and fluctuation characteristics of each historical period, a corresponding sample pollution distribution topology graph and sample index fluctuation feature graph are generated. Through the above steps, multiple historical sample pairs are formed, which respectively constitute a sample air pollution index mean array set - a sample pollution distribution topology graph set, and a sample index fluctuation array set - a sample index fluctuation feature graph set. Subsequently, the generator and discriminator in the generative adversarial network (GAN) structure are used to train the above two sample sets. Taking the training of the pollution distribution fitter as an example, first, the generative adversarial network architecture is constructed, wherein the generator (Generator) is a mapping network based on a convolutional neural network (CNN) for receiving the input sample air pollution index mean array and outputting the predicted pollution distribution topology graph. The discriminator is also a CNN structure for receiving the input topology graph and judging whether it comes from the real pollution distribution topology graph set. In the training process, each set of sample air pollution index mean array is input into the generator to generate a simulated pollution topology graph, and the generated graph and the real pollution distribution topology graph of the same sample are input into the discriminator. The discriminator attempts to distinguish whether the input graph is a real graph or a generated graph, and its output is a binary classification result (real / fake). The generator adjusts its parameters according to the feedback of the discriminator so that the generated graph becomes more and more realistic and as close as possible to the real pollution graph. In the training, the cross-entropy loss function is used as the optimization target of the discriminator and the generator, and the structural similarity index (SSIM) or mean square error (MSE) can also be introduced as an auxiliary constraint to improve the topology structure fidelity. The training process adopts an alternating optimization strategy, i.e. in each training round, the generator is fixed to train the discriminator, and then the discriminator is fixed to train the generator. The whole adversarial training process lasts for multiple rounds until the discrimination accuracy of the discriminator cannot be improved and the pollution topology graph output by the generator is highly consistent with the real graph, indicating that the network has converged. At this time, the generator weight is frozen and used as the pollution distribution fitter for subsequent actual pollution graph prediction. Similarly, another GAN network is trained with the sample index fluctuation array as input and the sample index fluctuation feature graph as output target to optimize its generation ability and discrimination accuracy, and finally an index fluctuation fitter is obtained. In summary, the above training process realizes the modeling and generalization ability of pollution spatial distribution and fluctuation trend through deep learning of historical data, so that the pollution topology and fluctuation graph of the continuous area can be quickly and accurately output under new data input, thereby providing data basis for dynamic optimization of sensor layout. Figure 1 and sent to the discriminator. The discriminator attempts to distinguish whether the input graph is a real graph or a generated graph, and its output is a binary classification result (real / fake). The generator adjusts its parameters according to the feedback of the discriminator so that the generated graph becomes more and more realistic and as close as possible to the real pollution graph. In the training, the cross-entropy loss function is used as the optimization target of the discriminator and the generator, and the structural similarity index (SSIM) or mean square error (MSE) can also be introduced as an auxiliary constraint to improve the topology structure fidelity. The training process adopts an alternating optimization strategy, i.e. in each training round, the generator is fixed to train the discriminator, and then the discriminator is fixed to train the generator. The whole adversarial training process lasts for multiple rounds until the discrimination accuracy of the discriminator cannot be improved and the pollution topology graph output by the generator is highly consistent with the real graph, indicating that the network has converged. At this time, the generator weight is frozen and used as the pollution distribution fitter for subsequent actual pollution graph prediction. Similarly, another GAN network is trained with the sample index fluctuation array as input and the sample index fluctuation feature graph as output target to optimize its generation ability and discrimination accuracy, and finally an index fluctuation fitter is obtained. In summary, the above training process realizes the modeling and generalization ability of pollution spatial distribution and fluctuation trend through deep learning of historical data, so that the pollution topology and fluctuation graph of the continuous area can be quickly and accurately output under new data input, thereby providing data basis for dynamic optimization of sensor layout.

[0028] obtain prediction weather change information and prediction people flow density distribution of the target area in a preset future period, adjust the pollution distribution topology graph and the index fluctuation feature graph, and output a corrected pollution distribution topology graph and a corrected index fluctuation feature graph.

