Water conservancy intelligent internet-of-things sensing method and system

Abstract

The invention relates to the technical field of hydrological data processing, in particular to a water conservancy intelligent internet-of-things sensing method and system, and the method comprises the following steps: combining statistical data into a feature set based on hydrological monitoring data including flow, water level and rainfall information, judging whether the data accords with normal distribution, and generating a data distribution feature vector set. According to the method, the change trend of the data is captured through multi-level data distribution characteristic analysis, and the monitoring precision of the data exception is enhanced. In the prediction process, new data and weight adjustment are gradually fused, so that the model can adapt to environmental changes in real time, and the hydrological state in a future short period is accurately predicted. Besides, abnormal data can be identified in real time based on analysis of quantiles, potential problems can be quickly found, and state division of different levels can be carried out according to abnormal degrees. In sampling scheduling, the priority of data acquisition is adjusted according to a probability model of state transition, so that monitoring resources are efficiently configured.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrological data processing, and in particular to a water conservancy intelligent Internet of Things perception method and system. Background Art

[0002] The field of hydrological data processing involves the collection, analysis, and processing of data related to hydrological phenomena such as water bodies, precipitation, rivers, lakes, and groundwater. Sensors, remote sensing technology, and data analysis algorithms are used to monitor and assess the dynamic changes and quality of water resources. Real-time collection and analysis of hydrological data can effectively support decision-making in flood prevention and disaster reduction, water resource management, water quality monitoring, and ecological and environmental protection.

[0003] The intelligent IoT sensing method for water conservancy projects applies IoT and intelligent sensing technologies to monitor hydrological data in real time and conduct dynamic analysis. Smart sensors and IoT devices deployed in river channels, dams, irrigation areas, and other areas collect hydrological information such as water level, flow, and quality. This data is then sent to a cloud platform for processing via a data transmission network. This approach aims to improve the monitoring and management efficiency of water conservancy projects.

[0004] Existing technologies for hydrological monitoring struggle to flexibly adjust to changing hydrological conditions, making it difficult to obtain critical data in a timely manner during hydrological anomalies, hindering the efficiency of emergency response. Furthermore, the predictive capabilities of collected data analysis are limited, hindering the accurate reflection of short-term hydrological dynamics. The identification and processing of abnormal data struggle to rapidly respond to multi-level anomalies under complex hydrological conditions, and the processing of large amounts of data can easily lead to network overload and storage pressure. Furthermore, the conflict between high-frequency sampling and inefficient data utilization often limits the monitoring accuracy and response speed of existing technologies in complex hydrological environments. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and propose a water conservancy intelligent Internet of Things perception method and system.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solution: a water conservancy intelligent Internet of Things perception method, comprising the following steps:

[0007] S1: Based on hydrological monitoring data, including flow, water level, and rainfall information, the statistical data is combined into a feature set to determine whether the data conforms to the normal distribution and generate a data distribution feature vector set;

[0008] S2: Based on the data distribution feature vector set, the observed hydrological data is updated, the influence of the latest data is increased by adjusting the data weight, the distribution characteristics after time decay are obtained, and future hydrological data are predicted to generate a dynamic distribution prediction value;

[0009] S3: Based on the dynamic distribution prediction value, extract the hydrological data within each time window, analyze the data change trend, directly compare the arrived hydrological data with the quantile boundaries, mark the data outside the boundaries as abnormal, and generate an abnormal data determination result;

[0010] S4: Based on the abnormal data determination results, the hydrological state is divided into multiple types, the state at the next moment is estimated according to the current state and previous changes, the priority of the data collection tasks in the monitoring area is adjusted, and a sampling priority scheduling result is generated;

[0011] S5: Based on the sampling priority scheduling result, the monitoring data is reconstructed in phase space, a compression threshold is set according to the complexity characteristics, the hydrological data is compressed section by section, and an optimized data set is generated;

[0012] S6: Based on the optimized data set and in combination with the feedback information from the IoT sensing devices, the sampling frequency and data transmission mode of each sensing node in the monitoring area are dynamically adjusted according to the data complexity and change trend to generate a monitoring frequency configuration result.

[0013] The present invention has improvements in that the data distribution feature vector set includes the mean, variance, skewness and kurtosis based on data statistics in the time window, the normal distribution judgment result, and the overall data distribution morphological characteristics; the dynamic distribution prediction value includes the hydrological data prediction value adjusted by time attenuation, the weighted correction result of the new data, and the data change trend prediction for the future time period; the abnormal data judgment result includes the data points that exceed the upper and lower quantile boundaries, and the out-of-bounds situation of each data point; the sampling priority scheduling result includes the sampling frequency adjustment according to the current hydrological state, the state transition probability, and the sampling priority division of the abnormal state; the optimized data set includes the feature retention result of the data, the compression ratio, and the segmented compression result of the data based on the compression threshold; the monitoring frequency configuration result includes the fluctuation area divided according to the local standard deviation, the sampling frequency setting of the regional boundary, and the weighted result of the partition sampling frequency.

[0014] The present invention has been improved in that, based on hydrological monitoring data, including flow, water level, and rainfall information, the statistical data are combined into a feature set, and whether the data conforms to a normal distribution is determined. The specific steps of generating a data distribution feature vector set are as follows:

[0015] S101: Based on various hydrological monitoring data such as flow, water level, and rainfall, the data is sliced according to the time window, the original hydrological observation values within each time period are extracted, and the original hydrological dataset within the time window is generated;

[0016] S102: Based on the original hydrological data set within the time window, calculate the statistics of flow, water level, and rainfall item by item, obtain the statistical distribution characteristics of the data by performing numerical operations on mean, variance, skewness, and kurtosis, and generate a basic statistical feature set of the hydrological data;

[0017] S103: Based on the basic statistical feature set of the hydrological data, an overall distribution model within the current time window is constructed, and the normality test of the statistics is performed to analyze the data distribution form, identify the central trend and dispersion of the distribution, and generate a data distribution feature vector set.

[0018] The present invention has been improved in that, based on the data distribution feature vector set, the observed hydrological data is updated, the influence of the latest data is increased by adjusting the data weight, the distribution characteristics after time decay are obtained, and future hydrological data are predicted. The specific steps of generating dynamic distribution prediction values are as follows:

[0019] S201: Based on the data distribution feature vector set, new observed hydrological data is acquired and imported, and added to the current data set according to the same time window standard to generate an updated hydrological data feature set;

[0020] S202: Based on the updated hydrological data feature set, weights are assigned to the latest data, and the influence of the latest data in the overall feature set is adjusted to generate a hydrological feature set with adjusted weights;

[0021] S203: Based on the weight-adjusted hydrological feature set, the features are gradually modified to predict future hydrological feature distribution and generate a dynamic distribution prediction value.

