Method for analyzing hydrological weather data in water supply system

Abstract

The invention discloses a method for analyzing hydrological weather data in a water supply system, and belongs to the technical field of water supply management, and the method comprises the following steps: S1, obtaining multi-source weather data, and carrying out formatting processing by adopting a unified timestamp and a unified coordinate system to obtain a standardized data set; s2, constructing feature vectors of the standardized data set, and performing dimension reduction processing on a set formed by the feature vectors by adopting a principal component analysis method to obtain a simplified feature set; s3, inputting the simplified feature set and the running parameters of the water supply system into a long-short term memory network model, and training an association relationship between the simplified feature set and the running parameters of the water supply system to obtain an association prediction model; s4, acquiring real-time multi-source weather data, inputting the real-time multi-source weather data into the association prediction model after the real-time multi-source weather data is subjected to standardization processing in the step S1 and dimension reduction processing in the step S2, and acquiring water supply system operation parameters output by the association prediction model.

Description

Technical Field

[0001] The present invention relates to the technical field of water supply management, and in particular to a method for analyzing hydrological and weather data in a water supply system. Background Art

[0002] The application of hydrological and weather data analysis in water supply systems is an important research field to ensure urban water resource management and scheduling. Its core lies in improving the water supply system's responsiveness to water resource demand and adaptability to extreme weather events through accurate meteorological information forecasting and analysis.

[0003] With the acceleration of urbanization and the intensification of climate change, water supply systems require more efficient data-driven decision support. However, existing methods often face problems such as data heterogeneity, difficulty in integration, and lack of real-time performance when processing multi-source weather data. This results in low accuracy in the correlation analysis between weather factors and water supply operating parameters, making it difficult to meet the needs of intelligent scheduling. In addition, when establishing the correlation model between weather data and water supply system operating parameters, existing methods often ignore the dynamic influence of key meteorological factors such as rainfall, temperature, and wind speed, resulting in insufficient robustness of the prediction model. These problems directly restrict the accuracy of water supply scheduling in the face of extreme weather such as heavy rain and drought. Summary of the Invention

[0004] In order to overcome the defects of the prior art, the present invention provides a method for analyzing hydrological and weather data in a water supply system to solve the above problems.

[0005] The technical solution adopted by the present invention to solve the technical problem is: a method for analyzing hydrological and weather data in a water supply system, comprising the following steps: S1: Acquire multi-source weather data, wherein the multi-source weather data includes rainfall, temperature, and wind speed parameters, and format the data using a unified timestamp and a unified coordinate system to obtain a standardized data set; S2: constructing a feature vector of the standardized data set, wherein the feature vector includes rainfall intensity, temperature change, and wind speed direction parameters, wherein the rainfall intensity is calculated by rainfall amount, the temperature change is calculated by temperature, and the wind speed direction is calculated by wind speed. The principal component analysis method is used to perform dimensionality reduction processing on the set of feature vectors to obtain a streamlined feature set; S3: Inputting the simplified feature set and the operating parameters of the water supply system into a long short-term memory network model, training the correlation between the simplified feature set and the operating parameters of the water supply system, and obtaining a correlation prediction model; S4: Acquire real-time multi-source weather data, input the real-time multi-source weather data into the association prediction model after the standardization processing of step S1 and the dimensionality reduction processing of step S2, and obtain the water supply system operation parameters output by the association prediction model.

[0006] Preferably, in step S1, the step of formatting using a unified timestamp includes: A timestamp detection method is used to identify the timestamp format of the multi-source weather data, and the heterogeneous timestamps are converted into ISO8601 format to obtain a time-aligned dataset.

[0007] Optionally, in step S1, the step of performing formatting processing using a unified coordinate system includes: converting the coordinate system of the time-aligned dataset into a WGS84 geographic coordinate system based on a coordinate conversion method to obtain a coordinate unified dataset.

[0008] Specifically, in step S1, a linear interpolation method is used to fill in the missing values ​​in the coordinate unified data set to obtain an interpolated data set; Outliers that exceed three times the standard deviation of the mean in the interpolated data set are removed by an outlier detection method to obtain a cleaned data set; The Z-score standardization method is used to convert the rainfall, temperature and wind speed parameters in the cleaned data set into a distribution with a mean of 0 and a variance of 1 to obtain a standardized data set.

