A multi-source collaborative energy supply intelligent water quality monitoring method, medium and system

Abstract

The application provides a kind of multi-source collaborative energy supply intelligent water quality monitoring method, medium and system, belong to water quality monitoring technical field, the application is by calling the energy scheduling model of multiple stages to calculate the power generation prediction value and load demand prediction value of future period, output each period's power supply switching instruction and power distribution scheme, control solar panel, small wind turbine generator and the power supply priority of battery, monitoring sensor collects chemical oxygen demand, total phosphorus concentration and total nitrogen concentration data, auxiliary sensor synchronously collects water temperature, pH value and turbidity data, edge computing module calls multi-parameter cross correction model to carry out interference suppression processing to monitoring data, calculates adaptive sampling adjustment factor according to the data after correction dynamically adjusts sampling interval, solve the technical problem that the continuity of monitoring data caused by insufficient energy supply stability of field water quality monitoring system in no city power environment.

Description

Technical Field

[0001] This invention belongs to the field of water quality monitoring technology, and specifically relates to an intelligent water quality monitoring method, medium and system with multi-source collaborative power supply. Background Technology

[0002] Water quality monitoring is a crucial means of environmental protection. Traditional water quality monitoring systems primarily rely on mains power or a single solar power source for long-term continuous monitoring. In remote water areas, nature reserves, and other areas without mains power coverage, existing monitoring devices typically use solar panels as the main energy source, supplemented by batteries for energy storage, to maintain the continuous operation of monitoring sensors, data acquisition modules, and communication modules. However, the efficiency of a single solar power source drops sharply during cloudy or rainy weather or at night, and the battery power is rapidly depleted, causing the monitoring equipment to frequently enter low-power mode or even shut down due to power outages, resulting in significant gaps in monitoring data. Although some solutions attempt to increase battery capacity or expand the solar panel area to extend the battery life, these measures significantly increase system costs and installation difficulty, and still cannot cope with several consecutive days of severe weather. In other words, existing technologies suffer from the technical problem of insufficient power supply stability in environments without mains power, leading to poor continuity of monitoring data. Summary of the Invention

[0003] In view of this, the present invention provides a smart water quality monitoring method, medium and system with multi-source collaborative power supply, which can solve the technical problem of poor continuity of monitoring data caused by insufficient power supply stability of field water quality monitoring systems in environments without mains power.

[0004] The invention is implemented as follows: The first aspect of the invention provides an intelligent water quality monitoring method with multi-source collaborative power supply. Environmental sensors collect data on light intensity, wind speed, and ambient temperature and transmit this data to an energy management unit. The energy management unit identifies the current operating condition based on the combination of light intensity and wind speed, calls a multi-stage energy scheduling model to calculate predicted power generation and load demand for future periods, outputs power switching commands and power allocation schemes for each period, and controls the power supply priority of solar panels, small wind turbines, and batteries. This achieves complementary advantages of solar and wind energy under different meteorological conditions, ensuring a stable and reliable power supply for the monitoring system under various meteorological conditions. Monitoring sensors collect data on chemical oxygen demand (COD), total phosphorus concentration (TP), and total nitrogen concentration (TN) in the water body. Auxiliary sensors simultaneously collect data on water temperature, pH, and turbidity. An edge computing module calls a multi-parameter cross-calibration model to suppress interference in the COD, TTP, and TN data. Based on the calibrated data, an adaptive sampling adjustment factor is calculated to dynamically adjust the sampling interval. The calibrated data is compressed and encoded before being uploaded to a mobile terminal via a wireless communication module.

[0005] Specifically, the identification of the operating condition type involves synchronously recording data on light intensity, wind speed, solar panel power generation, and small wind turbine power generation every hour within a monitoring period of no less than 12 months. The experimental data is then classified in two dimensions according to light intensity and wind speed. The average power generation of the solar panel and the average power generation of the small wind turbine under different combinations of light intensity and wind speed ranges are statistically analyzed. Based on the comparison relationship of the average power generation, the following operating conditions are defined: strong light and weak wind, weak light and strong wind, no wind and solar power, and mixed operating conditions.

[0006] Specifically, the rule for identifying the working condition type is that when the light intensity is greater than... Furthermore, a wind speed less than 3 m / s is identified as a strong light and weak wind condition, and a light intensity less than... And the wind speed is greater than When the light intensity is less than 100 pm, the condition is identified as low light and strong wind. Furthermore, when the wind speed is less than 3 m / s, it is identified as a windless and solar-free operating condition; other combinations of light intensity and wind speed are identified as mixed operating conditions.

[0007] Specifically, the multi-stage energy dispatch model divides the 24-hour monitoring cycle into 48 decision stages, each lasting 30 minutes. The state variables include the current remaining battery power, the real-time power generation of the solar panels, and the real-time power generation of the small wind turbine. The action variables include solar priority power supply mode, wind priority power supply mode, and hybrid power supply mode. The value function includes a power supply reliability penalty term and a battery loss cost term.

[0008] Specifically, the training dataset for the multi-stage energy dispatch model is established by selecting no less than 365 days of historical meteorological and electricity consumption data, sampling the daily light intensity curve, wind speed curve, and load power curve at 30-minute intervals to form time series samples, labeling the optimal power supply mode and actual battery state changes for each time period, and constructing a training sample set containing state transition trajectories and cumulative benefits.

[0009] Specifically, the multi-stage energy dispatch model training involves initializing the value function parameters for each decision stage, performing backward recursive calculations starting from the last decision stage, iterating through all state variable values ​​and action variable combinations for each decision stage, calculating the sum of the value of the current decision stage and the value of subsequent decision stages, selecting the action variable that minimizes the total value as the optimal strategy, and updating the value function parameters.

[0010] The weighting coefficient of the power supply reliability penalty item is determined based on the remaining battery power percentage. When the remaining battery power percentage is greater than or equal to 70%, the weighting coefficient of the power supply reliability penalty item is 0.3; when the remaining battery power percentage is between 40% and 70%, the weighting coefficient of the power supply reliability penalty item is 0.5; and when the remaining battery power percentage is less than 40%, the weighting coefficient of the power supply reliability penalty item is 0.8.

[0011] Specifically, the multi-parameter cross-calibration model involves an input layer that receives six parameters: raw values ​​of chemical oxygen demand (COD), total phosphorus concentration (TP), total nitrogen concentration (TN), water temperature, pH, and turbidity. The hidden layer uses partial least squares regression to extract principal component features. Singular value decomposition projects the six-dimensional input data into a three-dimensional principal component space. The output layer generates calibrated COD, TP, and TN data.

[0012] Specifically, the training dataset for the multi-parameter cross-calibration model is established by configuring no fewer than 500 sets of standard water samples with known concentrations under laboratory conditions, collecting raw data using monitoring sensors and auxiliary sensors under different water temperatures, pH levels, and turbidity conditions, and recording the deviations between the raw values ​​and the known concentration values ​​of the standard water samples to form training samples.

[0013] Specifically, the training of the multi-parameter cross-calibration model involves dividing the training samples into a training set and a validation set in a 7:3 ratio, constructing a redundant measurement matrix between the input parameters and the output calibration values ​​using the training set, extracting the principal component vectors corresponding to the top three largest singular values ​​by singular value decomposition of the redundant measurement matrix, establishing a linear mapping relationship from the principal component space to the calibration value space, and optimizing the mapping coefficients using the least squares method.

