A sampling frequency energy-saving control method and system of a water quality monitoring sensor

Abstract

The application discloses a kind of water quality monitoring sensor's sampling frequency energy-saving control method and system, belong to water quality monitoring and adaptive control field.This method includes obtaining real-time water quality monitoring data and energy consumption state, generates water quality change trend prediction model, based on model and energy consumption calculation dynamic energy-saving sampling frequency, control sensor dynamic adjustment sampling frequency, and continuously monitor feedback update model.This method and system use machine learning model to predict water quality change trend, realize multi-objective optimization calculation dynamic energy-saving sampling frequency, by gradual strategy smooth adjustment sampling frequency, and have pattern recognition and adaptive model update ability, can significantly reduce sensor energy consumption and data transmission cost under the premise of guaranteeing monitoring accuracy, effectively improve the intelligentization, stability and battery endurance of water quality monitoring system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of water quality monitoring and adaptive control, and particularly relates to a sampling frequency energy-saving control method and system for a water quality monitoring sensor. BACKGROUND

[0002] Water quality monitoring is an important part of environmental protection and water resource management, widely used in industrial production, agricultural irrigation, drinking water and ecological environment monitoring, etc. With the rapid development of Internet of Things and sensor technology, the number of water quality monitoring sensors deployed in remote or large-scale water areas is increasing, which continuously collects a large amount of water quality data and transmits it, providing real-time basis for water environment supervision.

[0003] However, the existing water quality monitoring system generally has challenges in real-time control and adaptive management. Sensors usually run in a fixed high-frequency sampling mode, which is a rough sampling method without intelligent control strategy. This mode leads to high power consumption, short battery life and high equipment maintenance frequency. At the same time, the large amount of redundant data generated also increases the burden of data storage and transmission, and fails to effectively adapt to the dynamics of water quality changes and energy consumption limitations. The existing control system fails to achieve dynamic and adaptive adjustment of the sampling frequency of water quality monitoring sensors, thereby limiting the long-term stable operation and wide deployment of water quality monitoring devices. SUMMARY

[0004] To solve the above problems, the present application provides a sampling frequency energy-saving control method for a water quality monitoring sensor, which uses a machine learning model to predict water quality change trend, and calculates a dynamic energy-saving sampling frequency based on a multi-objective optimization function, combined with a progressive strategy to smoothly adjust the sampling frequency, while having pattern recognition and adaptive model updating capabilities, which can significantly reduce sensor energy consumption and data transmission cost while ensuring monitoring accuracy, significantly improving the intelligence, stability and battery endurance of the water quality monitoring system.

[0005] The above objectives can be achieved by the following solutions:

[0006] A sampling frequency energy-saving control method for a water quality monitoring sensor, comprising: obtaining real-time water quality monitoring data and the current energy consumption state of the water quality monitoring sensor; generating a water quality change trend prediction model according to the water quality monitoring data; calculating a dynamic energy-saving sampling frequency of the water quality monitoring sensor based on the water quality change trend prediction model and the energy consumption state, combined with a preset energy-saving strategy; controlling the water quality monitoring sensor according to the dynamic energy-saving sampling frequency to dynamically adjust the sampling frequency; continuously monitoring water quality changes within a monitoring period and updating the water quality change trend prediction model according to the monitoring results.

[0007] Optionally, the acquiring real-time water quality monitoring data and current energy consumption state of the water quality monitoring sensor comprises: collecting water quality parameters and external environment parameters generated by the water quality monitoring sensor to generate real-time water quality monitoring data, wherein the external environment parameters comprise air temperature, rainfall, and water flow speed; and acquiring energy consumption information of the water quality monitoring sensor or a power supply module connected to the water quality monitoring sensor to generate an energy consumption state, wherein the energy consumption information comprises battery power, data transmission volume, and operation mode.

[0008] Optionally, the generating a water quality change trend prediction model comprises: constructing a machine learning model based on the real-time water quality monitoring data and historical water quality monitoring data to learn the spatiotemporal evolution law of water quality parameters; optimizing parameters of the machine learning model according to real-time water quality monitoring data and corresponding sampling frequency data, and verifying the prediction accuracy of the machine learning model to generate a water quality change trend prediction model.