[0009] It is worth noting that in step S2, the standardized rainfall intensity is calculated based on the rainfall in the standardized data set, the standardized temperature gradient is calculated based on the temperature in the standardized data set, and the standardized wind speed angle is calculated based on the wind speed in the standardized data set to form an initial parameter set.

[0010] Preferably, in step S2, the sliding window method is used to process the initial parameter set, the window size is set to a fixed length, the step size is set to a unit interval, and the rolling function of the pandas library is used to calculate the mean, variance, maximum value and minimum value of the rainfall intensity, temperature gradient and wind speed angle in the initial parameter set to obtain a time series feature set.

[0011] Specifically, in step S2, principal component analysis is performed on the time series feature set using the PCA class of the sklearn library, a variance contribution rate threshold is set, and features whose variance contribution rates exceed the variance contribution rate threshold are retained as main feature components to obtain a streamlined feature set.

[0012] Optionally, in step S3, the operating parameters of the water supply system are obtained, and the time sequence of the reduced feature set and the operating parameters of the water supply system are aligned using a timestamp matching method; The Z-score standardization method is used to standardize the water supply system operating parameters after time series alignment, and the mean and standard deviation of each parameter are calculated to obtain the standardized water supply system operating parameters; The standardized water supply system operating parameters and the streamlined feature set after time series alignment are used as input. The long short-term memory network model is used for training. A fixed time step is set to construct the correlation relationship between the water supply system operating parameters and the streamlined feature set, and obtain the initial correlation prediction model.

[0013] It is worth noting that, in step S3, if the prediction error of the initial association model exceeds a preset threshold, the initial association model is optimized by reducing the learning rate and increasing the number of hidden layer nodes to obtain an optimized association prediction model.

[0014] The beneficial effect of the present invention is that in the method for analyzing hydrological and weather data in the water supply system, multi-source weather data including radar, satellite and ground stations is first obtained, and the data is standardized and cleaned. Then, spatiotemporal features are extracted and feature vectors are constructed, and a simplified feature set is obtained by dimensionality reduction through principal component analysis. The simplified feature set and the operating parameters of the water supply system are input into the long short-term memory network model to train the correlation between weather data and operating parameters. Finally, the real-time weather data is input into the association prediction model to obtain the predicted operating parameters of the water supply system. Subsequently, the scheduling plan can be implemented based on the predicted operating parameters of the water supply system. This method can effectively integrate multi-source meteorological data, realize intelligent and precise scheduling of the water supply system, and improve the reliability of water supply. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flow chart of a method for analyzing hydrological and weather data in a water supply system according to an embodiment of the present invention; Figure 2 A step-by-step flow chart of step S1 in one embodiment of the present invention; Figure 3 A step-by-step flow chart of step S2 in one embodiment of the present invention; Figure 4 1 is a step-by-step flow chart of step S3 in one embodiment of the present invention. DETAILED DESCRIPTION