[0014] Specifically, the calculation of the adaptive sampling adjustment factor involves extracting the corrected chemical oxygen demand (COD), total phosphorus (TP), and total nitrogen (TN) data from the most recent 10 sampling periods, calculating the standard deviation of each data point, normalizing them to the 0-1 range, and taking the average value as the data fluctuation characteristic value. It also involves calculating the absolute value of the difference between adjacent sampling periods, normalizing it to the 0-1 range, and taking the maximum value as the gradient change characteristic value. Finally, the data fluctuation characteristic value and the gradient change characteristic value are weighted and summed using weighting coefficients of 0.4 and 0.6, respectively.

[0015] Specifically, the sampling interval adjustment means that when the adaptive sampling adjustment factor is between 0 and 0.3, the sampling interval is extended to twice the base sampling interval; when the adaptive sampling adjustment factor is between 0.3 and 0.7, the base sampling interval is maintained; and when the adaptive sampling adjustment factor is between 0.7 and 1, the sampling interval is shortened to 0.5 times the base sampling interval.

[0016] A second aspect of the present invention provides a computer-readable storage medium storing program instructions, which, when executed in a computer, are used to perform the above-described intelligent water quality monitoring method with multi-source collaborative power supply.

[0017] A third aspect of this invention provides a multi-source collaborative power supply intelligent water quality monitoring system, comprising the aforementioned computer-readable storage medium and a multi-source collaborative power supply intelligent water quality monitoring device. The multi-source collaborative power supply intelligent water quality monitoring device includes a support column, a solar panel, a small wind turbine, an environmental sensor, a channel opening, a battery, a cabinet, wiring, a monitoring module, an energy management unit, an edge computing module, and a wireless communication module. The support column is a hollow structure fixed to the ground. The solar panel and the small wind turbine are connected through wiring inside the support column. The cabinet houses the edge computing module, the energy management unit, the battery, and the wireless communication module. The monitoring module is placed in the water body being monitored and connected to the edge computing module through wiring. The edge computing module is a microcontroller or a small computer. The computer-readable storage medium is disposed within the edge computing module, and the edge computing module contains a microprocessor that executes program instructions stored in the computer-readable storage medium.

[0018] Furthermore, the monitoring module includes a monitoring sensor and an auxiliary sensor. The monitoring sensor collects data on chemical oxygen demand, total phosphorus concentration, and total nitrogen concentration in the water body, while the auxiliary sensor simultaneously collects data on water temperature, pH, and turbidity.

[0019] Furthermore, the environmental sensor is fixed on the support column and connected to the energy management unit via wiring. The energy management unit is connected to the control terminals of the solar panel, the small wind turbine, and the battery via wiring.

[0020] Furthermore, the solar panel and the small wind turbine are connected to the battery via a channel port, and the edge computing module is connected to the wireless communication module via a line. When the corrected chemical oxygen demand data, corrected total phosphorus concentration data, or corrected total nitrogen concentration data exceed the preset threshold range, the abnormal alarm function of the mobile terminal is triggered.

[0021] This invention constructs a multi-source collaborative energy supply mechanism, combining solar panels, small wind turbines, and batteries to form a complementary power supply architecture. Based on real-time changes in sunlight intensity and wind speed, it identifies four typical operating conditions: strong sunlight and weak wind, weak sunlight and strong wind, no wind and solar power, and mixed conditions. It then invokes a multi-stage energy dispatch model to predict future power generation capacity and load demand, outputting power switching commands and power allocation schemes for each time period, achieving complementary advantages of solar and wind energy under different meteorological conditions. When sunlight is sufficient but wind is insufficient, solar panels are prioritized for power supply; when sunlight is insufficient but wind is sufficient, small wind turbines are prioritized for power supply; and when both sunlight and wind are insufficient, batteries provide backup power. By dynamically adjusting the power supply mode and power allocation ratio, the monitoring system can obtain a stable and reliable power supply under various meteorological conditions, avoiding power outages caused by fluctuations in a single energy source. In summary, this invention solves the technical problem mentioned in the background art: insufficient power supply stability in field water quality monitoring systems in environments without mains power, leading to poor continuity of monitoring data. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method of the present invention.

[0023] Figure 2 This is a schematic diagram of the composition of an intelligent water quality monitoring device with multi-source collaborative power supply involved in the embodiment.

[0024] Figure 3 This is a schematic diagram of the power supply section of the intelligent water quality monitoring device with multi-source collaborative power supply involved in the embodiment.

[0025] Figure 4 This is a schematic diagram of the functional box of the intelligent water quality monitoring device with multi-source collaborative power supply involved in the embodiment.

[0026] Figure 5 This is a schematic diagram of the intelligent water quality monitoring device with multi-source collaborative power supply involved in the embodiment.

[0027] Figure 6 This is a graph showing the changes in sunlight intensity and wind speed throughout the day on January 10th.

[0028] Figure 7 This is a curve comparing the chemical oxygen demand (COD) data before and after correction.

[0029] Figure 8 This is a graph showing the changes in the adaptive sampling adjustment factor from January 10th to 16th.

[0030] Figure 9 This is a time series diagram of abnormal water quality parameters.

[0031] The reference numerals in the attached drawings are explained as follows: (1) solar panel, (2) small wind turbine, (3) support column, (4) environmental sensor, (5) passageway, (6) battery, (7) cabinet, (8) wiring, (9) monitoring module, (10) monitoring sensor, (11) auxiliary sensor, (12) energy management unit, (13) edge computing module. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0033] like Figure 1 The diagram shown is a flowchart of a multi-source collaborative power supply intelligent water quality monitoring method provided by the first aspect of the present invention. This method includes the following steps: S1. The environmental sensor collects data on light intensity, wind speed and ambient temperature, and transmits the data to the energy management unit. The energy management unit identifies the current operating condition type based on the combination of light intensity and wind speed. S2. The energy management unit calls the multi-stage energy dispatch model to calculate the predicted power generation and load demand for future periods, outputs power switching instructions and power allocation schemes for each period, and controls the power supply priority of solar panels, small wind turbines and batteries. S3. The monitoring sensor collects data on chemical oxygen demand, total phosphorus concentration, and total nitrogen concentration in the water body, and the auxiliary sensor simultaneously collects data on water temperature, pH, and turbidity. The data on chemical oxygen demand, total phosphorus concentration, and total nitrogen concentration, as well as water temperature, pH, and turbidity, are transmitted to the edge computing module. S4. The edge computing module calls the multi-parameter cross-calibration model to perform interference suppression processing on the chemical oxygen demand, total phosphorus concentration and total nitrogen concentration data, and outputs the calibrated chemical oxygen demand data, calibrated total phosphorus concentration data and calibrated total nitrogen concentration data. S5. The edge computing module calculates an adaptive sampling adjustment factor based on the corrected chemical oxygen demand data, corrected total phosphorus concentration data, and corrected total nitrogen concentration data. When the adaptive sampling adjustment factor is in the range of [0, 0.3), the sampling interval is extended to twice the baseline sampling interval. When the adaptive sampling adjustment factor is in the range of [0.3, 0.7), the baseline sampling interval is maintained. When the adaptive sampling adjustment factor is in the range of [0.7, 1], the sampling interval is shortened to 0.5 times the baseline sampling interval. S6. The edge computing module compresses and encodes the corrected chemical oxygen demand (COD), total phosphorus (TP) concentration, and total nitrogen (N) concentration data and then uploads them to the mobile terminal via the wireless communication module. When the corrected COD, TP, or N exceeds the preset threshold range, the mobile terminal's abnormal alarm function is triggered.