[0022] The present invention has the following improvements: based on the dynamic distribution prediction value, the hydrological data within each time window is extracted, the data change trend is analyzed, the arrived hydrological data is directly compared with the quantile boundary, and the data outside the boundary is marked as abnormal. The specific steps of generating the abnormal data determination result are as follows:

[0023] S301: Based on the dynamic distribution prediction value, extract the hydrological data within each time window, perform sorting processing on each segment of data, calculate the upper quantile and lower quantile of the data, determine the values of the upper and lower quantiles according to the calculated distribution position and the data sorting result, and generate a quantile boundary set;

[0024] S302: Based on the quantile boundary set, each newly arrived hydrological data segment is compared with the respective quantile boundary one by one, and each data point is checked to see whether it exceeds the upper and lower boundary ranges, thereby generating an out-of-bounds data identification result;

[0025] S303: Based on the out-of-bounds data identification result, the data exceeding the upper and lower quantile boundaries are marked, the exceeded time period and location are clarified, and the data are identified as abnormal hydrological data, and an abnormal data determination result is generated.

[0026] The present invention has been improved in that, based on the abnormal data determination results, the hydrological status is divided into multiple types, the status at the next moment is estimated according to the current status and past changes, and the priority of the data collection tasks in the monitoring area is adjusted. The specific steps of generating the sampling priority scheduling result are as follows:

[0027] S401: Based on the abnormal data determination result, the abnormal data is classified and processed, and the data is divided into multiple categories according to the hydrological state by referring to the degree of data deviation to generate a hydrological state classification result;

[0028] S402: Based on the hydrological state classification result and in combination with the change trajectory of the previous hydrological state, the transition frequency between multiple states is analyzed, the transition probability of the current state to other states is calculated, and a state transition probability analysis result is generated;

[0029] S403: Based on the state transition probability analysis result, with reference to the current state and the predicted state of the next period, the priority order of the data collection points in the monitoring area is adjusted, the data collection tasks are optimized, and a sampling priority scheduling result is generated.

[0030] The present invention has been improved by combining the changing trajectory of the hydrological state in the past and adopting the formula:

[0031]

[0032] Calculate in the time window, from the state s i 'Transfer to state s j ′’s conditional probability P(s j ′|s i ′);

[0033] Among them, n(s i ′→s j ′) represents the transition from state s i 'Transfer to state s j ′, ω i′j′ Indicates state s i ' to state s j ′ is the weight coefficient of , which represents the smoothing coefficient, i′ and j′ are state index symbols, and k′ represents the identifier of any state in the state index set.

[0034] The present invention has the following improvements: based on the sampling priority scheduling result, the monitoring data is reconstructed in phase space, a compression threshold is set according to the complexity characteristics, and the hydrological data is compressed section by section to generate an optimized data set:

[0035] S501: Based on the sampling priority scheduling result, the monitoring data is reconstructed in phase space, an embedding matrix of the time series is constructed, the dynamic characteristics of the time series are extracted, and a phase space reconstruction matrix is generated;

[0036] S502: Based on the phase space reconstruction matrix, determine the periodicity and randomness of data changes, and generate a complexity eigenvalue set by analyzing the periodic changes and random distribution of each segment of data;

[0037] S503: Based on the complexity feature value set, a compression threshold is set according to the complexity of each segment of data, detail features are retained for data with significant complexity that changes frequently and periodically, compression and data screening are performed on data with simple complexity, and an optimized data set is generated.

[0038] The present invention has been improved in that, based on the optimized data set and in combination with feedback information from IoT sensing devices, the sampling frequency and data transmission mode of each sensing node in the monitoring area are dynamically adjusted according to data complexity and change trends. The specific steps for generating the monitoring frequency configuration result are as follows:

[0039] S601: Based on the optimized data set, referring to the data complexity characteristics of each monitoring node, combined with the real-time monitoring feedback of the Internet of Things device, determine the data change of the node and generate a node complexity analysis result;

[0040] S602: Based on the node complexity analysis result, dynamically adjust the sampling frequency of the node according to the complexity and change trend of the monitored node, and generate a dynamic sampling frequency adjustment result;

[0041] S603: Based on the dynamic sampling frequency adjustment result, the data collection frequency and transmission interval of the monitoring node are adjusted in real time to optimize the sampling efficiency and generate a monitoring frequency configuration result.

[0042] A water conservancy intelligent Internet of Things sensing system, the system comprising:

[0043] The data feature extraction module is based on hydrological monitoring data, including flow, water level, and rainfall information. It combines the data into a feature set and extracts statistical features to determine whether it conforms to the normal distribution and generate a data distribution feature vector set;

[0044] The dynamic distribution prediction module updates the latest observed hydrological data based on the data distribution feature vector set, adjusts the data weight to highlight the impact of the latest data, predicts future hydrological data, and generates a dynamic distribution prediction value;

[0045] The anomaly detection module extracts the hydrological data within each time window based on the dynamic distribution prediction value, analyzes the data change trend, compares the data with the upper and lower quantile boundaries, identifies and marks abnormal data that exceeds the boundaries, and generates an abnormal data determination result;

[0046] The hydrological state classification module divides the hydrological state into multiple types based on the abnormal data judgment results, calculates the state transition probability by combining the current and historical state changes, and adjusts the data collection priority of the monitoring area according to the estimation results to generate a sampling priority scheduling result;

[0047] The sampling frequency adjustment module dynamically adjusts the sampling frequency and data transmission mode of the sensing nodes in the monitoring area based on the sampling priority scheduling result and the real-time feedback information of the Internet of Things devices to generate a monitoring frequency configuration result.