[0016] The following is a further description of specific embodiments of the present invention in conjunction with the accompanying drawings. It should be noted that the description of these embodiments is intended to facilitate understanding of the present invention and does not constitute a limitation of the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0017] like Figure 1-4 As shown, a method for analyzing hydrological and weather data in a water supply system includes the following steps: S1: Acquire multi-source weather data, which includes rainfall, temperature, and wind speed parameters, and format them using a unified timestamp and a unified coordinate system to obtain a standardized data set; illustratively, obtain radar, satellite, and ground station data through a multi-source data interface. For example, radar data provides hourly rainfall in millimeters; satellite data provides cloud top temperature in degrees Celsius; and ground station data records wind speed in meters per second. Suppose that within an hour in a certain area, the radar records 5 mm of rainfall, the satellite records a cloud top temperature of -10 degrees Celsius, and the ground station records a wind speed of 3 meters per second. These data are acquired from the meteorological database in real time through the API interface, ensuring the diversity and coverage of the data, and providing rich basic information for subsequent analysis; S2: constructing a feature vector of the standardized data set, wherein the feature vector includes rainfall intensity, temperature change, and wind speed direction parameters, wherein the rainfall intensity is calculated by rainfall amount, the temperature change is calculated by temperature, and the wind speed direction is calculated by wind speed. The principal component analysis method is used to perform dimensionality reduction processing on the set of feature vectors to obtain a streamlined feature set; S3: Inputting the simplified feature set and the operating parameters of the water supply system into a long short-term memory network model, training the correlation between the simplified feature set and the operating parameters of the water supply system, and obtaining a correlation prediction model; S4: Acquire real-time multi-source weather data, subject the real-time multi-source weather data to standardization processing in step S1 and dimensionality reduction processing in step S2, and then input it into the association prediction model to obtain the water supply system operating parameters output by the association prediction model. In this embodiment, the real-time multi-source weather data is acquired and subjected to standardization processing in step S1 to obtain a standardized dataset corresponding to the real-time multi-source weather data. The standardized dataset corresponding to the real-time multi-source weather data is subjected to dimensionality reduction processing in step S2 to obtain a reduced feature set corresponding to the real-time multi-source weather data, thereby conforming to the input data format of the association prediction model. In this embodiment, the reduced feature set and water supply system operating parameters input into the long short-term memory network model when training the association prediction model are historical data, while the reduced feature set input into the association prediction model during actual prediction is real-time data.

[0018] In the method for analyzing hydrological and weather data in the water supply system, multi-source weather data including radar, satellite and ground stations is first obtained, and the data is standardized and cleaned. Then, spatiotemporal features are extracted and feature vectors are constructed, and a simplified feature set is obtained through dimensionality reduction through principal component analysis. The simplified feature set and the operating parameters of the water supply system are input into the long short-term memory network model to train the association between weather data and operating parameters. Finally, the real-time weather data is input into the association prediction model to obtain the predicted operating parameters of the water supply system. Subsequently, the scheduling plan can be implemented based on the predicted operating parameters of the water supply system. This method can effectively integrate multi-source meteorological data, realize intelligent and precise scheduling of the water supply system, and improve the reliability of water supply.

[0019] It is worth noting that, in step S1, the step of formatting the data using a unified timestamp includes: identifying the timestamp format of the multi-source weather data using a timestamp detection method, converting the heterogeneous timestamps into the ISO8601 format, and obtaining a time-aligned dataset.

[0020] In one possible implementation, a timestamp detection method is used to identify the timestamp format of multi-source weather data. For example, the timestamp for rainfall corresponding to radar data is 2025-04-27 08:00:00+08:00, the timestamp for temperature corresponding to satellite data is 27 / 04 / 2025 08:00 UTC, and the timestamp for wind speed parameters corresponding to ground station data is 202504270800. After detection, if the formats are inconsistent, all timestamps are parsed and converted to the ISO8601 format, such as 2025-04-27T08:00:00Z. This conversion ensures consistency in the temporal dimension of the time-aligned datasets, facilitating subsequent spatiotemporal analysis and improving the accuracy of data fusion.

[0021] Specifically, in step S1, the step of formatting using a unified coordinate system includes: converting the coordinate system of the time-aligned dataset into a WGS84 geographic coordinate system based on a coordinate conversion method to obtain a coordinate unified dataset.

[0022] Specifically, the coordinate system of the time-aligned dataset is converted to the WGS84 geographic coordinate system using a coordinate conversion method. For example, the projected coordinates of rainfall corresponding to radar data are in UTM format, the projected coordinates of temperature corresponding to satellite data are in latitude and longitude format, and the projected coordinates of wind speed parameters corresponding to ground station data are in local plane coordinates. Through the coordinate conversion algorithm, all data points are mapped to the WGS84 coordinate system, for example, a point is converted from UTM coordinates (500000, 4649776) to latitude and longitude (120.5E, 42.0N). This unified dataset eliminates spatial reference differences and improves data compatibility in geospatial analysis.

[0023] Preferably, in step S1, a linear interpolation method is used to fill in missing values ​​in the coordinate unified data set to obtain an interpolated data set; Outliers that exceed three times the standard deviation of the mean in the interpolated data set are removed by an outlier detection method to obtain a cleaned data set; The Z-score standardization method is used to convert the rainfall, temperature and wind speed parameters in the cleaned data set into a distribution with a mean of 0 and a variance of 1 to obtain a standardized data set.