[0034] The above method employs a multi-source collaborative power supply intelligent water quality monitoring device for data collection, comprising a support column, solar panels, a small wind turbine, environmental sensors, a channel opening, a battery, a cabinet, wiring, a monitoring module, an energy management unit, an edge computing module, and a wireless communication module. The support column is a hollow structure, fixed underground, used to secure the solar panels, the small wind turbine, and the cabinet. The solar panels and the small wind turbine are connected via wiring inside the support column and via wiring to the battery through the channel opening. The lower part of the cabinet has a channel opening, and the edge computing module, energy management unit, battery, and wireless communication module are housed inside the cabinet. The monitoring module is placed in the water body being monitored and includes monitoring sensors and auxiliary sensors, connected to the edge computing module via wiring. The environmental sensors are fixed to the support column and connected to the energy management unit via wiring. The energy management unit is connected to the control terminals of the solar panels, the small wind turbine, and the battery via wiring. The edge computing module is connected to the wireless communication module via wiring.

[0035] The identification rules for the operating condition types were determined through experiments and experimental data analysis. The experimental steps were as follows: During a monitoring period of no less than 12 months, data on light intensity, wind speed, solar panel power generation, and small wind turbine power generation were recorded synchronously every hour, collecting no less than 8760 sets of experimental data. The experimental data analysis steps were as follows: The experimental data were classified in two dimensions according to light intensity and wind speed; the average power generation of solar panels and small wind turbines under different combinations of light intensity and wind speed ranges were statistically analyzed; when the average power generation of solar panels was greater than that of small wind turbines... The combination of sunlight intensity and wind speed when the average power output of the generator is more than three times that of the average power output of the solar panel is defined as a strong light and weak wind condition. The combination of sunlight intensity and wind speed when the average power output of the small wind turbine is more than three times that of the average power output of the solar panel is defined as a weak light and strong wind condition. The combination of sunlight intensity and wind speed when both the average power output of the solar panel and the average power output of the small wind turbine are less than 20% of their rated power is defined as a no-wind / solar condition. All other combinations of sunlight intensity and wind speed are defined as mixed conditions. Based on the analysis of experimental data, the identification rule for the condition type is: when the sunlight intensity is greater than 600... Furthermore, a wind speed less than 3 m / s is identified as a strong light and weak wind condition, and a light intensity less than 200 ppm is also considered a strong light and weak wind condition. Furthermore, wind speeds greater than 5 m / s are identified as low-light, high-wind conditions, and light intensity less than 200 ppm is also considered a low-light, high-wind condition. Furthermore, when the wind speed is less than 3 m / s, it is identified as a windless and solar-free operating condition; other combinations of light intensity and wind speed are identified as mixed operating conditions.

[0036] The structure of the multi-stage energy dispatch model is as follows: the 24-hour monitoring cycle is divided into 48 decision stages, each lasting 30 minutes. Each decision stage includes three components: state variables, action variables, and a value function. The state variables include the current remaining battery power, the real-time power generation of the solar panels, and the real-time power generation of the small wind turbine. The action variables include solar priority power supply mode, wind priority power supply mode, and hybrid power supply mode. The value function includes a power supply reliability penalty term and a battery loss cost term. The power supply reliability penalty term is calculated based on the duration of unmet load demand, and the battery loss cost term is calculated based on the battery charge / discharge depth and cycle count.

[0037] The steps for establishing the training dataset of the multi-stage energy dispatch model are as follows: select no less than 365 days of historical meteorological data and electricity consumption data, sample the daily light intensity curve, wind speed curve and load power curve at 30-minute intervals to form time series samples, label the optimal power supply mode and actual battery state changes for each time period, and construct a training sample set containing state transition trajectories and cumulative benefits.

[0038] The training steps of the multi-stage energy dispatch model are as follows: initialize the value function parameters of each decision stage, perform backward recursive calculation starting from the last decision stage, traverse all possible state variable values ​​and action variable combinations in each decision stage, calculate the sum of the value of the current decision stage and the value of the subsequent decision stages, select the action variable that minimizes the total value as the optimal strategy and update the value function parameters, repeat the recursion until the first decision stage, and complete the solution of the global optimal strategy.

[0039] The structure of the multi-parameter cross-calibration model is as follows: the input layer receives six parameters: raw values ​​of chemical oxygen demand (COD), total phosphorus concentration (TP), total nitrogen concentration (TN), water temperature, pH, and turbidity. The hidden layer uses partial least squares regression to extract principal component features. The output layer generates calibrated COD, TP, and TN data. The hidden layer projects the six-dimensional input data into a three-dimensional principal component space through singular value decomposition to suppress the cross-interference between different parameters.

[0040] The steps for establishing the training dataset of the multi-parameter cross-calibration model are as follows: Under laboratory conditions, no less than 500 sets of standard water samples with known concentrations are prepared. Each set of standard water samples contains chemical oxygen demand (COD), total phosphorus (TP), and total nitrogen (TN) with different concentration combinations. Under different water temperature, pH, and turbidity conditions, monitoring sensors and auxiliary sensors are used to collect the original values ​​of COD, TN, TN, water temperature, pH, and turbidity. The deviations between the original values ​​of COD, TN, and TN and the known concentrations of the standard water samples are recorded to form the training samples.

[0041] The training steps of the multi-parameter cross-calibration model are as follows: the training samples are divided into a training set and a validation set in a 7:3 ratio; a redundant measurement matrix between the input parameters and the output calibration values ​​is constructed using the training set; the principal component vectors corresponding to the top three largest singular values ​​are extracted by singular value decomposition of the redundant measurement matrix; a linear mapping relationship is established from the principal component space to the calibration value space; the mapping coefficients are optimized by the least squares method; the calibration accuracy is evaluated on the validation set and the number of principal components is adjusted until the calibration error is less than 5%.

[0042] The calculation steps of the adaptive sampling adjustment factor are as follows: extract the corrected chemical oxygen demand (COD), total phosphorus (TP), and total nitrogen (TN) concentration data from the most recent 10 sampling periods; calculate the standard deviations of the corrected COD, TP, and TN data respectively; normalize these standard deviations to the 0-1 range and take the average value as the data fluctuation characteristic value; calculate the absolute values ​​of the differences in corrected COD, TP, and TN data between adjacent sampling periods; normalize these differences to the 0-1 range and take the maximum value as the gradient change characteristic value; and weight the data fluctuation characteristic value and the gradient change characteristic value with weighting coefficients of 0.4 and 0.6 to obtain the adaptive sampling adjustment factor.

[0043] The weight coefficient of the power supply reliability penalty term in the multi-stage energy dispatch model is determined based on the remaining battery power percentage. It is usually based on empirical values, namely: when the remaining battery power percentage is greater than or equal to 70%, the weight coefficient of the power supply reliability penalty term is 0.3; when the remaining battery power percentage is between [40% and 70%), the weight coefficient of the power supply reliability penalty term is 0.5; and when the remaining battery power percentage is less than 40%, the weight coefficient of the power supply reliability penalty term is 0.8.

[0044] The baseline sampling interval is the data acquisition time interval of the monitoring sensor under normal operating conditions, ranging from 10 to 30 minutes. The chemical oxygen demand (COD) is the mass concentration of oxygen consumed when organic matter and reducing inorganic matter in water are oxidized by chemical oxidants. The partial least squares regression algorithm is a multivariate statistical data analysis method combining multiple linear regression, principal component analysis, and canonical correlation analysis. It achieves regression modeling between independent and dependent variables by extracting the comprehensive variable with the strongest explanatory power for the dependent variable. The singular value decomposition (SVD) is a mathematical operation that decomposes a matrix into the product of three matrices. The decomposed matrix contains the main feature information of the original matrix and is used for dimensionality reduction and noise reduction. The raw COD value is the uncorrected COD data directly measured by the monitoring sensor. The raw total phosphorus concentration is the uncorrected total phosphorus concentration data directly measured by the monitoring sensor. The raw total nitrogen concentration is the uncorrected total nitrogen concentration data directly measured by the monitoring sensor. The predicted power generation is the future solar panel power generation and small wind turbine power generation predicted by a multi-stage energy dispatch model based on light intensity and wind speed data collected by environmental sensors. The load demand forecast is the total power demand of the monitoring module, edge computing module, and wireless communication module for future periods, predicted by the multi-stage energy dispatch model based on historical electricity consumption data. The power switching command is a control signal output by the multi-stage energy dispatch model, used to instruct the energy management unit to switch between solar-priority power supply mode, wind-priority power supply mode, and hybrid power supply mode. The power allocation scheme is the output power allocation ratio of solar panels, small wind turbines, and batteries for each time period, output by the multi-stage energy dispatch model.