[0029] In one embodiment, first, based on public network base station data, weather prediction information of a target area in a preset future period is obtained, which can reflect the air flow state and pollution diffusion trend in the future period. At the same time, a prediction model is constructed by combining historical people flow activity trajectory data and weather change characteristics, and the people flow density distribution in the future period is calculated to represent the pollution disturbance that may be caused by people activity. Subsequently, the prediction weather change information and the prediction people flow density distribution are input into the feature correction branch that has been trained in advance as input features, and the original pollution distribution topology graph and the index fluctuation feature graph are corrected. In the correction process, the feature correction branch will correct the spatial offset according to the influence of future meteorological factors on the migration path of pollutants, and adjust the original graph by considering the pollution intensification or disturbance fluctuation that may be caused by high-density people flow areas, so as to output a corrected pollution distribution topology graph reflecting the real pollution diffusion situation in the future and a corrected index fluctuation feature graph reflecting the uncertainty level of future pollution. In summary, through the above correction steps, the original pollution distribution topology graph can adapt to future environmental factors, and the foresight and accuracy of subsequent sensor layout decision can be improved.

[0030] Further, the application provides obtaining prediction weather change information and prediction people flow density distribution of a target area in a preset future period, adjusting the pollution distribution topology graph and the index fluctuation feature graph, and outputting a corrected pollution distribution topology graph and a corrected index fluctuation feature graph, comprising:

[0031] obtain prediction weather change information of a target area in a preset future period based on public network base station data, wherein the prediction weather change information at least includes prediction temperature, prediction wind direction and prediction wind speed; obtain a historical people flow density distribution sequence of the target area based on public network base station data, and obtain prediction people flow density distribution by combining the prediction weather change information; adjust the pollution distribution topology graph and the index fluctuation feature graph according to the prediction weather change information and the prediction people flow density distribution, and output a corrected pollution distribution topology graph and a corrected index fluctuation feature graph.

[0032] Preferably, first, based on the public network base station data, the meteorological prediction information of the target area in the preset future period is extracted, including the predicted temperature, the predicted wind direction and the predicted wind speed, etc. The information is usually provided by a third-party meteorological service platform, and through the public network base station associated with the terminal geographic position, the high-resolution meteorological data matching at the regional level can be realized, and a set of weather change information with time sequence characteristics in the future period is formed. Subsequently, based on the historical positioning data of the public network base station, the number of terminal activities in the target area in multiple historical periods is counted to construct a human flow density distribution sequence, and then combined with the weather condition data corresponding to the historical periods, a training sample set is formed, and a long short-term memory network (LSTM) or other time sequence prediction model is used for modeling. Then, by inputting the predicted weather change information, the model can calculate the spatio-temporal distribution trend of the human flow in the future period, so as to output the predicted human flow density distribution, which accurately reflects the human flow and aggregation characteristics in the future area. Then, the predicted weather change information and the predicted human flow density distribution are taken as input features, and the existing pollution distribution topology map and the index fluctuation feature map are input into the pre-trained feature correction branch, the influence of wind speed and wind direction on the pollution migration path is simulated, the pollution distribution topology map is corrected by spatial offset, diffusion or convergence, the area where the pollution fluctuation is likely to occur in the future is identified by analyzing the trend of human flow density change, and the index fluctuation map is enhanced or smoothed accordingly, so as to output the corrected pollution distribution topology map and the corrected index fluctuation feature map. The corrected pollution distribution topology map reflects the real spatial distribution trend of the pollutants under the joint action of future meteorological conditions and human flow activities, and the corrected index fluctuation feature map reveals the potential fluctuation intensity of the pollution data in each area. This adjustment process provides more forward-looking and accurate decision support for subsequent sensor deployment.