[0024] Linear interpolation is used to fill missing values ​​in a unified coordinate dataset. For example, if a ground station has missing rainfall data at 9:00 AM, but rainfall data at 8:00 AM and 10:00 AM are 4 mm and 6 mm, respectively, linear interpolation can be used to estimate the rainfall at 9:00 AM to be 5 mm. This method utilizes the trend of the preceding and following data to fill in the gaps, enhancing the integrity of the interpolated dataset and providing continuous data support for subsequent analysis.

[0025] In one embodiment, an outlier detection method is used to remove outliers from the interpolated dataset that exceed three standard deviations of the mean. For example, if the mean rainfall in a region is 5 mm and the standard deviation is 1 mm, a record of 10 mm at a certain point exceeds the mean plus three standard deviations (8 mm) and is therefore removed. This cleaned dataset eliminates abnormal interference, improves data reliability, and lays the foundation for accurate meteorological analysis.

[0026] The Z-score normalization method was used to transform the rainfall, temperature, and wind speed parameters in the cleaned dataset into a distribution with a mean of 0 and a variance of 1. For example, if the rainfall at a point is 6 mm, the mean is 5 mm, and the standard deviation is 1 mm, the Z-score would be (6-5) / 1 = 1. Temperature and wind speed were also normalized to produce a standardized dataset. This method eliminates dimensional differences, making different parameters directly comparable and improving the data's suitability for machine learning models.

[0027] It is worth noting that in step S2, the standardized rainfall intensity is calculated based on the rainfall in the standardized data set, the standardized temperature gradient is calculated based on the temperature in the standardized data set, and the standardized wind speed angle is calculated based on the wind speed in the standardized data set to form an initial parameter set.

[0028] Rainfall intensity reflects the change in rainfall per unit time, temperature gradient indicates the spatial temperature change rate, and wind speed angle describes the dynamic distribution of wind direction. For example, the standardized rainfall intensity in a certain area is 0.2, the standardized temperature gradient is 0.5, and the standardized wind speed angle is 0.45. These parameters are extracted from the standardized dataset to form the initial parameter set.

[0029] Preferably, in step S2, the sliding window method is used to process the initial parameter set, the window size is set to a fixed length, the step size is set to a unit interval, and the rolling function of the pandas library is used to calculate the mean, variance, maximum value and minimum value of the rainfall intensity, temperature gradient and wind speed angle in the initial parameter set to obtain a time series feature set.

[0030] In one possible implementation, a sliding window approach is used to process the initial parameter set, with a window size of 6 hours and a step size of 1 hour. The rolling function in the pandas library is used to calculate the mean, variance, maximum, and minimum values ​​of rainfall intensity, the mean, variance, maximum, and minimum values ​​of temperature gradient, and the mean, variance, maximum, and minimum values ​​of wind speed angle. It should be noted that the sliding window approach can capture the temporal dynamics of parameters. For example, the six-hour mean of standardized rainfall intensity is 0.18, the variance is 0.3, the maximum is 0.25, and the minimum is 0.12. Similarly, the statistical characteristics of temperature gradient and wind speed angle are calculated using a sliding window approach to generate a temporal feature set. This approach preserves data trends and fluctuations through smoothing along the time dimension.

[0031] Specifically, in step S2, principal component analysis is performed on the time series feature set using the PCA class of the sklearn library, a variance contribution rate threshold is set, and features whose variance contribution rates exceed the variance contribution rate threshold are retained as main feature components to obtain a streamlined feature set.

[0032] Specifically, the time series feature set contains multidimensional statistics, which increases computational complexity due to its high dimensionality. To this end, principal component analysis is performed using the PCA class of the sklearn library, with the variance contribution threshold set to 90%. Features with a variance contribution exceeding 90% are retained as the main feature components. Preferably, PCA maps high-dimensional features to a low-dimensional space through a linear transformation. For example, assuming that the time series feature set contains 12-dimensional statistical features (rainfall intensity, temperature gradient, and wind speed angle each correspond to four statistical features, which together make up 12-dimensional statistical features), after PCA analysis, the variance contribution of three features exceeds 90%, so these three features are retained as the main feature components. In one embodiment, the mean and variance of rainfall intensity have higher weights in the principal components, indicating that it contributes more to the data set. This simplified feature set significantly reduces dimensionality while retaining key information.