[0045] The specific implementation methods of the above steps are described in detail below.

[0046] The specific implementation of step S1 is as follows: The environmental sensor collects data on light intensity, wind speed, and ambient temperature through a photoresistor and an anemometer, respectively. The collection frequency is set to once every 5 minutes, and the unit of the collected light intensity data is [missing information]. Wind speed data is in m / s, and ambient temperature data is in °C. After data collection, the light intensity, wind speed, and ambient temperature data are packaged and sent to the energy management unit via a digital signal transmission protocol. Upon receiving the data, the energy management unit immediately identifies the operating condition. The identification process is based on a preset threshold judgment logic, first determining whether the light intensity is greater than 600. If the conditions are met, further determine if the wind speed is less than 3 m / s. If both conditions are met, it is identified as a strong light and weak wind condition. If the light intensity is less than 200... The system checks if the wind speed is greater than 5 m / s. If it is, it is identified as a low-light, high-wind condition. If the light intensity is less than 200... Furthermore, if the wind speed is less than 3 m / s, it is identified as a windless and solar-free operating condition, and the other combinations are identified as mixed operating conditions. The threshold judgment logic is based on the statistical analysis of experimental data, which can accurately distinguish environmental conditions dominated by different energy supplies and provide a basis for decision-making in subsequent energy dispatch.

[0047] The specific implementation of step S2 is as follows: The energy management unit inputs the current operating condition type, remaining battery power, real-time solar panel power generation, and real-time small wind turbine power generation as input parameters into the multi-stage energy dispatch model. The multi-stage energy dispatch model is solved based on a dynamic programming algorithm. After dividing the 24-hour monitoring cycle into 48 decision stages, backward recursive calculation is performed starting from the last decision stage. In each decision stage, three action variables are traversed: solar priority power supply mode, wind priority power supply mode, and hybrid power supply mode. The power supply reliability penalty term and battery loss cost term are calculated for each action variable. The penalty is calculated by multiplying the duration of unmet load demand by a penalty coefficient. The battery loss cost is calculated by multiplying the battery charge / discharge depth by a cycle count conversion factor. The total value is obtained by adding the value of the current decision stage to the value of the subsequent decision stages. The action variable that minimizes the total value is selected as the optimal strategy. After the recursion is completed, the power switching instructions and power allocation schemes for each time period are output. The power switching instructions control the energy management unit to switch between different power supply modes. The power allocation scheme guides the solar panels, small wind turbines, and batteries to output power according to the calculated ratio. The dynamic programming algorithm ensures maximum energy utilization efficiency through global optimization.

[0048] The specific implementation of step S3 is as follows: The monitoring sensor collects data on chemical oxygen demand (COD), total phosphorus concentration (TP), and total nitrogen concentration (TN) in the water body using electrochemical analysis. The auxiliary sensor simultaneously collects water temperature, pH, and turbidity data using a temperature probe, pH electrode, and turbidity optical sensor. The collection frequency is set according to the current sampling interval. The unit for COD data is mg / L, the unit for TP and TN data is mg / L, the unit for water temperature data is °C, the unit for pH data is dimensionless, and the unit for turbidity data is NTU. After the monitoring sensor and auxiliary sensor complete the data collection, the signals are converted into digital signals by an analog-to-digital converter and transmitted to the edge computing module via a line. After receiving the COD, TP, and TN data, as well as the water temperature, pH, and turbidity data, the edge computing module performs data buffering to prepare for subsequent correction processing. The synchronous acquisition mechanism ensures that the water quality parameters are time-aligned with the environmental parameters, avoiding correction errors caused by time deviations.

[0049] The specific implementation of step S4 is as follows: The edge computing module uses the received chemical oxygen demand (COD), total phosphorus (TP), and total nitrogen (TN) concentration data as the original values ​​of COD, TTP, and TN concentration, and together with the water temperature, pH, and turbidity values, forms a 6-dimensional input vector, which is then fed into the multi-parameter cross-calibration model. The multi-parameter cross-calibration model performs interference suppression processing based on the partial least squares regression algorithm and the singular value decomposition principle. After receiving the 6-dimensional input vector, the input layer passes it to the hidden layer. The hidden layer first constructs a redundant measurement matrix between the input parameters and the output calibration values, and then performs singular value decomposition to extract the redundant measurement matrix. The principal component vectors corresponding to the first three largest singular values ​​project the 6-dimensional input vector into a 3-dimensional principal component space, suppressing the cross-influence of temperature fluctuations, pH changes, and turbidity on the measurement of chemical oxygen demand (COD), total phosphorus (TP), and total nitrogen (TN). The output layer calculates the corrected COD, TTP, and TN data based on the linear mapping relationship from the principal component space to the correction value space. The partial least squares regression algorithm achieves accurate correction by extracting the comprehensive variable with the strongest explanatory power for the dependent variable, and singular value decomposition eliminates noise interference through dimensionality reduction. The combination of these two methods significantly improves the accuracy of water quality parameter measurements.

[0050] The specific implementation of step S5 is as follows: The edge computing module extracts the corrected chemical oxygen demand (COD), total phosphorus (TP), and total nitrogen (TN) concentration data from the data cache for the most recent 10 sampling periods. It calculates the standard deviation of each of the three data types using the sample standard deviation formula. The standard deviations of the corrected COD, TP, and TN data are normalized by dividing each by its respective historical maximum standard deviation. The normalized values ​​range from 0 to 1. The arithmetic mean of the three normalized standard deviations is taken as the data fluctuation characteristic value, reflecting the stability of the water quality parameters. Subsequently, the absolute values ​​of the differences in corrected COD, TP, and TN data between adjacent sampling periods are calculated. These absolute values ​​are then normalized by dividing each by its respective historical maximum difference. The maximum value of the three normalized differences is taken as the standard deviation. The gradient change feature value reflects the degree of abrupt change in water quality parameters. The data fluctuation feature value and the gradient change feature value are weighted and summed with weighting coefficients of 0.4 and 0.6 to obtain the adaptive sampling adjustment factor. The weighting coefficients are set based on the principle that the importance of gradient change for pollution event detection is higher than that of data fluctuation. The adaptive sampling adjustment factor comprehensively evaluates the stability and trend of water quality. The sampling interval is adjusted according to the range to which the adaptive sampling adjustment factor belongs. When the adaptive sampling adjustment factor belongs to [0, 0.3), it indicates that the water quality is stable, and the sampling interval is extended to twice the baseline sampling interval to reduce energy consumption. When the adaptive sampling adjustment factor belongs to [0.3, 0.7), it indicates that the water quality fluctuates normally, and the baseline sampling interval is maintained. When the adaptive sampling adjustment factor belongs to [0.7, 1], it indicates that the water quality changes abnormally, and the sampling interval is shortened to 0.5 times the baseline sampling interval to increase the monitoring density. The adaptive adjustment mechanism balances energy consumption and monitoring accuracy.