[0033] Further, the application provides a method for obtaining the historical human flow density distribution sequence of the target area based on the public network base station data, and analyzing the predicted human flow density distribution based on the predicted weather change information, comprising:

[0034] Based on the public network base station data, a sample human flow density distribution sequence set and a sample weather change information set are collected, and a historical human flow density distribution after a historical period is obtained as a sample predicted human flow density distribution, to obtain a sample predicted human flow density distribution set; the sample human flow density distribution sequence set, the sample weather change information set and the sample predicted human flow density distribution set are used as training data to train a long short-term memory network to convergence, to obtain a human flow density prediction model; and the human flow density prediction model is used to analyze the predicted human flow density distribution according to the historical human flow density distribution sequence and the predicted weather change information.

[0035] Optionally, first, based on public network base station data, a plurality of representative historical time periods are selected, and the location activity information of terminal devices is extracted in the target area, the change of the number of terminals in each geographic grid per unit time is counted, a sample human flow density distribution sequence set is constructed, the sequence data has a clear time sequence, and the dynamic distribution characteristics of the crowd in different time periods are reflected. At the same time, the corresponding historical weather change information is extracted in these time periods to form a sample weather change information set, including temperature, wind speed, wind direction and other multi-dimensional meteorological factors. Further, human flow density distribution data is continuously collected in the continuous time period after the corresponding time period of each sample sequence as a sample predicted human flow density distribution set, which represents the model prediction target value. Subsequently, the above three types of data are combined to form a complete training sample (input: human flow density sequence and weather information, output: human flow density in the future period), and a long short-term memory network is used as the human flow density prediction model structure. The input layer of the human flow density prediction model accepts the historical human flow density sequence and weather information, processes the time dependence relationship through multiple LSTM units, and the output layer predicts the human flow density distribution at each time in the future. In the training process, mean square error (MSE) is used as the loss function, and the back propagation algorithm is used to continuously optimize the model weight until the validation set error is stable and the model converges. Finally, the trained human flow density prediction model is applied to the actual prediction task, that is, the current obtained historical human flow density distribution sequence and predicted weather change information are input into the human flow density prediction model for forward calculation, and the predicted human flow density distribution in the preset future period is output. This predicted human flow density distribution can show the high-risk area of crowd gathering in space and reflect the human flow change trend in time, providing key dynamic basis for pollution map correction and sensor layout.

[0036] Further, the application provides adjusting the pollution distribution topology map and the exponential fluctuation characteristic map according to the predicted weather change information and the predicted human flow density distribution, including:

[0037] Based on the public network base station data, a sample weather change information set, a sample people flow density distribution set, a sample pollution distribution topology graph set and a sample index fluctuation feature graph set are collected, and a historical pollution distribution topology graph and a historical index fluctuation feature graph after a historical period are obtained as a sample correction pollution distribution topology graph and a sample correction index fluctuation feature graph, to obtain a sample correction pollution distribution topology graph set and a sample correction index fluctuation feature graph set; taking the sample weather change information set, the sample people flow density distribution set and the sample pollution distribution topology graph set as inputs, taking the sample correction pollution distribution topology graph set as a supervised training to generate a generative adversarial network to convergence, to obtain a first feature correction branch; taking the sample weather change information set, the sample people flow density distribution set and the sample index fluctuation feature graph set as inputs, taking the sample correction index fluctuation feature graph set as a supervised training to generate a generative adversarial network to convergence, to obtain a second feature correction branch; the first feature correction branch and the second feature correction branch are used to adjust the pollution distribution topology graph and the index fluctuation feature graph according to the predicted weather change information and the predicted people flow density distribution.