[0009] Optionally, the calculating a dynamic energy-saving sampling frequency of the water quality monitoring sensor comprises: comprehensively evaluating a water quality fluctuation risk level according to an output result of the water quality change trend prediction model; based on the water quality fluctuation risk level and the energy consumption state, and in combination with a preset energy-saving benefit maximization rule, calculating a minimum sampling frequency that meets a monitoring accuracy requirement to generate a dynamic energy-saving sampling frequency.

[0010] Optionally, the dynamically adjusting the sampling frequency comprises: judging a difference between a current sampling frequency of the water quality monitoring sensor and the dynamic energy-saving sampling frequency; when the difference exceeds a preset threshold, generating a frequency adjustment instruction to control the water quality monitoring sensor to adjust the sampling frequency; using a gradual adjustment strategy to smoothly transition the sampling frequency of the water quality monitoring sensor from the current frequency to the dynamic energy-saving sampling frequency; and monitoring a working state of the water quality monitoring sensor in real time during the sampling frequency adjustment process, and triggering an alarm or a recovery mechanism when an abnormality occurs.

[0011] Optionally, the updating the water quality change trend prediction model according to the monitoring result feedback comprises: collecting real-time water quality monitoring data, corresponding dynamic energy-saving sampling frequencies, and actual energy consumption data obtained in a monitoring period to form a feedback data set; evaluating the prediction accuracy of the water quality change trend prediction model in the current monitoring period; and when the prediction accuracy is lower than an accuracy tolerance or a new significant water quality change pattern appears, retraining or fine-tuning the water quality change trend prediction model using the feedback data set.

[0012] Optionally, constructing the machine learning model based on the real-time water quality monitoring data and the historical water quality monitoring data comprises: performing data preprocessing on the real-time water quality monitoring data and the historical water quality monitoring data to generate preprocessed data, wherein the data preprocessing comprises outlier rejection, missing value filling, and data standardization; extracting time series features, statistical features, or correlation features related to the water quality change trend from the preprocessed data to construct a model training data set; designing a network structure or algorithm parameters of the machine learning model, and initializing weights or states of the machine learning model based on the training data set.

[0013] Optionally, the combination of the preset energy-saving benefit maximization rule comprises: taking water quality monitoring accuracy, energy consumption, and data transmission cost as optimization objectives to construct a multi-objective optimization function; based on the water quality fluctuation risk level and the energy consumption state, solving the multi-objective optimization function to minimize the energy consumption and the data transmission cost under the premise of meeting the water quality monitoring accuracy; and determining the lowest sampling frequency according to the solving result of the multi-objective optimization function.

[0014] Optionally, the retraining or parameter fine-tuning of the water quality change trend prediction model by using the feedback data set comprises: identifying a new water quality change mode that has a significant statistical difference or clustering anomaly with the learned mode of the water quality change trend prediction model by performing pattern recognition analysis on the feedback data set; and adaptively selecting a strategy of incremental training or global retraining of the water quality change trend prediction model according to the feature or the degree of prediction accuracy decrease of the new water quality change mode.

[0015] Based on the same inventive concept, the application also provides a sampling frequency energy-saving control system of a water quality monitoring sensor, the system comprising:

[0016] a data acquisition module configured to acquire real-time water quality monitoring data and a current energy consumption state of the water quality monitoring sensor;

[0017] a trend prediction module configured to generate a water quality change trend prediction model according to the water quality monitoring data;

[0018] a frequency calculation module configured to calculate a dynamic energy-saving sampling frequency of the water quality monitoring sensor based on the water quality change trend prediction model and the energy consumption state, and in combination with a preset energy-saving strategy;

[0019] a frequency adjustment control module configured to control the water quality monitoring sensor to dynamically adjust the sampling frequency according to the dynamic energy-saving sampling frequency;

[0020] a model updating module configured to continuously monitor water quality changes in a monitoring period, and update the water quality change trend prediction model according to the monitoring result feedback.