[0051] The specific implementation of step S6 is as follows: The edge computing module compresses and encodes the corrected chemical oxygen demand (COD), total phosphorus (TP) concentration, and total nitrogen (TN) concentration data. The compression encoding uses a lossless compression algorithm to reduce the amount of data transmitted. After encoding, the data is uploaded to the mobile terminal via the wireless communication module. The upload process uses a low-power wireless transmission protocol. After receiving the data, the mobile terminal decodes and restores it. Before uploading the data, the edge computing module determines whether the corrected COD data exceeds a preset threshold range. The preset threshold range is COD less than 15 mg / L or greater than 40 mg / L. At the same time, it determines whether the corrected TP concentration data exceeds the range of 0.02 mg / L to 0.2 mg / L and whether the corrected TN concentration data exceeds the range of 0.2 mg / L to 2.0 mg / L. If any parameter exceeds the preset threshold range, the abnormal alarm function of the mobile terminal is triggered. The mobile terminal reminds the user to check the abnormal data through sound and vibration. The preset threshold range is set based on the national surface water environmental quality standards to ensure timely detection of water pollution events.

[0052] It should be noted that the key technical ideas of this invention include dynamic programming optimization of a multi-stage energy dispatch model, interference suppression of a multi-parameter cross-calibration model, and an adaptive sampling adjustment mechanism. The multi-stage energy dispatch model divides the 24-hour monitoring cycle into 48 decision stages and uses a dynamic programming algorithm for global optimization, overcoming the limitations of traditional fixed power supply strategies in coping with environmental fluctuations. It solves the optimal power switching strategy and power allocation scheme for each stage through backward recursion, minimizing battery loss costs while ensuring power supply reliability. Compared to a single energy supply method, this extends battery life and improves system endurance. The multi-parameter cross-calibration model combines partial least squares regression and singular value decomposition principles to project 6-dimensional input parameters into a 3-dimensional principal component space, effectively suppressing the cross-interference of temperature, pH, and turbidity on the measurement of chemical oxygen demand, total phosphorus concentration, and total nitrogen concentration. This solves the problem of measurement deviation caused by environmental factors in traditional single sensors, significantly improving the accuracy and reliability of water quality parameter measurements. The adaptive sampling adjustment mechanism dynamically adjusts the sampling interval based on data fluctuation and gradient change characteristics. It reduces the sampling frequency during periods of stable water quality to save energy, and increases the sampling density to capture pollution events during periods of abnormal water quality changes. This overcomes the drawbacks of traditional fixed sampling strategies, such as high energy consumption or slow response, achieving a dynamic balance between energy consumption and monitoring accuracy. The three technical approaches work synergistically to construct a complete water quality monitoring closed loop. A multi-stage energy scheduling model ensures power supply stability, providing energy security for long-term monitoring; a multi-parameter cross-calibration model provides high-precision water quality data, offering a reliable basis for decision-making; and the adaptive sampling adjustment mechanism optimizes energy allocation based on water quality conditions, forming a positive feedback loop. These three elements support each other, achieving a balance between efficient energy utilization and accurate water quality monitoring. Compared to existing technologies, this approach offers significant advantages in power supply stability, measurement accuracy, and system endurance.

[0053] It should be noted that this invention also solves the following technical problem: the problem of decreased measurement accuracy of water quality monitoring sensors due to interference from environmental factors such as water temperature, pH, and turbidity in complex field environments. This invention utilizes an edge computing module to call a multi-parameter cross-calibration model. The input layer receives six parameters: raw values ​​of chemical oxygen demand (COD), total phosphorus concentration (TPC), total nitrogen concentration (TN), water temperature, pH, and turbidity. The hidden layer uses partial least squares regression to extract principal component features. Singular value decomposition projects the six-dimensional input data into a three-dimensional principal component space, suppressing cross-interference between different parameters. The output layer generates calibrated COD, TPC, and TN data, effectively eliminating the impact of environmental factors on monitoring accuracy and ensuring the accuracy and reliability of water quality monitoring data.

[0054] A second aspect of the present invention provides a computer-readable storage medium storing program instructions, which, when executed in a computer, are used to perform the above-described intelligent water quality monitoring method with multi-source collaborative power supply.

[0055] A third aspect of this invention provides a multi-source collaborative power supply intelligent water quality monitoring system, comprising the aforementioned computer-readable storage medium and a multi-source collaborative power supply intelligent water quality monitoring device. The multi-source collaborative power supply intelligent water quality monitoring device includes a support pillar, a solar panel, a small wind turbine, an environmental sensor, a channel opening, a battery, a cabinet, wiring, a monitoring module, an energy management unit, an edge computing module, and a wireless communication module. The support pillar is a hollow structure fixed underground. The solar panel and the small wind turbine are connected through wiring inside the support pillar. The cabinet houses the edge computing module, the energy management unit, the battery, and the wireless communication module. The monitoring module is placed in the water body being monitored and connected to the edge computing module through wiring. The edge computing module is a microcontroller or a small computer. The computer-readable storage medium is disposed within the edge computing module, and the edge computing module contains a microprocessor that executes program instructions stored in the computer-readable storage medium. Further, the monitoring module includes a monitoring sensor and an auxiliary sensor. The monitoring sensor collects data on chemical oxygen demand, total phosphorus concentration, and total nitrogen concentration in the water body, while the auxiliary sensor simultaneously collects data on water temperature, pH, and turbidity. Furthermore, the environmental sensor is fixed to the support pillar and connected to the energy management unit via wiring. The energy management unit is connected to the control terminals of the solar panel, the small wind turbine, and the battery via wiring. Furthermore, the solar panel and the small wind turbine are connected to the battery via a channel, and the edge computing module is connected to the wireless communication module via wiring. When the corrected chemical oxygen demand (COD), total phosphorus concentration (TP), or total nitrogen concentration (TN) data exceeds a preset threshold range, an abnormal alarm function on the mobile terminal is triggered.

[0056] Specifically, the principle of this invention is as follows: This invention collects real-time data on light intensity and wind speed using environmental sensors. The energy management unit categorizes the operating conditions into four types based on the combination of light intensity and wind speed: strong light and weak wind, weak light and strong wind, no wind and solar, and a mixed type. This classification accurately reflects the complementary characteristics of solar and wind energy. The multi-stage energy dispatch model divides the 24-hour monitoring cycle into 48 decision stages. Each decision stage includes state variables, action variables, and a value function. The globally optimal power supply strategy is solved through backward recursive calculation, ensuring that power supply reliability penalties and battery loss costs are minimized while meeting load demands. When solar power generation capacity is strong, solar panels are prioritized for power supply and battery charging. When wind power generation capacity is strong, small wind turbines are prioritized for power supply. When both renewable energy sources are insufficient, batteries provide backup power. This dynamic switching mechanism achieves synergistic complementarity among multiple energy sources. Because solar and wind energy have natural complementarity in their temporal distribution—solar power generation efficiency is high during the day when there is sufficient sunlight, while strong winds often accompany nighttime or rainy weather—the multi-source collaborative energy supply mechanism fully utilizes this complementary characteristic, significantly improving power supply stability.

[0057] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0058] The specific implementation of step S1 involves an environmental sensor collecting data on light intensity, wind speed, and ambient temperature. This data is then transmitted to the energy management unit via a line. Upon receiving the light intensity and wind speed data, the energy management unit determines the current operating condition type according to preset identification rules. When the light intensity... Greater than 600 And wind speed When the speed is less than 3 m / s, it is identified as a strong light and weak wind condition. When the light intensity is less than 3 m / s, it is identified as a strong light and weak wind condition. Less than 200 And wind speed A speed greater than 5 m / s is identified as a low-light, high-wind condition. When the light intensity... Less than 200 And wind speed When the speed is less than 3 m / s, it is identified as a windless and solar-free operating condition; other combinations of light intensity and wind speed are identified as mixed operating conditions. In the formula... The light intensity collected by the environmental sensor, in units of , The wind speed is measured by environmental sensors, in m / s.