[0038] Optionally, first, based on public network base station data, sample weather change information sets, sample human flow density distribution sets, sample pollution distribution topology graph sets, and sample index fluctuation characteristic graph sets are collected and sorted in multiple historical time windows, and actual pollution change results of each group of sample time periods after a certain number of hours or days are obtained, that is, corresponding historical pollution distribution topology graphs and historical index fluctuation characteristic graphs, as correction targets, respectively constituting sample correction pollution distribution topology graph sets and sample correction index fluctuation characteristic graph sets, for supervised learning. Subsequently, two generative adversarial networks (GANs) are constructed for feature correction, wherein the first group of networks is used to train the first feature correction branch for pollution distribution graph correction, the generator in the first group of networks takes the sample weather change information, sample human flow density distribution, and sample pollution topology graph as input, and outputs the fitted correction pollution distribution topology graph, and the discriminator is used to judge the difference between the generated correction pollution distribution topology graph and the true sample correction pollution distribution topology graph. The specific training steps are similar to the foregoing and will not be described here. The second group of networks is used to train the second feature correction branch for index fluctuation graph correction. Similarly, the generator takes the sample weather change information, sample human flow density distribution, and sample index fluctuation characteristic graph as input, and outputs the fitted correction index fluctuation characteristic graph, and the discriminator identifies the authenticity of the generated characteristic graph. Finally, the predicted future weather change information and the predicted human flow density distribution graph are input as input, and the current pollution distribution topology graph and the fluctuation characteristic graph are input into the first feature correction branch and the second feature correction branch, respectively, for forward inference, to obtain the correction pollution distribution topology graph and the correction index fluctuation characteristic graph, which together reflect the pollution intensity distribution and its fluctuation characteristics in the future period, thereby providing a more realistic pollution environment perception basis for subsequent sensor deployment optimization, improving the forward-looking and rationality of subsequent sensor deployment, and ensuring high precision of air quality monitoring.

[0039] According to the correction pollution distribution topology graph and the correction index fluctuation characteristic graph, a sensor deployment scheme optimization is performed to obtain an adaptive deployment scheme, and sensor deployment position adjustment is performed in the preset future period.

[0040] In one embodiment, after obtaining the corrected pollution distribution topology map and the corrected index fluctuation feature map, the corrected index fluctuation feature map and the corrected index fluctuation feature map are used to perform adaptability analysis on the randomly generated sensor layout scheme, to evaluate the coverage and response capability of the layout scheme in the target area, and to comprehensively form a layout adaptability score. In each iteration, the schemes with higher adaptability are retained and adjusted and combined, and the sensor position distribution is gradually optimized, until a preset optimization number of rounds or adaptability convergence is reached. After the iteration ends, the layout scheme with the highest adaptability score is taken as the adaptive layout scheme, and in a preset future period, the actual installation position of the sensor is dynamically adjusted according to the adaptive layout scheme, to ensure that the sensor can be preferentially deployed in areas with higher pollution risk and greater volatility, thereby improving the coverage and real-time response capability of the entire monitoring, effectively avoiding the redundant deployment of sensors in low-value areas, and improving the resource utilization efficiency and the accuracy of air quality monitoring.

[0041] Further, the application provides a sensor layout scheme optimization method based on the corrected pollution distribution topology map and the corrected index fluctuation feature map, to obtain an adaptive layout scheme, comprising:

[0042] Randomly laying out a first sensor layout scheme based on a preset number of sensor layouts in the target area; performing layout adaptability analysis based on the corrected pollution distribution topology map, the corrected index fluctuation feature map, and the first sensor layout scheme, to output a first scheme adaptability; iteratively generating the layout scheme and iteratively analyzing the scheme adaptability, until a preset number of iterations is reached, and outputting a sensor layout scheme with the maximum scheme adaptability as the adaptive layout scheme.

[0043] Preferably, first, a preset sensor deployment quantity N and a spatial boundary condition of a target area (such as a latitude and longitude range, an installable area constraint) are set, and a certain number of initial deployment individuals (i.e., an initial population) are randomly generated in the target area based on this, each individual representing a sensor deployment scheme, denoted as a first sensor deployment scheme, including a spatial coordinate group of N sensors. Subsequently, for each deployment individual, a deployment fitness analysis is performed in combination with the previously generated correction pollution distribution topology map and correction index fluctuation feature map, and the matching degree of the deployment scheme to the pollution high-value area and the fluctuation strong area is calculated, so as to obtain the fitness score of each deployment individual, i.e., the first scheme fitness. Then, based on the first scheme fitness, a roulette selection or tournament selection method is used to select individuals with high fitness as parents, and a coordinate level exchange operation (such as position point crossover, local deployment segment replacement, etc.) is performed on the offspring to generate a new sensor deployment scheme, denoted as a second sensor deployment scheme. In addition, some sensor positions in the individual are disturbed (such as slightly moving the coordinates) with a certain probability to enhance the local exploration ability, and the individual with the lowest fitness is replaced with the newly generated scheme to maintain a constant population size. The above iteration process is repeated until a preset maximum iteration number or population fitness convergence is reached. After the iteration ends, the deployment scheme with the highest fitness is selected from all individuals in the iteration process, and is taken as the final adaptive deployment scheme. This adaptive deployment scheme effectively avoids local optimal traps and can globally optimize the sensor position distribution under complex pollution environments and deployment constraints, improve the coverage rate of high pollution areas and the response density of high fluctuation areas, and significantly enhance the overall effectiveness and accuracy of monitoring.