[0033] In one embodiment, the sliding window size can be adjusted based on specific needs. For example, for short-term heavy rainfall forecasts, the window can be narrowed to 3 hours with a step size of 30 minutes to capture finer-grained changes. For long-term trend analysis, the window can be expanded to 24 hours with a step size of 6 hours. This flexibility ensures the adaptability of the time series feature set.

[0034] It should be noted that the variance contribution threshold of PCA can also be adjusted according to the analysis objectives. For example, if higher accuracy is required, the threshold can be set to 95% to retain more principal components, while if it is not, it can be lowered to 85% to further simplify the calculation.

[0035] Optionally, in step S3, the operating parameters of the water supply system are obtained, and the time sequence of the reduced feature set and the operating parameters of the water supply system are aligned using a timestamp matching method; The Z-score standardization method is used to standardize the water supply system operating parameters after time series alignment, and the mean and standard deviation of each parameter are calculated to obtain the standardized water supply system operating parameters; The standardized water supply system operating parameters and the streamlined feature set after time series alignment are used as input. The long short-term memory network model is used for training. A fixed time step is set to construct the correlation relationship between the water supply system operating parameters and the streamlined feature set, and obtain the initial correlation prediction model.

[0036] For example, when obtaining water supply system operating parameters, these parameters include water pressure, flow rate, and pump station power. The timestamp matching method achieves temporal alignment by unifying the time resolution. Since the streamlined feature set has already been formatted with unified timestamps in step S1, it is only necessary to unify the timestamps of the water supply system operating parameters into the ISO8601 format to achieve the purpose of matching the timestamps of the streamlined feature set. This method ensures temporal consistency between the features in the streamlined feature set and the water supply system operating parameters, facilitating subsequent modeling.

[0037] In one possible implementation, Z-score normalization of water supply system operating parameters requires calculating the mean and standard deviation of each parameter. For example, the water pressure data has a mean of 2.4 bar and a standard deviation of 0.2 bar, while the flow rate has a mean of 95 m³ / h and a standard deviation of 10 m³ / h. After normalization, a water pressure of 2.5 bar is converted to a Z-score of 0.5, and a flow rate of 100 m³ / h is converted to 0.5. This normalization eliminates dimensional differences and makes different parameters comparable. It should be noted that the Z-score method can effectively handle outliers in the data and improve the stability of model training.

[0038] Specifically, the standardized water supply parameters and the reduced feature set were input into a long short-term memory (LSTM) model, with a time step of six hours. This model captures the dynamic correlation between the features in the reduced feature set and the water supply system operating parameters. For example, an increase in the mean rainfall intensity feature in the reduced feature set may lead to an increase in flow. The LSTM analyzes this temporal dependency through its memory cells.

[0039] It is worth noting that, in step S3, if the prediction error of the initial association model exceeds a preset threshold, the initial association model is optimized by reducing the learning rate and increasing the number of hidden layer nodes to obtain an optimized association prediction model.

[0040] During initial model training, it is assumed that the initial correlation model predicts water pressure, a water supply system operating parameter, with a mean squared prediction error of 0.1, which is higher than the preset threshold of 0.05. During model optimization, the learning rate can be reduced from 0.01 to 0.005, and the number of hidden layer nodes can be increased from 64 to 128. Optimally, the adjusted model can fit the data more accurately, reducing the error to 0.03, which is lower than the preset threshold of 0.05. This optimization method improves prediction accuracy by refining the learning step size and increasing the model capacity.

[0041] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. It is apparent to those skilled in the art that various changes, modifications, substitutions, and variations to these embodiments may be made without departing from the principles and spirit of the present invention, and these changes and modifications still fall within the scope of protection of the present invention.