[0059] The specific implementation of step S2 involves the energy management unit calling a multi-stage energy dispatch model, dividing the 24-hour monitoring cycle into 48 decision stages, each lasting 30 minutes. Each decision-making stage has a state variable represented as... ,in The current remaining battery charge, in units of , Real-time power generation of the solar panel, in units of , This refers to the real-time power generation of a small wind turbine, in units of... Action variables This indicates the choice of power supply mode, including solar-first, wind-first, and hybrid power supply modes. The value function is expressed as: ; In the formula For the first The value function for each decision stage is a dimensionless number. The weighting coefficient for the power supply reliability penalty term is a dimensionless number. The duration of unmet load demand, in units of , For reference duration, the value is 30. , This is the battery loss cost weighting coefficient, with an empirical value of 0.2, and is a dimensionless number. This refers to the cost of battery wear and tear, expressed in yuan. For reference cost, a value of 100 yuan is used. Weighting coefficient for power supply reliability penalty item. The remaining battery capacity percentage is determined based on the battery's remaining charge percentage. The formula for calculating the remaining battery charge percentage is as follows: ; In the formula This represents the percentage of remaining battery charge, and is a dimensionless number. The rated capacity of the battery is expressed in units of 1000 kilowatts. .when hour ,when hour ,when hour The formula for calculating battery loss costs is as follows: ; In the formula The charge / discharge depth cost factor is empirically valued at 50 yuan. For the first The depth of charge / discharge of a staged battery is a dimensionless number, ranging from 0 to 1. The cost coefficient for the number of cycles is empirically set at 0.1 yuan. The number of battery cycles is a dimensionless number.

[0060] The specific implementation method of step S3 is the same as described above, and will not be repeated in detail here.

[0061] The specific implementation of step S4 involves the edge computing module calling a multi-parameter cross-calibration model to perform interference suppression processing on the chemical oxygen demand, total phosphorus concentration, and total nitrogen concentration data. The input vector is represented as follows: ,in This is the raw value of chemical oxygen demand, in mg / L. This is the original value of total phosphorus concentration, in mg / L. This is the original value of total nitrogen concentration, in mg / L. This is the water temperature value, in °C. The pH value is a dimensionless number. This is the turbidity value, in units of... By projecting the 6-dimensional input data into a 3-dimensional principal component space through singular value decomposition, the normalized input vector is represented as: ; In the formula Let be the normalized input vector, and be a dimensionless number. The reference value for chemical oxygen demand is 100 mg / L. The total phosphorus concentration is a reference value, taken as 1 mg / L. The total nitrogen concentration is a reference value of 10 mg / L. This is a reference value for water temperature, taken as 25℃. The pH reference value is 7. This is a turbidity reference value, with a value of 10. The projection matrix is The principal component vector is represented as: ; In the formula These are the principal component vectors corresponding to the first three largest singular values, each vector having a dimension of 6. is a 3D principal component eigenvector, and is a dimensionless number. The output correction value is expressed as: ; In the formula The normalized and corrected data vector is a dimensionless number. The data is the corrected chemical oxygen demand (COD) in mg / L. The total phosphorus concentration data is corrected and is in mg / L. The total nitrogen concentration data is corrected and is expressed in mg / L. The mapping coefficient matrix has dimensions of . , is a dimensionless number. is the bias vector, with a dimension of 3, and is a dimensionless number.

[0062] The specific implementation of step S5 is as follows: the edge computing module first extracts the corrected chemical oxygen demand data, corrected total phosphorus concentration data, and corrected total nitrogen concentration data from the most recent 10 sampling periods, and calculates the standard deviation for each. , and In the formula The standard deviation of the corrected chemical oxygen demand (COD) data is given in mg / L. The standard deviation of the corrected total phosphorus concentration data is given in mg / L. This represents the standard deviation of the corrected total nitrogen concentration data, in mg / L. The formula for calculating the data fluctuation characteristic value is: ; In the formula The data fluctuation characteristic value is a dimensionless number. This represents the maximum standard deviation of chemical oxygen demand (COD) in historical statistics, expressed in mg / L. This represents the maximum standard deviation of total phosphorus concentration in historical statistics, expressed in mg / L. This represents the maximum standard deviation of total nitrogen concentration in historical statistics, expressed in mg / L. The formula for calculating the gradient change characteristic value is: ; In the formula Here, represents the gradient change eigenvalue, and is a dimensionless number. For the first Corrected chemical oxygen demand data for each sampling period, in mg / L. For the first Corrected chemical oxygen demand data for each sampling period, in mg / L. For the first Corrected total phosphorus concentration data for each sampling period, in mg / L. For the first Corrected total phosphorus concentration data for each sampling period, in mg / L. For the first Corrected total nitrogen concentration data for each sampling period, in mg / L. For the first Corrected total nitrogen concentration data for each sampling period, in mg / L. This represents the maximum difference in chemical oxygen demand (COD) between adjacent sampling periods in historical statistics, expressed in mg / L. This represents the maximum difference in total phosphorus concentration between adjacent sampling periods in historical statistics, expressed in mg / L. This represents the maximum difference in total nitrogen concentration between adjacent sampling periods in historical statistics, expressed in mg / L. The adaptive sampling adjustment factor is calculated using the following formula: ; In the formula is the adaptive sampling adjustment factor, and is a dimensionless number. When Sampling interval Extend to the reference sampling interval 2 times ,when Maintain the reference sampling interval at the time. ,when The sampling interval will be shortened to 0.5 times the reference sampling interval. In the formula The adjusted sampling interval, in units of , The reference sampling interval is expressed in units of 1000 m / s. The value ranges from 10 minutes to 30 minutes.

[0063] The specific implementation method of step S6 is the same as described above, and will not be repeated in detail here.

[0064] To better understand and implement this invention, the following is a specific application scenario of this invention, Example 2: To address the issues of unstable power supply and insufficient data accuracy in water quality monitoring at a certain reservoir, technicians deployed a multi-source collaborative power supply intelligent water quality monitoring device in the reservoir area, such as... Figures 2-5 As shown, a multi-source collaborative energy supply system is constructed by integrating solar panels, a small wind turbine, and a battery. Combined with a multi-parameter cross-calibration model and an adaptive sampling and adjustment mechanism, long-term stable water quality monitoring is achieved. The test area was selected in the center of the reservoir, at a depth of approximately 8 meters, with no obstructions and good sunlight and wind conditions. Technicians fixed support pillars to the bottom of the reservoir and installed solar panels and a small wind turbine through these pillars. An energy management unit, edge computing module, battery, and wireless communication module were placed inside the cabinet. The monitoring module was placed 2 meters underwater, and environmental sensors were fixed to the top of the support pillars.

[0065] Specifically, the aforementioned intelligent water quality monitoring device with multi-source collaborative power supply includes a multi-source collaborative power supply body consisting of a solar panel 1, a small wind turbine 2, and a battery 6, an environmental sensor 4, a monitoring module 9, an energy management unit 12, and an edge computing module 13; the support column 3 is a hollow structure, fixed underground, and fixes the solar panel 1, the small wind turbine 2, and the cabinet 7; the solar panel 1 and the small wind turbine 2 are connected through the wiring inside the support column 3, and are connected to the battery 6 through the channel 5.

[0066] Furthermore, the lower part of the cabinet 7 has a passageway 5 through which the wiring 8 can pass. The cabinet 7 contains an edge computing module 13, an energy management unit 12, and a battery 6.

[0067] Furthermore, the monitoring module 9 is placed in the water body being monitored, and includes a monitoring sensor 10 and an auxiliary sensor 11.