[0021] Compared with the prior art, the present application has the following advantages:

[0022] 1、The present application innovatively uses a machine learning model to generate a water quality change trend prediction model, and through data preprocessing and feature extraction on historical data, real-time data and external environmental parameters, it can effectively learn the spatio-temporal evolution law of water quality parameters. This intelligent control decision based on accurate prediction significantly improves the accuracy and foresight of water quality fluctuation risk assessment, and constitutes the control basis for dynamic sampling frequency adaptive adjustment, which can avoid improper control strategy or excessive sampling caused by lag response, thereby optimizing resource allocation and improving system operation efficiency. This fully demonstrates the intelligent capability of the system in environmental perception and control optimization.

[0023] 2、The present application introduces a multi-objective optimization function based on water quality fluctuation risk level and energy consumption state, which can comprehensively solve the system energy consumption and data transmission cost as the target, and realize the maximum energy saving benefit under the premise of meeting the system operation requirements. This multi-objective optimization algorithm is the core mechanism of intelligent decision and adaptive adjustment. At the same time, in the sampling frequency adjustment, the progressive strategy is adopted combined with real-time state evaluation and abnormal processing mechanism, which effectively guarantees the stability of the control process, the stable operation of the system and the rapid recovery ability in abnormal conditions, avoids the system instability or power consumption impact caused by traditional adjustment method, and significantly improves the robustness, energy efficiency and long-term stable operation ability of the system. This fully demonstrates the advancement of the present application in adaptive control and optimization scheduling of complex systems.

[0024] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be achieved and obtained by the structures indicated in the specification, claims and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0026] Figure 1 is a flowchart of a sampling frequency energy-saving control method of a water quality monitoring sensor according to an embodiment of the present application.

[0027] Figure 2 is a change trend line graph of a typical water quality parameter dissolved oxygen in a specific time period according to an embodiment of the present application.

[0028] Figure 3 This is a heat map of the error distribution in water quality parameter prediction according to an embodiment of the present invention.

[0029] Figure 4 This is a power consumption curve of a typical water quality monitoring sensor according to an embodiment of the present invention at different sampling frequencies.

[0030] Figure 5 This is a schematic diagram of the energy-saving control system for the sampling frequency of a water quality monitoring sensor according to an embodiment of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Reference Figure 1 One embodiment of the present invention proposes an energy-saving control method for the sampling frequency of a water quality monitoring sensor. The method uses a machine learning model to predict the trend of water quality changes and calculates the dynamic energy-saving sampling frequency based on a multi-objective optimization function. It combines a progressive strategy to smoothly adjust the sampling frequency and has the ability to recognize patterns and update models. It can significantly reduce sensor energy consumption and data transmission costs while ensuring monitoring accuracy, and significantly improve the intelligence, stability and battery life of the water quality monitoring system.

[0033] The method described in this embodiment specifically includes:

[0034] Acquire real-time water quality monitoring data and the current energy consumption status of water quality monitoring sensors;

[0035] Specifically, in practical applications, water quality monitoring sensors deployed in water bodies or water quality monitoring stations can collect various water quality parameters in real time, such as pH value, dissolved oxygen concentration, turbidity, and conductivity, and periodically send this data to a data processing unit. Simultaneously, the processing unit also receives operational status information from the sensor itself or its power supply unit, such as battery level, current consumption, and data transmission volume, to comprehensively understand the sensor's current energy consumption status.

[0036] Based on the water quality monitoring data, a water quality change trend prediction model is generated.

[0037] Specifically, the application utilizes historical accumulated water quality monitoring data and currently acquired real-time water quality data to construct a mathematical model capable of capturing the time evolution law of water quality parameters. The model learns the water quality change patterns under different time points and different environmental conditions to predict the possible trend of future water quality. For example, the model can predict whether the water quality parameters will remain stable, fluctuate or change significantly in the next few hours or days.

[0038] Based on the water quality change trend prediction model and the energy consumption state, a dynamic energy-saving sampling frequency of the water quality monitoring sensor is calculated in combination with a preset energy-saving strategy.

[0039] Specifically, after receiving the prediction result of the water quality change trend, the system will combine the current energy consumption state of the water quality monitoring sensor and refer to a set of pre-set energy-saving strategies to determine an optimal sampling frequency. The calculation process aims to balance the accuracy required for water quality monitoring and the energy consumption that the system can bear, ensuring that the monitoring needs are met while maximizing the endurance time of the equipment.