[0044] Further, the application provides a deployment fitness analysis based on the correction pollution distribution topology map, the correction index fluctuation feature map, and the first sensor deployment scheme, and outputs the first scheme fitness, including:

[0045] The correction pollution distribution topology map is divided according to a preset area size, and a plurality of high pollution areas are analyzed and determined, wherein the pollution index identification value of the high pollution area is greater than a preset index threshold. The monitoring coverage range of the PM2.5 sensor is obtained, the first sensor deployment scheme is rendered, and it is judged whether the first sensor deployment scheme covers the plurality of high pollution areas comprehensively. If not, the first sensor deployment scheme is discarded. If yes, the sensor matching degree in the plurality of high pollution areas is calculated based on the correction pollution distribution topology map and the correction index fluctuation feature map, and the first scheme fitness is obtained by weighting.

[0046] Optionally, according to the set area size (for example, 50 meters x 50 meters per grid), the corrected pollution distribution topology is spatially gridded to form a plurality of independent spatial units. Subsequently, the pollution index identification value in each grid unit is counted, which is the pollution index average of the area where the corresponding grid is located, representing the pollution level of the area. By comparing the pollution index identification value of each spatial unit with the preset pollution index threshold value, if the pollution index identification value of a certain area is greater than the threshold value, it is marked as a high pollution area, and these high pollution areas will be used for subsequent sensor coverage matching analysis. Then, the monitoring coverage radius (for example, 100 meters) of the PM2.5 sensor is obtained, and the coordinate positions of all sensors in the first sensor layout scheme are spatially rendered, that is, the circular range covered by each sensor is calculated, and these coverage areas are compared with the high pollution areas to determine whether the first sensor layout scheme completely covers all high pollution areas. If any high pollution area is not covered, it is considered that the scheme does not meet the basic layout requirements, and the scheme is directly discarded and does not enter the adaptation degree calculation link. On the contrary, if the first sensor layout scheme meets the coverage requirement, it enters the adaptation degree calculation stage, that is, the sensor matching degree of each high pollution area is calculated according to the corrected pollution distribution topology, the corrected index fluctuation feature map and the number of sensors in the high pollution area, and then the first scheme adaptation degree of the first sensor layout scheme is calculated by weighting. This first scheme adaptation degree can effectively improve the configuration efficiency of sensor resources in pollution hotspots and change sensitive areas, and enhance the scientificity and practicality of the overall monitoring.

[0047] Further, the application provides a method for calculating a sensor matching degree in each of the plurality of high pollution areas based on the corrected pollution distribution topology and the corrected index fluctuation feature map, and obtaining a first scheme adaptation degree by weighting, comprising:

[0048] Based on the corrected pollution distribution topology and the corrected index fluctuation feature map, a plurality of pollution index identification values and a plurality of index fluctuation degree identification values of the plurality of high pollution areas are obtained, and after the plurality of pollution index identification values and the plurality of index fluctuation degree identification values are dimensionless processed, a plurality of pollution characteristic values are obtained by weighting. Based on the first sensor layout scheme, a plurality of sensor numbers of the plurality of high pollution areas are obtained. A plurality of sensor matching degrees are calculated according to the plurality of pollution characteristic values and the plurality of sensor numbers, wherein the sensor matching degree is the ratio of the sensor number to the pollution characteristic value. The plurality of high pollution areas are set with pollution scale weights according to the plurality of pollution characteristic values, and the plurality of sensor matching degrees are weighted and calculated to output the first scheme adaptation degree, wherein the pollution scale weight and the pollution characteristic value are positively correlated.