[0048] Compared with the prior art, the advantages and positive effects of the present invention are:

[0049] In the present invention, the changing trend of data is captured through multi-level data distribution feature analysis. The monitoring accuracy of data anomalies is enhanced. In the prediction process, the gradual integration of new data and weight adjustment enables the model to adapt to environmental changes in real time and accurately predict the hydrological status in the short term in the future. In addition, the quantile-based analysis can identify abnormal data in real time, quickly discover potential problems, and divide the status into different levels according to the degree of abnormality. In sampling scheduling, the priority of data collection is adjusted according to the probability model of state transition, so that monitoring resources are efficiently configured. By analyzing the complexity of the data, the monitoring data is compressed accordingly, which not only ensures the detailed integrity of highly complex data, but also reduces the storage and transmission burden of redundant data. Combined with the dynamic adjustment strategy of device feedback information, the sampling frequency and data transmission method of the sensing node can be flexibly changed, which improves the timeliness and effectiveness of monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a workflow diagram of the present invention;

[0051] Figure 2 This is a detailed flow chart of step S1 of the present invention;

[0052] Figure 3 This is a schematic diagram of a detailed process of step S2 of the present invention;

[0053] Figure 4 This is a detailed flow chart of step S3 of the present invention;

[0054] Figure 5 This is a detailed flow chart of step S4 of the present invention;

[0055] Figure 6 This is a detailed flow chart of step S5 of the present invention;

[0056] Figure 7 This is a detailed flow chart of step S6 of the present invention;

[0057] Figure 8 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0059] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description. They do not indicate or imply that the devices or elements referred to must have a specific direction, be constructed and operate in a specific direction, and therefore should not be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0060] See also Figure 1 The present invention provides a technical solution: a water conservancy intelligent Internet of Things perception method, comprising the following steps:

[0061] S1: Based on hydrological monitoring data, including flow, water level, and rainfall information, obtain the original hydrological data within the time window, calculate the mean, variance, skewness, and kurtosis item by item, combine the statistical data into a feature set, determine whether the data conforms to the normal distribution, construct the overall distribution characteristics of the current time window, and generate a data distribution feature vector set;

[0062] S2: Based on the data distribution feature vector set, the observed hydrological data is updated. The new data is added to the current feature set and gradually accumulated to correct the data distribution characteristics. The influence of the latest data is increased by adjusting the data weight, and the distribution characteristics after time decay are obtained. The correction results are combined to complete the future short-term hydrological data forecast and generate a dynamic distribution forecast value.

[0063] S3: Based on the dynamic distribution prediction value, extract the hydrological data within each time window, calculate the upper and lower quantiles of each data segment, analyze the data change trend, directly compare the arriving hydrological data with the quantile boundaries, determine whether there is out-of-bounds data, mark the data that exceeds the boundary as abnormal, and generate the abnormal data determination result;

[0064] S4: Based on the abnormal data judgment results, the hydrological status is divided into multiple types, including mild anomaly, moderate anomaly and severe anomaly. The current hydrological status and the changes in the past status are comprehensively considered, the transition probability between states is calculated, and the state at the next moment is estimated based on the current status and past changes. The priority of the data collection tasks in the monitoring area is adjusted to generate the sampling priority scheduling result;

[0065] S5: Based on the sampling priority scheduling results, the monitoring data is reconstructed in phase space, the approximate entropy of the time series is calculated, and the complexity characteristics of the data are determined. The complexity characteristics include periodicity and randomness. The compression threshold is set according to the complexity characteristics, and the hydrological data is compressed section by section. More feature details are retained for high-complexity data, and a larger ratio is compressed for low-complexity data to generate an optimized data set.

[0066] S6: Based on the optimized data set and the real-time feedback information from IoT sensing devices, the sampling frequency and data transmission mode of each sensing node in the monitoring area are dynamically adjusted according to the complexity and change trend of the data. The sensing nodes include water level monitors, flow sensors, and rainfall gauges, and the monitoring frequency configuration results are generated.

[0067] The data distribution feature vector set includes the mean, variance, skewness and kurtosis based on data statistics within the time window, the normal distribution judgment results, and the overall data distribution morphological characteristics. The dynamic distribution prediction value includes the hydrological data prediction value adjusted by time attenuation, the weighted correction result of the new data, and the data change trend prediction for the future period. The abnormal data judgment results include the data points that exceed the upper and lower quantile boundaries, and the out-of-bounds situation of each data point. The sampling priority scheduling results include the sampling frequency adjustment based on the current hydrological status, the state transition probability, and the sampling priority division of the abnormal state. The optimized data set includes the feature retention results of the data, the compression ratio, and the segmented compression results of the data based on the compression threshold. The monitoring frequency configuration results include the fluctuation area divided according to the local standard deviation, the sampling frequency setting of the regional boundary, and the weighted results of the partition sampling frequency.

[0068] See also Figure 2 ,Based on the hydrological monitoring data, including flow, water level, and rainfall information, ,the statistical data are combined into a feature set, and the ,data are judged to be in line with the normal distribution. ,The specific steps of generating the data distribution feature vector set are as follows:

[0069] S101: Based on various hydrological monitoring data such as flow, water level, and rainfall, the data is sliced according to the time window, the original hydrological observation values within each time period are extracted, and the original hydrological dataset within the time window is generated;

[0070] Based on various hydrological monitoring data such as flow, water level, rainfall, etc., slice processing is performed according to the set time window. By dividing the hydrological monitoring data of different time periods, the original hydrological observation values of each time period are extracted. First, the time consistency of each monitoring data must be ensured, and the missing or abnormal data must be preliminarily processed. For missing data, the previous value interpolation method can be used for processing, and for abnormal data, the upper and lower limit threshold methods are used to mark the abnormal values, and the observation values out of the range are eliminated. Then, the observation data of each time window are extracted one by one, combined with the flow, water level, rainfall and other data, and arranged into a new time series in the same time order. Finally, the data from different monitoring points are combined to form a time series set to obtain the original hydrological data set within the time window.

[0071] S102: Based on the original hydrological data set within the time window, the statistics of flow, water level, and rainfall are calculated item by item. By performing numerical operations on mean, variance, skewness, and kurtosis, the statistical distribution characteristics of the data are obtained to generate a basic statistical feature set of the hydrological data;

[0072] Calculate the statistics of flow, water level and rainfall item by item according to the formula:

[0073]

[0074]

[0075] The mean, variance, skewness and kurtosis of flow, water level and rainfall were calculated respectively.