[0068] The working method of the above-mentioned intelligent water quality monitoring device with multi-source coordinated power supply includes the following steps: Step 1: Fix the support column 3 to the ground, and the monitoring device is powered by the solar panel 1, the small wind turbine 2 and the battery 6. Step 2: The environmental sensor 4 monitors real-time meteorological data of the environment and transmits it to the energy management unit 12. The energy management unit 12 defines the power supply priority and switching logic under operating conditions such as "strong light and weak wind", "weak light and strong wind", and "no wind and light" based on historical power consumption patterns. Step 3: The monitoring sensor 10 in the monitoring module 9 monitors data such as the amount of oxygen consumed, total phosphorus content, and total content of all nitrogen-containing compounds when organic matter and reducing inorganic matter are chemically oxidized. The temperature, pH, turbidity and other conditions monitored by the auxiliary sensor 11 are used to correct the deviation of the data monitored by the monitoring sensor 10. Step 4: The monitoring module 9 transmits data to the edge computing module 13 via line 8 for local preprocessing, anomaly detection, and compression of the collected data. The data is then uploaded to the mobile device via an adaptive communication protocol. Anomalies will trigger a notification on the mobile device for viewing. The edge computing module executes the steps provided in the specific implementation embodiment.

[0069] In the initial stage of actual operation, technicians conducted continuous monitoring of the system for 7 days. Environmental sensors collected data on light intensity, wind speed, and ambient temperature every 5 minutes. Figure 6 As shown, the curves of light intensity and wind speed variation throughout the day on January 10th exhibit obvious diurnal and weather fluctuation characteristics. From 6:00 AM to 9:00 AM, the light intensity increased from 50... Gradually rising to 720 The wind speed remained around 2.5 m / s. The energy management unit identified it as a strong light, weak wind condition. The multi-stage energy dispatch model output a solar priority power supply mode. The solar panel power generation reached 85W, while the small wind turbine power generation was only 12W. The battery was charging. The solar irradiance reached a peak of 950 ppm between 12:00 PM and 2:00 PM. The wind speed dropped to 1.8 m / s, and the system continued to maintain the solar-first power supply mode. The remaining battery charge increased from 68% to 89%. After 4 PM, the sunlight intensity dropped rapidly, reaching 180 at 6 PM. Simultaneously, the wind speed rose to 6.2 m / s, which the system identified as a low-light, high-wind condition. It switched to a wind-priority power supply mode, increasing the power output of the small wind turbine to 58W. From 8 PM to 5 AM the following day, the solar irradiance decreased to 0. When the wind speed fluctuates between 3.5 and 5.8 m / s, the system enters a hybrid power supply mode, with small wind turbines and batteries working together to supply power. The remaining battery power drops from 89% to 72%. The energy dispatch strategy is switched 6 times throughout the day, and the power supply reliability remains at 100%.

[0070] The monitoring sensors and auxiliary sensors collected water quality parameters at a baseline sampling interval of 15 minutes. A total of 672 sets of data were collected from January 10 to 16, as shown in Table 1. The raw monitoring data for some periods showed obvious environmental interference.

[0071] Table 1. Raw monitoring data for some time periods

[0072] Technicians observed that the raw values ​​of chemical oxygen demand (COD), total phosphorus (TP), and total nitrogen (TN) concentrations on January 11th were significantly higher than at other times. However, turbidity also increased significantly, leading to a preliminary assessment that the measurement bias was caused by turbidity interference. The edge computing module used a multi-parameter cross-calibration model to process the data, constructing a 6-dimensional input vector from the raw values ​​of COD, TP, TN, water temperature, pH, and turbidity. Principal component features were extracted using partial least squares regression and singular value decomposition. The hidden layer projected the 6-dimensional input data into a 3-dimensional principal component space, and the output layer generated corrected COD, TP, and TN data. Figure 7 As shown in the curve, the comparison of chemical oxygen demand (COD) data before and after correction shows that the original COD values ​​on January 11 were 22.4 mg / L and 23.1 mg / L, respectively. After processing by the multi-parameter cross-calibration model, the values ​​were corrected to 18.9 mg / L and 19.3 mg / L, respectively. This eliminated the 4 to 5 mg / L deviation caused by turbidity interference. The corrected data maintained continuity with the data from the preceding and following periods, verifying the effectiveness of the multi-parameter cross-calibration model.

[0073] The edge computing module calculates an adaptive sampling adjustment factor based on the corrected chemical oxygen demand (COD), total phosphorus (TP), and total nitrogen (TN) concentration data, such as... Figure 8 As shown, the adaptive sampling adjustment factor variation curve from January 10th to 16th exhibits obvious fluctuation characteristics. From January 10th to the morning of January 11th, the adaptive sampling adjustment factor remained between 0.25 and 0.35, with the system maintaining a baseline sampling interval of 15 minutes. Starting at 2 PM on January 11th, the adaptive sampling adjustment factor rapidly increased, reaching 0.78 at 3 PM. The system then shortened the sampling interval to 7.5 minutes. Data from the most recent 10 sampling periods were extracted, yielding a data fluctuation characteristic value of 0.42 and a gradient change characteristic value of 0.95. These values ​​were then weighted and summed using weighting coefficients of 0.4 and 0.6 to obtain the adaptive sampling adjustment factor. The sampling adjustment factor was set to 0.78, and high-frequency sampling continued until 6 PM. A total of 24 sets of data were collected during this period, capturing the complete process of chemical oxygen demand rising from 19.3 mg / L to 28.7 mg / L and then falling back to 20.1 mg / L. In the early morning of January 12, the adaptive sampling adjustment factor dropped to 0.18, and the system extended the sampling interval to 30 minutes to reduce energy consumption. From January 13 to 16, the adaptive sampling adjustment factor stabilized between 0.28 and 0.42, and the system maintained the baseline sampling interval. The sampling interval was adjusted 19 times during the 7-day monitoring period, reducing energy consumption by approximately 32% compared to the fixed sampling interval.

[0074] At 2:00 AM on January 15th, the corrected chemical oxygen demand (COD) data suddenly rose to 42.5 mg / L, exceeding the preset threshold range of 40 mg / L. The edge computing module immediately triggered the abnormal alarm function, such as... Figure 9As shown, after receiving the abnormal alarm information, the mobile terminal displayed an abnormal water quality parameter curve. Technicians checked historical data and found that the chemical oxygen demand (COD) had been rising continuously from 19.6 mg / L at 23:00 on January 14 to a peak of 42.5 mg / L at 2:00 on January 15. The total phosphorus concentration rose from 0.067 mg / L to 0.23 mg / L, and the total nitrogen concentration rose from 0.84 mg / L to 2.38 mg / L. All three indicators were abnormal simultaneously. Technicians rushed to the site to investigate and found that fertilization of upstream farmland had caused rainwater to wash pollutants into the reservoir. After taking emergency measures, the water quality parameters returned to normal at 9:00 am on January 15. After correction, the COD data dropped to 21.3 mg / L. Throughout the entire abnormal event handling process, the system accurately captured the pollution event and promptly issued an alarm, providing key data support for rapid response.

[0075] The multi-stage energy dispatch model exhibits good adaptability within a 7-day monitoring period. As shown in Table 2, the energy dispatch statistics under different operating conditions show that the system can dynamically adjust the power supply strategy according to environmental conditions.