[0049] Optionally, based on the constructed correction pollution distribution topology and correction index fluctuation feature map, for each identified high pollution area, the pollution index identification value and the index fluctuation degree identification value are extracted, and the two identification values are the average air pollution index value and the index fluctuation degree average value of the corresponding area respectively. Subsequently, the pollution index identification value and the index fluctuation degree identification value of all high pollution areas are respectively processed dimensionless, for example, by minimum-maximum normalization or Z-score standardization, so that the numerical values are in the same scale, so as to subsequent weighted processing. After normalization, the two standardized indicators of each region are linearly weighted according to the set weight, and the pollution feature value of the region is obtained. Then, based on the current first sensor layout scheme, the number of sensors falling into each high pollution area is counted, and the number of sensors in each high pollution area is calculated by ratio with the pollution feature value, and the sensor matching degree of each region is obtained. The higher the matching degree, the more sufficient the sensor configuration of the region, and the lower the matching degree, the less the monitoring resources of the region. Then, in order to comprehensively evaluate the adaptation effect of the whole layout scheme, it is necessary to weight and summarize the matching degrees of various regions, and for this purpose, according to the pollution feature value of each region, the pollution scale weight of each region is set, and the higher the weight value, the greater the pollution risk of the region, which should be covered first. Under normal circumstances, the pollution weight is the ratio of the pollution feature value of the region to the total pollution feature value. Finally, according to the set pollution weight, the weighted calculation of the multiple sensor matching degrees is carried out, and the first scheme adaptation degree of the first sensor layout scheme is obtained. This first scheme adaptation degree can scientifically measure the monitoring matching degree of the layout scheme to the high-risk pollution area, provide accurate basis for optimizing the sensor resource configuration, and significantly improve the monitoring efficiency and scheduling rationality of the sensor network.

[0050] In summary, the embodiments of the present application have at least the following technical effects:

[0051] The embodiment of the application first acquires air quality monitoring data of a target region in a historical time zone based on public network base station data, and calculates a plurality of air pollution index means and a plurality of index fluctuation degrees; then, a pollution distribution topology graph and an index fluctuation feature graph are fitted and constructed according to the plurality of air pollution index means and the plurality of index fluctuation degrees; subsequently, predicted weather change information and predicted human flow density distribution of the target region in a preset future period are acquired, the pollution distribution topology graph and the index fluctuation feature graph are adjusted, and a corrected pollution distribution topology graph and a corrected index fluctuation feature graph are output; finally, sensor layout scheme optimization is performed according to the corrected pollution distribution topology graph and the corrected index fluctuation feature graph, and an adaptive layout scheme is obtained to adjust the sensor layout position in the preset future period. These technical effects collectively solve the technical problem of lack of dynamic and accurate layout of PM2.5 sensor layout positions in the prior art, which leads to insufficient air quality monitoring precision, and achieve the technical effects of dynamically analyzing the terminal position of public network base station data by using big data analysis technology, accurately optimizing the sensor layout position, and improving the air quality monitoring precision.

[0052] It should be noted that the above sequence of the embodiments of the application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of the present application. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0053] The above only describes the preferred embodiments of the application and does not limit the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

[0054] The present application and the drawings are only exemplary descriptions of the application, and are considered to cover any and all modifications, changes, combinations or equivalents within the scope of the application. Obviously, those skilled in the art can make various modifications and changes to the application without departing from the scope of the application. Thus, if these modifications and changes of the application belong to the scope of the application and its equivalents, the application intends to include these modifications and changes.