[0076] In the formula, μ represents the mean, which indicates the average level of the monitored hydrological data (such as flow), σ 2 represents variance, indicating the degree of dispersion of hydrological data. S represents skewness, reflecting the symmetry of data distribution. Skewness greater than 0 indicates positive skewness, with the distribution shifting to the right. Skewness less than 0 indicates negative skewness, with the distribution shifting to the left. Skewness is the cubic average of the multiples of the standard deviation of each data point from the mean. K represents kurtosis, reflecting the peak or flatness of data distribution. A kurtosis greater than 3 indicates a peak, while a kurtosis less than 3 indicates flatness. Kurtosis is the fourth average of the multiples of the standard deviation of each data point from the mean. i is the observation value of the i-th hydrological data, which is usually obtained through hydrological monitoring equipment (water level meter, flow sensor, rain gauge, etc.), and n is the number of observation data, that is, the total number of observation points in the time window.

[0077] If the flow rate measurements in the observation data are x1=50, x2=55, and x3=53, then the mean is:

[0078]

[0079] variance:

[0080]

[0081] Skewness:

[0082]

[0083] Kurtosis:

[0084]

[0085] Based on historical data, if the mean flow rate for a certain period is 55 and the variance is 8.5, compared to the currently measured mean of 52.67 and variance of 6.22, the current flow rate data is relatively stable, and the skewness and kurtosis are close to a normal distribution. This indicates that the current hydrological conditions are close to the historical stable state and there are no significant abnormal changes in the hydrological data. The results show that the mean reflects the overall level of the observations, the variance indicates the degree of dispersion of the data, the skewness reflects the symmetry of the data, and the kurtosis reflects the degree of peaks or flatness of the data. All these results are used to generate the basic statistical feature set of hydrological data.

[0086] S103: Based on the basic statistical feature set of hydrological data, an overall distribution model within the current time window is constructed. By performing a normality test on the statistical quantity, the data distribution morphology is analyzed, the central tendency and dispersion of the distribution are identified, and a data distribution feature vector set is generated.

[0087] First, based on the basic statistical features generated previously, extract the mean, variance and other values, and select the appropriate test method to determine whether the current data conforms to the normal distribution. The normality test can be performed by D'

[0088] The Agostino test method is performed. This method tests the symmetry and peakness of the data by calculating the third and fourth order moments. According to the theory of normal distribution, for data that conforms to the normal distribution, its skewness is close to 0 and the kurtosis is close to 3. Therefore, based on the skewness and kurtosis calculated previously, the distribution form of the data can be judged. If the data passes the normality test, the central tendency and dispersion of its distribution are analyzed, and the quantiles of the observed values under the normal distribution are calculated to further identify the central and discrete trends of the data. Finally, through these distribution feature analyses, a set of data distribution feature vectors is generated.

[0089] See also Figure 3 ,Based on the data distribution feature vector set, the observed hydrological data is updated, the influence of the latest data is increased by adjusting the data weight, the distribution characteristics after time decay are obtained, and future hydrological data are predicted. The specific steps to generate dynamic distribution prediction values are as follows:

[0090] S201: Based on the data distribution feature vector set, new observed hydrological data is acquired and imported, and added to the current data set according to the same time window standard to generate an updated hydrological data feature set;

[0091] First, the hydrological data from different sources (such as flow, water level, and rainfall) are time-aligned, and synchronized timestamps are used to ensure that all data have consistent time tags. Then, data quality control methods are used, such as using mean interpolation to fill missing values and using upper and lower bounds to remove outliers. After preprocessing, each data is combined according to the same time window standard and combined with existing historical observation data to form a new time series data set. Subsequently, each observation is normalized using standardization operations to ensure the comparability of different physical quantities during the data processing process. After the data processing is completed, all processed observation data are gradually added to the existing data set according to the time window to generate an updated hydrological data feature set.

[0092] S202: Based on the updated hydrological data feature set, weights are assigned to the latest data, and the influence of the latest data in the overall feature set is adjusted to generate a hydrological feature set with adjusted weights;

[0093] First, extract the latest observations (such as flow and water level) for each time period, and adjust the impact of the new data on the overall data set through gradually increasing weight coefficients. Use dynamic weight allocation methods, such as time-weighted method, to gradually increase the weight of new data to ensure that the impact of the latest data is amplified in the overall data set. Fusion of new and old data is performed through weighted average function. In specific operations, different weighting factors can be assigned to each observation, and these factors can be dynamically set according to the trend and change amplitude of historical data to ensure that data with higher volatility obtains a larger weight allocation. After all observation data are weighted, a weight-adjusted hydrological feature set is generated.

[0094] S203: Based on the weight-adjusted hydrological feature set, the features are gradually modified to predict the future distribution of hydrological features and generate a dynamic distribution prediction value;

[0095] In stepwise correction, a time window is first defined; the data within the window serves as the basis for correction. At each time step, the feature values within the window are collected and compared with the current features to analyze any deviations. Next, correction factors are generated, using a regression analysis model to describe this relationship. The feature values are then updated accordingly to capture the changes.

[0096] The characteristics are gradually modified according to the future distribution of hydrological characteristics, according to the formula:

[0097]

[0098] Perform data prediction and generate dynamic distribution prediction values;

[0099] in, is the hydrological characteristic value at the future time t+k, representing the predicted value for a certain time point in the future. y(t) is the hydrological data observation value at the current time t, such as the flow, water level or rainfall data at the current time point, which is measured in real time by hydrological monitoring equipment (such as flow meter, water level gauge or rain gauge). k is the time step, which represents the prediction time interval. For example, the hydrological conditions in the future 1 hour, 2 hours or longer can be predicted. The time step can be set according to the actual application requirements. j is the weight of the jth observation data, representing the importance of each observation data in the prediction model. The weight is dynamically adjusted by time weighting or based on the volatility of the data. Generally, newer data or data with greater volatility are given higher weights. j (t) is the value of the j-th observation data at time t, which is the measured value of flow, water level or rainfall obtained by real-time monitoring equipment. ν(t) is the mean of the hydrological data at time t, reflecting the average level of the observation data. The mean is calculated by averaging all the data in the current time window. ξ(t) is the standard deviation of the hydrological data at time t, reflecting the fluctuation amplitude of the data. The standard deviation measures the degree of dispersion of the data by calculating the degree of deviation of the data from the mean.

[0100] If the current time t = 10 hours, the observed flow value y(t) = 60 cubic meters per second, the weight values v1 = 0.5, v2 = 0.2, v3 = 0.3, the observed data y1(t) = 58, y2(t) = 62, y3(t) = 59, the current mean ν(t) = 60, the standard deviation ξ(t) = 2, and the time step k = 2 hours, then we can get the following formula:

[0101]

[0102] The results show that the predicted flow rate in the next 2 hours is 59.1 cubic meters per second.