[0040] According to the dynamic energy-saving sampling frequency, the water quality monitoring sensor is controlled to dynamically adjust the sampling frequency.

[0041] Specifically, once the optimal dynamic energy-saving sampling frequency is determined, the system will send a control instruction to the corresponding water quality monitoring sensor. After receiving the instruction, the sensor will immediately adjust the frequency of data collection to keep consistent with the dynamic energy-saving sampling frequency calculated by the system. This adjustment can be real-time or phased to ensure that the sensor can work smoothly and reliably according to the new frequency.

[0042] During the monitoring period, the water quality change is continuously monitored, and the water quality change trend prediction model is updated according to the monitoring results.

[0043] Specifically, the system will not stop monitoring the water quality and evaluating the working state of the sensor. While the sensor operates according to the new sampling frequency, the system will continuously receive new water quality monitoring data. These new data will be used to verify the accuracy of the water quality change trend prediction model, and by comparing the actual monitoring results with the model prediction results, the internal parameters or structure of the prediction model will be adjusted and optimized periodically or when significant deviations occur, ensuring that the model can continuously learn and adapt to new water quality environments and change patterns.

[0044] By employing real-time data acquisition and energy consumption status assessment, machine learning models to predict water quality change trends, dynamic energy-saving sampling frequency calculation based on multi-objective optimization and energy-saving benefit maximization rules, and smooth and gradual adjustment control of sensors, combined with continuous monitoring and feedback update mechanisms to optimize the model, the system can minimize sensor energy consumption, extend equipment battery life, and reduce data transmission and storage burden while ensuring water quality monitoring accuracy. At the same time, it ensures the system's rapid response to water quality changes and stable operation, significantly improving the intelligence level, resource utilization efficiency, adaptability, and overall reliability of the water quality monitoring system.

[0045] Optionally, acquiring real-time water quality monitoring data and the current energy consumption status of the water quality monitoring sensor includes:

[0046] The system collects water quality parameters and external environmental parameters generated by water quality monitoring sensors to generate real-time water quality monitoring data. The external environmental parameters include air temperature, rainfall, and water flow velocity.

[0047] Specifically, water quality monitoring sensors are deployed in specific water bodies such as rivers, lakes, or industrial discharge outlets to periodically collect key water quality parameters such as pH, dissolved oxygen concentration, turbidity, and conductivity. Simultaneously, external environmental data, such as real-time temperature, rainfall, and water flow velocity data for the monitored area, can be acquired through independent sensors, meteorological station interfaces, or hydrological monitoring stations, and transmitted synchronously with the water quality parameters to the data processing center to form comprehensive real-time water quality monitoring data.

[0048] For example, in a specific body of water, water quality monitoring sensors collect pH and dissolved oxygen concentration data every 5 minutes. Simultaneously, real-time air temperature and rainfall are acquired via a connected wireless weather sensor, and water flow velocity is measured by an ultrasonic flow meter. This data is packaged and transmitted via a cellular network to a cloud server, where it is initially integrated to construct real-time water quality monitoring data for the current moment.

[0049] The energy consumption information of the water quality monitoring sensor or the power supply module connected to the water quality monitoring sensor is obtained, and an energy consumption status is generated. The energy consumption information includes battery power, data transmission volume, and operating mode.

[0050] Specifically, the system acquires energy consumption information through two main channels. First, the power management unit inside the water quality monitoring sensor monitors and reports its own battery percentage, the instantaneous power consumption of sensor modules (such as measurement probes and data acquisition chips), and the current operating mode (such as sleep mode, sampling mode, and data transmission mode) in real time. Second, if the sensor uses an independent power supply module (such as a solar power system or an external battery pack), this module also monitors its output current, voltage, remaining energy, and the duration of power supply to the sensor. This information is then aggregated to quantify and generate the current energy consumption status.

[0051] For example, a water quality monitoring sensor reports its remaining battery power and total data transmission volume for the day to an edge gateway via Bluetooth Low Energy. Assume the remaining battery power is 75% and the total data transmission volume is 15MB. Simultaneously, the sensor may currently be in "interval sampling mode" rather than continuous sampling mode. These specific data points are received and integrated by the system to comprehensively assess the sensor's current power consumption status.