Claims

1. A method for analyzing hydrological and weather data in a water supply system, characterized in that: The following steps are involved: S1: Acquire multi-source weather data, wherein the multi-source weather data includes rainfall, temperature, and wind speed parameters, and format the data using a unified timestamp and a unified coordinate system to obtain a standardized data set; S2: constructing a feature vector of the standardized data set, wherein the feature vector includes rainfall intensity, temperature change, and wind speed direction parameters, wherein the rainfall intensity is calculated by rainfall amount, the temperature change is calculated by temperature, and the wind speed direction is calculated by wind speed. The principal component analysis method is used to perform dimensionality reduction processing on the set of feature vectors to obtain a streamlined feature set; S3: Inputting the simplified feature set and the operating parameters of the water supply system into a long short-term memory network model, training the correlation between the simplified feature set and the operating parameters of the water supply system, and obtaining a correlation prediction model; S4: Acquire real-time multi-source weather data, input the real-time multi-source weather data into the association prediction model after the standardization processing of step S1 and the dimensionality reduction processing of step S2, and obtain the water supply system operation parameters output by the association prediction model.

2. The method for analyzing hydrological and weather data in a water supply system according to claim 1, characterized in that: In step S1, the steps of formatting using a unified timestamp include: A timestamp detection method is used to identify the timestamp format of the multi-source weather data, and the heterogeneous timestamps are converted into ISO8601 format to obtain a time-aligned dataset.

3. The method for analyzing hydrological and weather data in a water supply system according to claim 2, characterized in that: In step S1, the step of formatting using a unified coordinate system includes: converting the coordinate system of the time-aligned dataset into a WGS84 geographic coordinate system based on a coordinate conversion method to obtain a coordinate unified dataset.

4. The method for analyzing hydrological and weather data in a water supply system according to claim 3, characterized in that: In step S1, a linear interpolation method is used to fill in missing values ​​in the coordinate unified data set to obtain an interpolated data set; Outliers that exceed three times the standard deviation of the mean in the interpolated data set are removed by an outlier detection method to obtain a cleaned data set; The Z-score standardization method is used to convert the rainfall, temperature and wind speed parameters in the cleaned data set into a distribution with a mean of 0 and a variance of 1 to obtain a standardized data set.

5. The method for analyzing hydrological and weather data in a water supply system according to claim 4, characterized in that: In step S2, the standardized rainfall intensity is calculated based on the rainfall in the standardized data set, the standardized temperature gradient is calculated based on the temperature in the standardized data set, and the standardized wind speed angle is calculated based on the wind speed in the standardized data set to form an initial parameter set.

6. The method for analyzing hydrological and weather data in a water supply system according to claim 5, characterized in that: In step S2, the sliding window method is used to process the initial parameter set, the window size is set to a fixed length, the step size is set to a unit interval, and the rolling function of the pandas library is used to calculate the mean, variance, maximum value and minimum value of the rainfall intensity, temperature gradient and wind speed angle in the initial parameter set to obtain a time series feature set.

7. The method for analyzing hydrological and weather data in a water supply system according to claim 6, characterized in that: In step S2, principal component analysis is performed on the time series feature set using the PCA class of the sklearn library, a variance contribution rate threshold is set, and features whose variance contribution rates exceed the variance contribution rate threshold are retained as main feature components to obtain a streamlined feature set.

8. The method for analyzing hydrological and weather data in a water supply system according to claim 7, characterized in that: In step S3, the operating parameters of the water supply system are obtained, and the time sequence of the reduced feature set and the operating parameters of the water supply system are aligned using a timestamp matching method; The Z-score standardization method is used to standardize the water supply system operating parameters after time series alignment, and the mean and standard deviation of each parameter are calculated to obtain the standardized water supply system operating parameters; The standardized water supply system operating parameters and the streamlined feature set after time series alignment are used as input. The long short-term memory network model is used for training. A fixed time step is set to construct the correlation relationship between the water supply system operating parameters and the streamlined feature set, and obtain the initial correlation prediction model.

9. The method for analyzing hydrological and weather data in a water supply system according to claim 8, characterized in that: In step S3, if the prediction error of the initial association model exceeds a preset threshold, the initial association model is optimized by reducing the learning rate and increasing the number of hidden layer nodes to obtain an optimized association prediction model.

Images