[0103] If the historical data for a particular period shows a mean of 60.5 cubic meters per second and a standard deviation of 3 cubic meters per second, and the current predicted flow rate is 59.5 cubic meters per second with minimal fluctuation, this indicates that the current hydrological flow rate is relatively stable. Compared to historical values, the current flow rate remains close to the mean with minimal deviation, indicating relatively stable hydrological conditions in the future. The resulting dynamic distribution forecast reflects the changing trends in future hydrological conditions and can provide support for future hydrological scheduling and response measures.

[0104] See also Figure 4Based on the dynamic distribution prediction value, the hydrological data within each time window is extracted. By analyzing the data change trend, the arrived hydrological data is directly compared with the quantile boundary, and the data exceeding the boundary is marked as abnormal. The specific steps for generating the abnormal data judgment result are as follows:

[0105] S301: Based on the dynamic distribution prediction value, extract the hydrological data within each time window, perform sorting processing on each data segment, calculate the upper and lower quantiles of the data, determine the values of the upper and lower quantiles based on the calculated distribution position and data sorting results, and generate a quantile boundary set;

[0106] Extract the hydrological data within each time window and sort them, and calculate the upper and lower quantiles according to the formula:

[0107]

[0108] Calculate the set of quantile boundaries;

[0109] Among them, Q u It is the upper quantile, which indicates the data point value above a certain percentage and is used to determine the upper limit of the data. l It is the lower quantile, which indicates the value of data points below a certain percentage and is used to determine the lower bound of the data. i is the ith hydrological data observation value after sorting, which is usually collected in real time by sensors, such as flow sensors or water level meters. Is the ceiling symbol, which means rounding the decimal up to the nearest integer, p u is the percentile value of the upper quantile, such as the value corresponding to 75%, which is used to determine the upper limit of the data, p l is the percentile value of the lower quantile, such as the value corresponding to 25%, which is used to determine the lower limit of the data. N is the total amount of observation data in the time window, which is obtained by statistically counting the observation data in the specified time period.

[0110] Suppose the flow observation data in a certain time window is X = {12, 15, 17, 18, 22, 24, 29}, with a total of N = 7 data points. Select the 75% and 25% percentile values as the upper and lower quantiles, that is, p u =0.75, p l =0.25.

[0111] Calculate the quantile index position according to the formula:

[0112]

[0113]

[0114] Therefore, the upper quantile Q u=X6=24, lower quantile Q l =X2=15.

[0115] If the upper quantile of the historical data is 26 and the lower quantile is 14, then by comparison, we can see that the upper quantile of the current dataset, 24, is lower than the historical upper limit of 26, indicating that the current flow extreme value is relatively low; at the same time, the lower quantile, 15, is slightly higher than the historical lower limit of 14, indicating that the low value part tends to be stable.

[0116] S302: Based on the quantile boundary set, each newly arrived hydrological data segment is compared with its respective quantile boundary one by one, and each data point is checked to see whether it exceeds the upper and lower boundary ranges, thereby generating an out-of-bounds data identification result;

[0117] By extracting each observation data and comparing it with the upper and lower quantile boundaries in the corresponding time window, the newly arrived data values are compared point by point with the quantile boundaries calculated in the time window, and the boundary screening method is used to mark the values that exceed the upper and lower limits. If the observation value is lower than the lower quantile or higher than the upper quantile, it is judged as out-of-bounds data. During the detection process, the timestamps and data points of all new data are matched one by one with the upper and lower quantiles in the corresponding time window. Independent quantile sets are used for boundary comparison of different types of data (such as flow, water level, rainfall, etc.) to generate out-of-bounds data identification results.

[0118] S303: Based on the out-of-bounds data identification results, mark the data that exceeds the upper and lower quantile boundaries, clarify the exceeded time period and location, identify them as abnormal hydrological data, and generate an abnormal data determination result;

[0119] The specific time and location information of these out-of-bounds data are recorded in sequence, and the abnormal points are classified into corresponding time periods through segmented storage. The specific categories of these data (such as flow anomaly, water level anomaly) are marked. The marked out-of-bounds data are sorted and formed into a traceable anomaly log. The trend and duration of its occurrence are analyzed, and the degree of exceeding the boundary is evaluated in combination with the historical data trend. By recording the location and time period of all abnormal data, potential abnormal events within the current observation range can be tracked and abnormal data judgment results can be generated.

[0120] See also Figure 5 Based on the abnormal data judgment results, the hydrological state is divided into multiple types. The state at the next moment is estimated according to the current state and previous changes. The priority of the data collection tasks in the monitoring area is adjusted. The specific steps for generating the sampling priority scheduling results are as follows:

[0121] S401: Based on the abnormal data determination result, the abnormal data is classified and processed, and the data is divided into multiple categories according to the hydrological state by referring to the degree of data deviation, thereby generating a hydrological state classification result;

[0122] By analyzing the degree of deviation of the observed data one by one, comparing the data with historical data within the normal range, using the deviation based on the standard deviation as the classification basis, the observed values are divided into normal, mild anomaly and severe anomaly categories according to the hydrological status. In the classification process, the historical fluctuation range of different hydrological variables (such as flow, water level, etc.) is referred to, and the classification threshold is dynamically set in combination with the historical quantile and fluctuation standard deviation of each variable. Time information is attached to each type of data to track its frequency of occurrence and time period distribution. Anomaly classification tables are used to record all classification results, and the data of each category are arranged in chronological order to form a structured classification log to generate hydrological status classification results.

[0123] S402: Based on the hydrological state classification results and in combination with the past hydrological state change trajectories, the transition frequencies between various states are analyzed, the transition probabilities from the current state to other states are calculated, and a state transition probability analysis result is generated;

[0124] Combined with the change trajectory of the previous hydrological state, the formula is adopted:

[0125]

[0126] Calculate in the time window, from the state s i 'Transfer to state s j ′’s conditional probability P(s j ′|s i ′);

[0127] Among them, n(s i ′→s j ′) represents the transition from state s i 'Transfer to state s j ′, which is obtained by monitoring the frequency of changes in historical status data, ω i′j′ Indicates state s i ' to state s j ′ is a weight coefficient, which is used to reflect the importance of different state transitions. The weight value can be set based on the fluctuation amplitude of specific hydrological variables (such as water level and flow) in the monitoring area or the degree of event risk. For example, rare but risky transition events are given higher weights. It represents the smoothing coefficient, which is used to avoid the situation where the denominator is zero and ensure the stability of the calculation. It is generally set to a positive number close to zero (such as 0.001). → represents the path relationship from one state to another. The states before and after the arrow are defined by the continuous state changes in the time series data. i′ and j′ are state index symbols used to identify different state sequences. k′ represents the identifier of any state in the set of all possible state indices, which is used to accumulate all states from s i ’The number of transitions to other states.