Claims

1. A method for analyzing terminal location distribution based on public network base station data, characterized in that, The methods include: Based on public network base station data, air quality monitoring data of the target area within the historical time zone is obtained, and several average air pollution indices and several index fluctuations are calculated. A pollution distribution topology map and an index fluctuation feature map are constructed by fitting the mean values ​​of several air pollution indices and the volatility of several indices. Obtain predicted weather changes and predicted population density distribution for the target area within a preset future time period, adjust the pollution distribution topology map and index fluctuation feature map, and output the corrected pollution distribution topology map and corrected index fluctuation feature map; Based on the corrected pollution distribution topology map and the correction index fluctuation characteristic map, the sensor deployment scheme is optimized to obtain an adaptive deployment scheme, and the sensor deployment positions are adjusted within the preset future time period. The adapted deployment schemes include: The first sensor deployment scheme is generated by randomly deploying sensors within the target area based on a preset number of sensors. Based on the corrected pollution distribution topology map, the correction index fluctuation characteristic map and the first sensor deployment scheme, a deployment adaptability analysis is performed, and the adaptability of the first scheme is output. The deployment scheme is iteratively generated and the first scheme adaptability is iteratively analyzed until the preset number of iterations is reached. The sensor deployment scheme with the maximum scheme adaptability is output as the adapted deployment scheme. The adaptability of the first solution output includes: The corrected pollution distribution topology map is divided according to the preset area size, and multiple high-pollution areas are analyzed and identified, wherein the pollution index label value of the high-pollution area is greater than the preset index threshold. Obtain the monitoring coverage of the PM2.5 sensor, render the first sensor deployment scheme, and determine whether the first sensor deployment scheme fully covers the multiple high-pollution areas. If not, discard the first sensor deployment scheme. If so, based on the corrected pollution distribution topology map and the correction index fluctuation feature map, the sensor matching degree in the multiple high-pollution areas is calculated respectively, and the first scheme fit degree is obtained by weighting.

2. The method for analyzing terminal location distribution based on public network base station data according to claim 1, characterized in that, Based on public network base station data, air quality monitoring data for the target area within the historical time zone is obtained, and several average air pollution index values ​​and several index fluctuations are calculated, including: Based on public network base station data, air quality monitoring data of the target area within the historical time zone is obtained. The air quality monitoring data is obtained based on the initial sensor deployment scheme and includes several air pollution index sequences. The initial sensor deployment scheme includes several PM2.5 sensors located at different locations. The mean values ​​of the air pollution indexes are calculated by performing mean calculations on the aforementioned air pollution index sequences to obtain the mean values ​​of the air pollution indexes. Volatility analysis is performed on the aforementioned air pollution index series to output several index volatility values, where the index volatility is the ratio of the index standard deviation to the index mean.

3. The method for analyzing terminal location distribution based on public network base station data according to claim 2, characterized in that, Based on the average values ​​of several air pollution indices and the volatility of several indices, a pollution distribution topology map and an index volatility characteristic map are constructed, including: Based on the position coordinates of several sensors in the initial sensor deployment scheme, the average values ​​of several air pollution indices and the fluctuation values ​​of several indices are arranged to construct an array of average air pollution indices and an array of fluctuation values. Pre-trained contamination distribution fitter and exponential fluctuation fitter; The pollution distribution fitter and the exponential fluctuation fitter are used to fit the air pollution index mean array and the exponential fluctuation array, and output the pollution distribution topology map and the exponential fluctuation feature map.

4. The method for analyzing terminal location distribution based on public network base station data according to claim 3, characterized in that, Pre-trained contamination distribution fitters and exponential fluctuation fitters include: Based on the historical environmental monitoring records of the target area, the following data were processed to obtain a sample air pollution index mean array set, a sample pollution distribution topology map set, a sample index volatility array set, and a sample index volatility characteristic map set. Using the sample air pollution index mean array set and sample pollution distribution topology map set, the generator and discriminator of the generative adversarial network are trained until the network converges, and a pollution distribution fitter is obtained. Using the sample exponential volatility array set and the sample exponential volatility feature map set, the generator and discriminator of the generative adversarial network are trained until the network converges, resulting in an exponential volatility fitter.