[0052] Optionally, the water quality change trend prediction model includes:

[0053] A machine learning model based on the real-time water quality monitoring data and historical water quality monitoring data is constructed to learn the spatiotemporal evolution patterns of water quality parameters;

[0054] Specifically, a powerful machine learning model is constructed using accumulated historical water quality monitoring datasets and real-time acquired water quality monitoring data. This model, through deep learning on this multi-dimensional data, can capture the trends and periodic patterns of water quality parameters over time, as well as the correlations between different spatial locations or parameters. The model's design enables it to identify potential patterns of water quality change, such as seasonal fluctuations and subtle changes preceding sudden pollution events. During the construction process, if a certain water quality parameter... At the point of time The value is represented as Its change over time can be modeled as follows:

[0055] ,

[0056] in, This represents the non-linear mapping relationship learned by the machine learning model. Historical water quality parameters This represents the external environment parameters at the current moment, while This indicates a characteristic related to sampling frequency. Through training, this model can accurately learn how water quality parameters change based on their historical state, external environment, and sampling frequency. This machine learning model can employ, for example, long short-term memory networks or convolutional neural networks.

[0057] like Figure 2 As shown, this illustrates the comparison between actual dissolved oxygen monitoring values ​​and model predictions.

[0058] For example, at an industrial wastewater discharge outlet, the system collects hourly water quality data and corresponding environmental data accumulated over the past year. This historical data, along with current data collected in real time by sensors, is input into a pre-trained Long Short-Term Memory (LSTM) network. Through its unique gating mechanism, this LTM network can effectively learn wastewater discharge patterns, the impact of seasonal temperature on water quality, and changes in data characteristics at different sampling frequencies, thus providing a foundation for subsequent water quality trend prediction.

[0059] Based on real-time water quality monitoring data and corresponding sampling frequency data, the parameters of the machine learning model are optimized, the prediction accuracy of the machine learning model is verified, and a water quality change trend prediction model is generated.

[0060] Specifically, after the machine learning model is initially built and trained, the system continuously optimizes and adjusts the model's parameters using new real-time water quality monitoring data and the corresponding sampling frequency data. This process aims to ensure that the model can adapt to the latest changes in the water quality environment and the sensor's operating status. Simultaneously, the system compares the model's predicted output... Compared with actual monitoring results Calculate the prediction error, for example, using mean squared error or mean absolute error as an evaluation metric to quantify the model's predictive accuracy. Loss function It can be defined as the sum of squared errors between the predicted and actual values:

[0061] ,

[0062] in, This refers to the number of data points used for optimization and validation. The system will periodically, or when it detects significant deviations in model predictions, utilize new monitoring data. The model weights are adjusted using algorithms such as gradient descent or its variants, based on the corresponding input features (including sampling frequency data). and bias To minimize its loss function:

[0063] ,

[0064] ,

[0065] in, It is the learning rate. Through this iterative optimization and verification, a highly accurate predictive model that can accurately reflect the trend of water quality changes is finally formed.

[0066] like Figure 3 As shown, the average prediction error of the model is displayed under different sampling frequencies and water quality fluctuation risk levels.

[0067] For example, the system automatically fine-tunes the model parameters weekly, updating the model using the latest week's water quality monitoring data and sampling frequency data. If, during this period, the predicted MAE suddenly exceeds 0.5 units (e.g., pH prediction error exceeds 0.5), the system immediately triggers an additional model optimization process to ensure the model's continued accuracy.

[0068] Optionally, the dynamic energy-saving sampling frequency of the calculated water quality monitoring sensor includes:

[0069] Based on the output of the water quality change trend prediction model, the risk level of water quality fluctuations is comprehensively assessed.

[0070] Specifically, after completing the prediction, the water quality change trend prediction model outputs the changing trends of various water quality parameters over a future period, along with their prediction confidence levels. Based on these outputs, the system comprehensively assesses the potential volatility of the water quality environment. This assessment considers not only the expected magnitude or rate of change of individual water quality parameters but also the correlation between multiple parameters and the uncertainty of the prediction results. For example, when multiple key water quality parameters are predicted to change rapidly, or when the prediction confidence is low, the water quality volatility risk level will be assessed as high. Water Quality Volatility Risk Level The following methods can be used for evaluation:

[0071] ,

[0072] in, It is an evaluation function. Representing the The predicted range or rate of change of each water quality parameter. This represents the prediction confidence level, while This indicates the interaction between different water quality parameters or external environmental factors. Risk Level It can be quantified into discrete levels of low, medium, high, or continuous numerical values.