[0128] If the hydrological status of the monitoring area has three types: normal s1′, mild abnormal s2′ and severe abnormal s3′. In the past month, the number of times the normal state shifted to mild abnormality was 8 times, the number of times the normal state shifted to severe abnormality was 2 times, and the number of times the normal state remained normal was 10 times. The transfer weight coefficient is set to ω 1′2′ =1.2,ω 1′3′ =1.5,ω 1′1′ =1, smoothing coefficient = 0.001.

[0129] Enter the formula to calculate:

[0130]

[0131] If historical data shows a transition probability of 0.35 from normal to mildly abnormal, while the current analysis shows a probability of 0.398, this indicates an increase in recent hydrological fluctuations. The historical transition probability from normal to severely abnormal is 0.15, while the current probability is 0.124, indicating a decrease in the frequency of severe anomalies. This probability analysis provides a clear understanding of the current hydrological state trends and, by combining historical values with current calculations, can further optimize monitoring strategies and resource allocation.

[0132] The results show that the probability of transitioning from a normal state to a mild abnormal state is high, so it is necessary to focus on monitoring the evolution of mild abnormal states and take early warning measures. At the same time, the probability of transitioning to a severe abnormal state is low, but emergency response capabilities must still be maintained to deal with possible emergencies.

[0133] S403: Based on the state transition probability analysis results, with reference to the current state and the predicted state of the next period, the priority order of the data collection points in the monitoring area is adjusted to optimize the data collection tasks and generate a sampling priority scheduling result;

[0134] The priority scheduling algorithm is used to reallocate collection resources, refer to high-probability state transition paths, and increase the priority of relevant collection points. During the scheduling process, the frequency and risk level of state transitions in different areas are combined to divide the collection tasks into high, medium, and low priorities, and more frequent collection is performed in high-priority areas. In the scheduling optimization, a dynamic weighted adjustment algorithm is used to match collection tasks with sensor resources to ensure that high-risk areas have sufficient monitoring resources. All scheduling results are recorded in a priority table and automatically synchronized to the data acquisition system to generate sampling priority scheduling results.

[0135] See also Figure 6 ,Based on the sampling priority scheduling results, the phase space of the monitoring data is reconstructed, the compression threshold is set according to the complexity characteristics, the hydrological data is compressed section by section, and the specific steps to generate the optimized data set are as follows:

[0136] S501: Based on the sampling priority scheduling result, the monitoring data is reconstructed in phase space, the embedding matrix of the time series is constructed, the dynamic characteristics of the time series are extracted, and the phase space reconstruction matrix is generated;

[0137] By extracting data points of different variables in the time series, the one-dimensional time series is reconstructed into an embedding matrix. By setting the embedding delay time and embedding dimension, the time series is expanded in the multidimensional phase space. First, the autocorrelation function is used to calculate the correlation between different lag points of the time series to determine the optimal embedding delay time. The result of the autocorrelation function provides a correlation measure for each pair of lag points, and the position where the autocorrelation decays fastest is used as the reference value of the embedding delay; then, the false neighbor point method is used to evaluate the aggregation of data points under different embedding dimensions, and the embedding dimension is increased until the proportion of false neighbor points drops to an acceptable range, so that the reconstructed phase space can capture the global dynamic behavior of the system; all embedded data points are combined in chronological order to form an embedding matrix. Each column of the embedding matrix represents the state of the time series under a specific embedding dimension, ensuring that the temporal correlation and intrinsic dynamic characteristics of the monitoring data are effectively preserved, and generating a phase space reconstruction matrix.

[0138] S502: Based on the phase space reconstruction matrix, determine the periodicity and randomness of data changes, and generate a complexity eigenvalue set by analyzing the periodic changes and random distribution of each data segment;

[0139] First, the autocorrelation function is used to analyze the periodicity within the time series. The autocorrelation function determines the pattern of periodicity based on the changes in the correlation of the time series at different lag steps. If the autocorrelation value peaks after a certain lag time, this lag time is the period length. Then, the Fourier transform is used to convert the time series into frequency domain form to obtain the main frequency components of the time series. The peak frequency in the spectrum diagram represents the main frequency of periodic changes. In addition, to evaluate the randomness of the data, the Shannon entropy is used to calculate the uncertainty of the information in the time series. The higher the entropy value, the more random components in the sequence, and the lower the entropy value, the more dominant the periodic components. After combining the periodic results of the Fourier transform with the random characteristics calculated by the entropy value, a set of complexity feature values describing the periodicity of the time series and the proportion of random components is generated.

[0140] S503: Based on the complexity feature value set, a compression threshold is set according to the complexity of each data segment, detail features are retained for data with significant complexity that changes frequently and periodically, and compression and data screening are performed on data with simple complexity to generate an optimized data set;

[0141] First, a compression threshold is set, and data of different complexities are classified according to the compression standard. Data with obvious periodic changes and high complexity are retained to avoid the loss of important periodic features. For simple randomly distributed data, piecewise linear interpolation is used to downsample the data. The endpoints in each segment of data are selected as key sampling points, and the missing data are supplemented according to the linear trend between the endpoints to reduce the data storage volume while ensuring that the overall trend remains unchanged. In addition, downsampling is performed on some data that lack periodic features to reduce data redundancy. All filtered and compressed data are recombined in chronological order to generate an optimized data set.

[0142] See also Figure 7 Based on the optimized data set and the feedback information from IoT sensing devices, the sampling frequency and data transmission mode of each sensing node in the monitoring area are dynamically adjusted according to the data complexity and change trend. The specific steps for generating the monitoring frequency configuration result are as follows:

[0143] S601: Based on the optimized data set, referring to the data complexity characteristics of each monitoring node and combining with the real-time monitoring feedback of the IoT device, the data change of the node is determined and the node complexity analysis results are generated;

[0144] Referring to the data complexity characteristics of each monitoring node and combining with the real-time monitoring feedback of IoT devices, the changes of the nodes are analyzed. The real-time data of each node is obtained through the data aggregation module of the IoT platform. Autocorrelation analysis is used to identify the trend of the time series, and spectrum analysis is used to detect the periodicity of the data. Combined with these analysis results, the fluctuation amplitude and change rate of the monitoring data are judged. By comparing the characteristic changes of historical data, it is identified whether the current node data is in an abnormal state or a trend change is occurring, and the node complexity analysis results are generated for each node.