[0076] Table 2. Energy Dispatch Statistics for Different Operating Conditions

[0077] Technicians continuously monitored the battery status, and the remaining battery charge percentage remained between 68% and 92% over a 7-day period, without any over-discharge or overcharge. The battery charge / discharge depth was controlled within 30%, with a cycle count of 53 times. Compared to a single solar power supply solution, this reduces the number of deep charge / discharge cycles by approximately 60%, effectively extending battery life. The weighting coefficient for power supply reliability penalties is dynamically adjusted based on the remaining battery charge percentage. When the remaining battery charge percentage is above 70%, the weighting coefficient is 0.3, prioritizing energy efficiency. When the remaining battery charge percentage is below 40%, the weighting coefficient increases to 0.8, prioritizing power supply reliability. This dynamic weighting adjustment mechanism ensures a balance between energy optimization and power supply assurance.

[0078] This invention represents a significant advancement over traditional water quality monitoring methods. Traditional monitoring devices rely on a single solar power source, consuming battery storage energy at night and during cloudy or rainy weather. Frequent deep charging and discharging shortens battery life. This invention, through a multi-source collaborative energy supply system and a multi-stage energy dispatch model, achieves spatiotemporal complementarity between solar and wind power. Utilizing a dynamic programming algorithm for global optimization, it dynamically switches power supply modes and allocates power ratios based on environmental conditions, minimizing battery wear costs while ensuring power supply reliability. This overcomes the limitations of single-energy supply. Traditional monitoring devices use a single sensor to measure water quality parameters, making them susceptible to interference from environmental factors such as temperature, pH, and turbidity, resulting in significant measurement data deviations. This invention addresses these issues by employing a multi-parameter cross-calibration model combined with partial maximum / minimum power generation. The least squares regression algorithm and the singular value decomposition principle project multidimensional input parameters onto the principal component space, suppressing the cross-interference between different parameters and significantly improving the accuracy of water quality parameter measurement. Traditional monitoring devices use fixed sampling intervals, which cannot adapt to dynamic changes in water quality, resulting in energy waste during periods of stable water quality and delayed response during pollution events. This invention uses an adaptive sampling adjustment mechanism to dynamically adjust the sampling interval based on data fluctuation characteristics and gradient change characteristics. During periods of stable water quality, the sampling interval is extended to reduce energy consumption, while during periods of abnormal water quality changes, the sampling interval is shortened to increase monitoring density, achieving a dynamic balance between energy consumption and monitoring accuracy. The three technological innovations work synergistically to construct a complete closed loop for water quality monitoring, providing reliable technical support for water source protection and environmental supervision.

[0079] It should be noted that the variables involved in this invention are explained in detail in Tables 3 and 4.

[0080] Table 3. Variable Explanation Table (Part 1)

[0081] Table 4. Variable Explanation Table (Part Two)

[0082] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A smart water quality monitoring method with multi-source coordinated power supply, characterized in that, Environmental sensors collect data on light intensity, wind speed, and ambient temperature and transmit it to the energy management unit. The energy management unit identifies the current operating condition based on the combination of light intensity and wind speed, calls a multi-stage energy dispatch model to calculate the predicted power generation and load demand for future periods, outputs power switching commands and power allocation schemes for each period, and controls the power supply priority of solar panels, small wind turbines, and batteries. This enables the complementary advantages of solar and wind power under different meteorological conditions, ensuring a stable and reliable power supply for the monitoring system under various meteorological conditions. Monitoring sensors collect data on chemical oxygen demand (COD), total phosphorus concentration (TP), and total nitrogen concentration (TN) in the water body. Auxiliary sensors simultaneously collect data on water temperature, pH, and turbidity. The edge computing module calls a multi-parameter cross-calibration model to perform interference suppression processing on the COD, TTP, and TN concentration data. Based on the calibrated data, it calculates an adaptive sampling adjustment factor to dynamically adjust the sampling interval. The calibrated data is then compressed and encoded before being uploaded to the mobile terminal via a wireless communication module.

2. The intelligent water quality monitoring method with multi-source coordinated power supply according to claim 1, characterized in that, The identification of the operating condition type specifically involves synchronously recording data on light intensity, wind speed, solar panel power generation, and small wind turbine power generation every hour within a monitoring period of no less than 12 months. The experimental data are classified in two dimensions according to light intensity and wind speed. The average power generation of solar panels and small wind turbines under different combinations of light intensity and wind speed ranges are statistically analyzed. Based on the comparison relationship of the average power generation, the following operating conditions are defined: strong light and weak wind operating condition, weak light and strong wind operating condition, no wind and solar operating condition, and mixed operating condition.

3. The intelligent water quality monitoring method with multi-source coordinated power supply according to claim 2, characterized in that, The specific rule for identifying the operating condition type is that when the light intensity is greater than... Furthermore, a wind speed less than 3 m / s is identified as a strong light and weak wind condition, and a light intensity less than... And the wind speed is greater than When the light intensity is less than 100 pm, the condition is identified as low light and strong wind. Furthermore, when the wind speed is less than 3 m / s, it is identified as a windless and solar-free operating condition; other combinations of light intensity and wind speed are identified as mixed operating conditions.

4. The intelligent water quality monitoring method with multi-source coordinated power supply according to claim 3, characterized in that, The multi-stage energy dispatch model specifically divides the 24-hour monitoring cycle into 48 decision stages, each lasting 30 minutes. The state variables include the current remaining battery power, the real-time power generation of the solar panels, and the real-time power generation of the small wind turbine. The action variables include solar priority power supply mode, wind priority power supply mode, and hybrid power supply mode. The value function includes a power supply reliability penalty term and a battery loss cost term.

5. The intelligent water quality monitoring method with multi-source coordinated power supply according to claim 4, characterized in that, The training dataset for the multi-stage energy dispatch model is established by selecting no less than 365 days of historical meteorological and electricity consumption data, sampling the daily light intensity curve, wind speed curve, and load power curve at 30-minute intervals to form time series samples, labeling the optimal power supply mode and actual battery state changes for each time period, and constructing a training sample set containing state transition trajectories and cumulative revenue.

6. The intelligent water quality monitoring method with multi-source coordinated power supply according to claim 5, characterized in that, The training of the multi-stage energy dispatch model specifically involves initializing the value function parameters for each decision stage, performing backward recursive calculations starting from the last decision stage, iterating through all state variable values ​​and action variable combinations for each decision stage, calculating the sum of the value of the current decision stage and the value of subsequent decision stages, selecting the action variable that minimizes the total value as the optimal strategy, and updating the value function parameters.

7. The intelligent water quality monitoring method with multi-source coordinated power supply according to claim 6, characterized in that, The weighting coefficient of the power supply reliability penalty item is determined based on the remaining battery power percentage. When the remaining battery power percentage is greater than or equal to 70%, the weighting coefficient of the power supply reliability penalty item is 0.3; when the remaining battery power percentage is between 40% and 70%, the weighting coefficient of the power supply reliability penalty item is 0.5; and when the remaining battery power percentage is less than 40%, the weighting coefficient of the power supply reliability penalty item is 0.

8.

8. The intelligent water quality monitoring method with multi-source coordinated power supply according to claim 7, characterized in that, The multi-parameter cross-calibration model specifically involves an input layer that receives six parameters: raw values ​​of chemical oxygen demand (COD), total phosphorus concentration (TP), total nitrogen concentration (TN), water temperature, pH, and turbidity. The hidden layer uses partial least squares regression to extract principal component features. Singular value decomposition projects the six-dimensional input data into a three-dimensional principal component space. The output layer generates calibrated COD, TP, and TN data.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions, which, when executed in a computer, are used to perform the intelligent water quality monitoring method with multi-source coordinated power supply as described in any one of claims 1-8.

10. A multi-source collaborative power supply intelligent water quality monitoring system, characterized in that, The system includes the computer-readable storage medium of claim 9, wherein the system is an edge computing device, the computer-readable storage medium is disposed within the edge computing device, and the edge computing device is provided with a microprocessor that executes program instructions stored in the computer-readable storage medium.

Images