[0073] For example, if a machine learning model predicts that dissolved oxygen concentration will decrease by more than 2 mg / L and turbidity will increase by more than 5 NTU within the next 2 hours, the system will assess it as "high risk." This indicates that significant changes may be occurring in the water body, requiring more intensive monitoring.

[0074] Based on the water quality fluctuation risk level and energy consumption status, and combined with the preset rules for maximizing energy-saving benefits, the minimum sampling frequency that meets the monitoring accuracy requirements is calculated, and a dynamic energy-saving sampling frequency is generated.

[0075] Specifically, the system utilizes the previously assessed water quality fluctuation risk level. And the current energy consumption status of water quality monitoring sensors (Including battery power, data transmission volume, operating mode, etc.), and combined with preset rules for maximizing energy efficiency, calculate the minimum sampling frequency that can meet the necessary monitoring accuracy and maximize energy saving under the current conditions. This rule aims to find the optimal balance between monitoring quality and resource consumption. The calculation process considers the impact of different sampling frequencies on data accuracy, energy consumption, and data transmission volume. The system determines this by solving a multi-objective optimization problem. The optimization problem can be expressed as:

[0076] ,

[0077] ,

[0078] ,

[0079] in, Indicates the sampling frequency Operating costs below Indicates energy consumption. Indicates the amount of data transmitted. At frequency The monitoring accuracy must meet the requirements of the water quality fluctuation risk level. The required precision of the decision . This is a preset selectable sampling frequency level. The solution process determines a dynamic energy-saving sampling frequency through iterative calculation or lookup table.

[0080] For example, if the water quality fluctuation risk level is assessed as "high risk," the system will require higher monitoring accuracy. In this case, even if the battery power is low, the system will prioritize a relatively high sampling frequency, for example, adjusting from 1 hour / time to 10 minutes / time, because it meets the high accuracy requirements. However, if the water quality is at "low risk" and the battery power is sufficient, the system may choose a very low sampling frequency, for example, adjusting from 1 hour / time to 6 hours / time, to maximize battery life. Specific rules for maximizing energy efficiency may include weighting or prioritization.

[0081] Optionally, the dynamic adjustment of the sampling frequency includes:

[0082] Determine the difference between the current sampling frequency and the dynamic energy-saving sampling frequency of the water quality monitoring sensor;

[0083] Specifically, the system continuously acquires the current actual sampling frequency of the water quality monitoring sensors. Simultaneously, the frequency calculation results indicate a dynamic energy-saving sampling frequency derived from current water quality trends and energy consumption status. These two frequency values ​​are compared, and their absolute or relative differences are calculated to assess whether frequency adjustment is necessary.

[0084] For example, suppose the water quality monitoring sensor currently samples every 30 minutes, while frequency calculations indicate that the dynamic energy-saving sampling frequency should be every 2 hours. The system detects a significant difference between these two frequencies and determines that an adjustment is needed.

[0085] When the difference exceeds a preset threshold, a frequency adjustment command is generated to control the water quality monitoring sensor to adjust the sampling frequency.

[0086] Specifically, when the difference between the current sampling frequency and the dynamic energy-saving sampling frequency exceeds a preset allowable range (e.g., a difference of less than 10% is negligible), a frequency adjustment process is triggered. At this time, based on the calculated dynamic energy-saving sampling frequency, a control instruction package containing the new frequency value and adjustment command is generated. This instruction is sent to the target water quality monitoring sensor via the communication module, instructing it to perform the frequency adjustment operation.

[0087] For example, if the difference between the current sampling frequency and the dynamic energy-saving sampling frequency exceeds a threshold of 1 minute per sampling, and the new frequency is lower than the current frequency, an instruction containing the new frequency parameters will be immediately generated and sent to the corresponding sensor via the wireless network. After receiving the instruction, the sensor will adjust its internal clock or sampling period according to the instruction requirements.