[0145] S602: Based on the node complexity analysis result, dynamically adjust the node sampling frequency according to the complexity and change trend of the monitored node, and generate a dynamic sampling frequency adjustment result;

[0146] The sampling period is dynamically set through an adaptive frequency control algorithm. The adaptive sampling interval is selected by real-time analysis of the node change amplitude. The sampling interval is shortened for nodes with a higher data change frequency to ensure the integrity of data collection. The sampling interval is extended for nodes with slower data changes or stable trends to reduce device power consumption. In specific operations, according to the trend graph of the complexity analysis results and combined with the detection of the slope change rate, the sampling interval of different nodes is dynamically calculated and set to generate a dynamic sampling frequency adjustment result.

[0147] S603: Based on the dynamic sampling frequency adjustment result, adjust the data collection frequency and transmission interval of the monitoring node in real time to optimize the sampling efficiency and generate a monitoring frequency configuration result;

[0148] The data transmission efficiency is optimized through wireless communication protocols. A short-interval data transmission frequency is adopted for nodes with higher complexity to ensure the real-time nature of the data. For nodes with lower complexity and slowly changing data, a batch upload mode is used to reduce the transmission frequency. During the configuration process, the sampling interval and transmission delay parameters of the nodes are gradually adjusted to dynamically optimize the coordination relationship between sampling and transmission, ensuring that the data transmission and sampling frequency of each node reach the optimal state, and generating the monitoring frequency configuration result.

[0149] See also Figure 8 , a water conservancy intelligent Internet of Things sensing system, the system includes:

[0150] The data feature extraction module is based on hydrological monitoring data, including flow, water level, and rainfall information. It combines the data into a feature set and extracts statistical features to determine whether it conforms to the normal distribution and generate a data distribution feature vector set;

[0151] The dynamic distribution prediction module updates the latest observed hydrological data based on the data distribution feature vector set, adjusts the data weight to highlight the impact of the latest data, predicts future hydrological data, and generates dynamic distribution prediction values;

[0152] The anomaly detection module extracts hydrological data within each time window based on the dynamic distribution prediction value. By analyzing the data change trend, the data is compared with the upper and lower quantile boundaries, and the abnormal data that exceeds the boundaries is identified and marked to generate the abnormal data judgment result.

[0153] The hydrological status classification module divides the hydrological status into multiple types based on the abnormal data judgment results. By combining the changes in the current and historical status, it calculates the state transition probability, adjusts the data collection priority of the monitoring area based on the estimation results, and generates the sampling priority scheduling results;

[0154] The sampling frequency adjustment module dynamically adjusts the sampling frequency and data transmission mode of the sensing nodes in the monitoring area based on the sampling priority scheduling results and the real-time feedback information of the IoT devices to generate the monitoring frequency configuration results.

[0155] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A water conservancy intelligent Internet of Things perception method, characterized in that: The following steps are involved: Based on hydrological monitoring data, including flow, water level, and rainfall information, the statistical data is combined into a feature set to determine whether the data conforms to the normal distribution and generate a data distribution feature vector set; Based on the data distribution feature vector set, the observed hydrological data is updated, the influence of the latest data is increased by adjusting the data weight, the distribution characteristics after time decay are obtained, and future hydrological data are predicted to generate a dynamic distribution prediction value; Based on the dynamic distribution prediction value, extract the hydrological data within each time window, analyze the data change trend, directly compare the arrived hydrological data with the quantile boundary, mark the data outside the boundary as abnormal, and generate an abnormal data determination result; Based on the abnormal data determination results, the hydrological state is divided into multiple types, the state at the next moment is estimated according to the current state and previous changes, the priority of the data collection tasks in the monitoring area is adjusted, and the sampling priority scheduling result is generated; Based on the sampling priority scheduling result, the monitoring data is reconstructed in phase space, a compression threshold is set according to the complexity characteristics, the hydrological data is compressed section by section, and an optimized data set is generated; Based on the optimized data set and the feedback information from IoT sensing devices, the sampling frequency and data transmission mode of each sensing node in the monitoring area are dynamically adjusted according to the data complexity and change trend to generate a monitoring frequency configuration result.

2. The water conservancy intelligent IoT sensing method according to claim 1 is characterized by: The data distribution feature vector set includes the mean, variance, skewness and kurtosis of data statistics within the time window, the normal distribution judgment result, and the overall data distribution morphological characteristics. The dynamic distribution prediction value includes the hydrological data prediction value adjusted by time attenuation, the weighted correction result of the new data, and the data change trend prediction for the future time period. The abnormal data judgment result includes the data points that exceed the upper and lower quantile boundaries, and the out-of-bounds situation of each data point. The sampling priority scheduling result includes the sampling frequency adjustment according to the current hydrological state, the state transition probability, and the sampling priority division of the abnormal state. The optimized data set includes the feature retention result of the data, the compression ratio, and the segmented compression result of the data based on the compression threshold. The monitoring frequency configuration result includes the fluctuation area divided according to the local standard deviation, the sampling frequency setting of the regional boundary, and the weighted result of the partition sampling frequency.

3. The water conservancy intelligent IoT sensing method according to claim 1 is characterized by: Based on hydrological monitoring data, including flow, water level, and rainfall information, the statistical data is combined into a feature set to determine whether the data conforms to the normal distribution. The specific steps for generating the data distribution feature vector set are as follows: Based on various hydrological monitoring data such as flow, water level, and rainfall, the data is sliced according to the time window, the original hydrological observation values within each time period are extracted, and the original hydrological data set within the time window is generated; Based on the original hydrological data set within the time window, the statistics of flow, water level and rainfall are calculated item by item, and the statistical distribution characteristics of the data are obtained by performing numerical operations on the mean, variance, skewness and kurtosis to generate a basic statistical feature set of the hydrological data; Based on the basic statistical feature set of the hydrological data, an overall distribution model within the current time window is constructed. By testing the normality of the statistical quantity, the form of the data distribution is analyzed, the central trend and the degree of dispersion of the distribution are identified, and a data distribution feature vector set is generated.