5. The method for analyzing terminal location distribution based on public network base station data according to claim 1, characterized in that, Obtain predicted weather changes and predicted population density distribution for a target area within a preset future time period; adjust the pollution distribution topology map and index fluctuation feature map; and output a corrected pollution distribution topology map and a corrected index fluctuation feature map, including: Based on public network base station data, obtain the predicted weather change information of the target area within a preset future time period, wherein the predicted weather change information includes at least the predicted temperature, predicted wind direction and predicted wind speed; Based on the historical population density distribution sequence of the target area obtained from public network base station data, the predicted population density distribution is obtained by combining the predicted weather change information. Based on the predicted weather change information and predicted population density distribution, the pollution distribution topology map and index fluctuation feature map are adjusted to output a corrected pollution distribution topology map and a corrected index fluctuation feature map.

6. The method for analyzing terminal location distribution based on public network base station data according to claim 5, characterized in that, Based on historical pedestrian density distribution sequences obtained from public network base station data for the target area, and combined with the predicted weather change information, a predicted pedestrian density distribution is obtained, including: Based on public network base station data, a sample pedestrian density distribution sequence set and a sample weather change information set are collected, and the historical pedestrian density distribution after historical time periods is obtained as the sample predicted pedestrian density distribution to obtain the sample predicted pedestrian density distribution set. The sample pedestrian density distribution sequence set, sample weather change information set, and sample predicted pedestrian density distribution set are used as training data to train a long short-term memory network until convergence, thus obtaining a pedestrian density prediction model. The predicted population density distribution is obtained by analyzing the historical population density distribution sequence and the predicted weather change information using the population density prediction model.

7. The method for analyzing terminal location distribution based on public network base station data according to claim 5, characterized in that, The pollution distribution topology map and index fluctuation characteristic map are adjusted based on the predicted weather change information and predicted population density distribution, including: Based on public network base station data, sample weather change information set, sample human flow density distribution set, sample pollution distribution topology map set, and sample index fluctuation feature map set are collected. Historical pollution distribution topology map and historical index fluctuation feature map after historical time period are obtained as sample corrected pollution distribution topology map and sample corrected index fluctuation feature map, resulting in sample corrected pollution distribution topology map set and sample corrected index fluctuation feature map set. Using the sample weather change information set, sample pedestrian density distribution set, and sample pollution distribution topology map set as inputs, and using the sample corrected pollution distribution topology map set as supervised training, a generative adversarial network is trained until convergence to obtain the first feature correction branch; Using the sample weather change information set, sample crowd density distribution set, and sample index fluctuation feature map set as inputs, and using the sample corrected index fluctuation feature map set as supervision, the generative adversarial network is trained until convergence to obtain the second feature correction branch. Using the first feature correction branch and the second feature correction branch, the pollution distribution topology map and the index fluctuation feature map are adjusted according to the predicted weather change information and the predicted population density distribution.

8. The method for analyzing terminal location distribution based on public network base station data according to claim 1, characterized in that, Based on the corrected pollution distribution topology map and the correction index fluctuation characteristic map, the sensor matching degree in the multiple high-pollution areas is calculated respectively, and the weighted average is used to obtain the first scheme fit degree, including: Based on the corrected pollution distribution topology map and the corrected index fluctuation feature map, multiple pollution index identifiers and multiple index fluctuation identifiers of the multiple high-pollution areas are obtained. After dimensionless processing of the multiple pollution index identifiers and multiple index fluctuation identifiers, multiple pollution feature values ​​are obtained by weighting. Based on the first sensor deployment scheme, the number of sensors in the multiple highly polluted areas is obtained; Multiple sensor matching degrees are calculated based on the multiple pollution characteristic values ​​and the multiple number of sensors, wherein the sensor matching degree is the ratio of the number of sensors to the pollution characteristic values; Based on the multiple pollution characteristic values, pollution scale weights are set for the multiple highly polluted areas, and the matching degree of the multiple sensors is weighted and calculated to output the first scheme fit degree, wherein the pollution scale weights and pollution characteristic values ​​are positively correlated.

Images