[0088] A gradual adjustment strategy is adopted to smoothly transition the sampling frequency of the water quality monitoring sensor from the current frequency to a dynamic energy-saving sampling frequency;

[0089] Specifically, to avoid sudden changes in sampling frequency impacting sensor hardware or data stream stability, a gradual adjustment strategy is employed. This means that frequency adjustment is not done in one step, but rather gradually approaches the target frequency through multiple small steps. If it is necessary to reduce the sampling frequency from a high frequency to a low frequency, the frequency will be reduced slightly first, and after running for a period of time, a second reduction will be made until the target frequency is reached. Conversely, the same process is repeated to ensure the smoothness of the adjustment process and the continuity of data.

[0090] For example, when the sampling frequency needs to be adjusted from once every 5 minutes to once every 30 minutes, it doesn't jump abruptly. It might first adjust the frequency to once every 10 minutes, run for 15 minutes, then adjust it to once every 20 minutes, and finally reach the target frequency of once every 30 minutes after another 15 minutes. This smooth transition reduces drastic changes in the data sampling interval.

[0091] During the sampling frequency adjustment process, the working status of the water quality monitoring sensor is monitored in real time, and an alarm or recovery mechanism is triggered when an abnormality occurs.

[0092] Specifically, during sampling frequency adjustment, the operating status of the water quality monitoring sensor should be closely monitored, including but not limited to its power consumption level, data transmission success rate, internal component temperature, and the stability of core water quality parameter readings. If any abnormality is detected, such as a sudden surge in power consumption, continuous data transmission failures, or the sensor reporting an internal error, a preset alarm mechanism will be immediately triggered, and a corresponding recovery mechanism will be initiated according to the type of abnormality. This could include pausing frequency adjustment, restoring to the most recent stable frequency, or attempting to restart the sensor module to ensure the normal operation of the equipment and the integrity of the data.

[0093] For example, during the frequency adjustment process of a water quality monitoring sensor from high to low, its data transmission success rate suddenly drops from 99% to 20%. This would be considered abnormal. At this point, the system immediately sends an SMS alert to the maintenance personnel and automatically pauses the current frequency adjustment, temporarily restoring the sensor's sampling frequency to its stable state before the adjustment, pending intervention from the maintenance personnel for troubleshooting.

[0094] Optionally, the step of updating the water quality change trend prediction model based on monitoring results includes:

[0095] Collect real-time water quality monitoring data, corresponding dynamic energy-saving sampling frequency, and actual energy consumption data acquired during the monitoring period to form a feedback dataset;

[0096] Specifically, during the water quality monitoring cycle, real-time water quality monitoring data is continuously received and stored. Simultaneously, the dynamic energy-saving sampling frequency of the water quality monitoring sensors and the actual energy consumption of the sensors at different frequencies are also recorded. These different types and multi-dimensional data are integrated and archived to form a comprehensive dataset for model feedback updates, ensuring that the feedback information reflects the actual operating conditions.

[0097] For example, during a one-month monitoring period, real-time water quality parameter data is stored hourly. The actual sampling frequency after each frequency adjustment command is sent, as well as the cumulative energy consumption data reported by the sensors every 24 hours, are recorded synchronously. This data is compiled into a time-series database as a feedback dataset.

[0098] Assess the accuracy of the water quality change trend prediction model within the current monitoring period;

[0099] Specifically, the feedback dataset is used to quantitatively evaluate the actual predictive performance of the water quality change trend prediction model in the most recent monitoring period. This includes comparing the model's predictions at the beginning of the period or after adjustments with the actual water quality data monitored during the period. Commonly used evaluation metrics include, but are not limited to, root mean square error, mean absolute percentage error, or correlation coefficient. These metrics objectively determine the degree of agreement between the model's predictions and the actual situation.

[0100] For example, the root mean square error (RMSE) between the turbidity predicted by the model and the actual turbidity over the next 24 hours is calculated. If the calculated RMSE value is 0.8 NTU, it is compared with a pre-defined performance standard. This RMSE value can serve as a key indicator for evaluating the accuracy of the model's predictions.