4. The water conservancy intelligent IoT sensing method according to claim 1, characterized in that: Based on the data distribution feature vector set, the observed hydrological data is updated, the influence of the latest data is increased by adjusting the data weight, the distribution characteristics after time decay are obtained, and future hydrological data are predicted. The specific steps for generating dynamic distribution prediction values are as follows: Based on the data distribution feature vector set, new observed hydrological data is acquired and imported, and added to the current data set according to the same time window standard to generate an updated hydrological data feature set; Based on the updated hydrological data feature set, weighting the latest data is assigned, and the influence of the latest data in the overall feature set is adjusted to generate a hydrological feature set with adjusted weights; Based on the weight-adjusted hydrological feature set, the features are gradually modified to predict the future distribution of hydrological features and generate a dynamic distribution prediction value.

5. The water conservancy intelligent IoT sensing method according to claim 1 is characterized by: Based on the dynamic distribution prediction value, the hydrological data within each time window is extracted. By analyzing the data change trend, the arrived hydrological data is directly compared with the quantile boundaries, and the data outside the boundaries is marked as abnormal. The specific steps for generating the abnormal data determination result are as follows: Based on the dynamic distribution prediction value, extract the hydrological data within each time window, perform sorting processing on each segment of data, calculate the upper quantile and lower quantile of the data, determine the values of the upper and lower quantiles according to the calculated distribution position and the data sorting result, and generate a quantile boundary set; Based on the quantile boundary set, each newly arrived hydrological data segment is compared with the respective quantile boundary one by one, and each data point is checked to see whether it exceeds the upper and lower boundary ranges, thereby generating an out-of-bounds data identification result; Based on the out-of-bounds data identification result, the data exceeding the upper and lower quantile boundaries are marked, the exceeded time period and location are clearly identified, and the data is identified as abnormal hydrological data, and an abnormal data determination result is generated.

6. The water conservancy intelligent IoT sensing method according to claim 1, characterized in that: Based on the abnormal data determination results, the hydrological state is divided into multiple types, the state at the next moment is estimated based on the current state and previous changes, and the priority of the data collection tasks in the monitoring area is adjusted. The specific steps for generating the sampling priority scheduling results are as follows: Based on the abnormal data determination result, the abnormal data is classified and processed, and the data is divided into multiple categories according to the hydrological state by referring to the degree of data deviation to generate a hydrological state classification result; Based on the hydrological state classification results and in combination with the change trajectory of the previous hydrological state, the transition frequencies between multiple states are analyzed, the transition probability of the current state to other states is calculated, and the state transition probability analysis results are generated; Based on the state transition probability analysis results, with reference to the current state and the predicted state of the next period, the priority order of data collection points in the monitoring area is adjusted, the data collection tasks are optimized, and the sampling priority scheduling results are generated.

7. The water conservancy intelligent IoT sensing method according to claim 1 is characterized by: Combined with the change trajectory of the previous hydrological state, the formula is adopted: Calculate in the time window, from the state s i 'Transfer to state s j ′’s conditional probability P (s j ′∣s i ′); Among them, n(s i ′→s j ′) represents the transition from state s i 'Transfer to state s j ′, ω i′j′ Indicates state s i ' to state s j ′ is the weight coefficient of , which represents the smoothing coefficient, i′ and j′ are state index symbols, and k′ represents the identifier of any state in the state index set.

8. The water conservancy intelligent IoT sensing method according to claim 1 is characterized by: Based on the sampling priority scheduling results, the monitoring data is reconstructed in phase space, the compression threshold is set according to the complexity characteristics, and the hydrological data is compressed section by section to generate the optimized data set. The specific steps are as follows: Based on the sampling priority scheduling result, the monitoring data is reconstructed in phase space, an embedding matrix of the time series is constructed, the dynamic characteristics of the time series are extracted, and a phase space reconstruction matrix is generated; Based on the phase space reconstruction matrix, the periodicity and randomness of data changes are judged, and a complexity eigenvalue set is generated by analyzing the periodic changes and random distribution of each segment of data; Based on the complexity feature value set, a compression threshold is set according to the complexity of each segment of data, detailed features are retained for data with significant complexity that changes frequently and periodically, compression and data screening are performed on data with simple complexity to generate an optimized data set.

9. The water conservancy intelligent IoT sensing method according to claim 1, characterized in that: Based on the optimized data set and the feedback information from IoT sensing devices, the sampling frequency and data transmission method of each sensing node in the monitoring area are dynamically adjusted according to the data complexity and change trend. The specific steps for generating the monitoring frequency configuration result are as follows: Based on the optimized data set, referring to the data complexity characteristics of each monitoring node, combined with the real-time monitoring feedback of the Internet of Things devices, the data changes of the node are judged and the node complexity analysis results are generated; Based on the node complexity analysis result, dynamically adjust the node sampling frequency according to the complexity and change trend of the monitored node, and generate a dynamic sampling frequency adjustment result; Based on the dynamic sampling frequency adjustment result, the data collection frequency and transmission interval of the monitoring node are adjusted in real time to optimize the sampling efficiency and generate a monitoring frequency configuration result.

10. A water conservancy intelligent IoT sensing system, characterized in that: According to the water conservancy intelligent Internet of Things perception method according to any one of claims 1 to 9, the system includes: The data feature extraction module is based on hydrological monitoring data, including flow, water level, and rainfall information. It combines the data into a feature set and extracts statistical features to determine whether it conforms to the normal distribution and generate a data distribution feature vector set; The dynamic distribution prediction module updates the latest observed hydrological data based on the data distribution feature vector set, adjusts the data weight to highlight the impact of the latest data, predicts future hydrological data, and generates a dynamic distribution prediction value; The anomaly detection module extracts the hydrological data within each time window based on the dynamic distribution prediction value, analyzes the data change trend, compares the data with the upper and lower quantile boundaries, identifies and marks abnormal data that exceeds the boundaries, and generates an abnormal data determination result; The hydrological state classification module divides the hydrological state into multiple types based on the abnormal data judgment results, calculates the state transition probability by combining the current and historical state changes, and adjusts the data collection priority of the monitoring area according to the estimation results to generate a sampling priority scheduling result; The sampling frequency adjustment module dynamically adjusts the sampling frequency and data transmission mode of the sensing nodes in the monitoring area based on the sampling priority scheduling result and the real-time feedback information of the Internet of Things devices to generate a monitoring frequency configuration result.

Images