[0101] When the prediction accuracy is below the accuracy tolerance or a new significant water quality change pattern emerges, the water quality change trend prediction model is retrained or its parameters are fine-tuned using the feedback dataset.

[0102] Specifically, the accuracy of the model's predictions is continuously checked and evaluated. If the accuracy falls below the preset tolerance, or if pattern recognition analysis reveals a new water quality change pattern in the feedback dataset that is significantly different from the model's historical learning patterns, the model update mechanism is triggered. At this point, the existing water quality change trend prediction model is retrained or its parameters are fine-tuned using the latest collected feedback dataset to adapt to the new changes, thereby generating a more accurate and adaptable prediction model.

[0103] For example, if a water quality model's accuracy in predicting dissolved oxygen concentration changes exceeds the accuracy tolerance of 0.6 mg / L for three consecutive days, this indicates that the model may be outdated. Alternatively, cluster analysis might reveal a previously unseen pattern of strong positive correlation between water quality parameters. In such cases, an incremental training process for the predictive model would be immediately triggered, using real-time data from the past week as new training samples to adapt to this new pattern and improve the model's predictive capabilities.

[0104] Optionally, constructing a machine learning model based on the real-time water quality monitoring data and historical water quality monitoring data includes:

[0105] The real-time water quality monitoring data and historical water quality monitoring data are preprocessed to generate preprocessed data. The data preprocessing includes outlier removal, missing value filling, and data standardization.

[0106] Specifically, before building the machine learning model, rigorous data preprocessing is performed on the raw water quality monitoring data and historical data. Outlier removal aims to identify and remove data points that significantly deviate from the normal range, such as extremely high or low readings caused by sensor malfunctions. This can be achieved through statistical methods or rule-based filtering. Missing value imputation handles data gaps caused by transmission interruptions or sensor malfunctions during data acquisition. Interpolation or machine learning methods can be used to estimate missing values. Data standardization transforms data with different dimensions or value ranges to a uniform scale to eliminate the impact of dimensional differences on model training and ensure that the model can fairly process all input features.

[0107] For example, for real-time dissolved oxygen data, if a reading drops sharply from 8 mg / L to 0.5 mg / L within a short period and then quickly recovers, it may be identified as an outlier and removed. If pH data is lost for one consecutive hour, the missing value can be estimated using linear interpolation, utilizing the pH values ​​at time points before and after which data is available. Meanwhile, to avoid significant differences in temperature and turbidity ranges affecting model convergence, all values ​​will be normalized to the [0,1] interval.

[0108] Extract time-series features, statistical features, or correlation features related to water quality change trends from the preprocessed data to construct a model training dataset;

[0109] Specifically, after data preprocessing, feature extraction is performed to extract features from the preprocessed data that are significant for predicting water quality change trends. Time-series features may include moving averages, moving standard deviations, and autocorrelation function values, used to capture the dynamic changes and periodicity of the data over time. Statistical features may include mean, variance, skewness, and kurtosis, used to describe the distribution characteristics of the data. Correlation features may include lagged correlations between different water quality parameters or between water quality parameters and external environmental parameters. These extracted features, combined with corresponding water quality change labels, constitute a structured dataset for training machine learning models.

[0110] For example, the moving average pH value over the past 6 hours is extracted from the preprocessed pH data as a time-series feature. Simultaneously, the standard deviation of dissolved oxygen concentration over the past 24 hours is calculated as a statistical feature to reflect its volatility. Furthermore, the cross-correlation between the current air temperature and the water flow velocity over the past 3 hours can be calculated as a correlation feature. These features, along with corresponding historical water quality change trend labels (e.g., "stable," "slightly fluctuating," "drastic change"), form the input samples for model training.

[0111] Design the network structure or algorithm parameters of the machine learning model, and initialize the weights or states of the machine learning model based on the training dataset.

[0112] Specifically, for the task of predicting water quality change trends, it is necessary to select and design appropriate machine learning model network structures or algorithm parameters. For example, if a neural network model is chosen, this includes determining the number of layers, the number of neurons in each layer, the choice of activation function (such as ReLU, Sigmoid), and the configuration of the output layer. For other types of machine learning algorithms, their specific algorithm parameter settings are involved. After completing the structure and parameter design, the initial weights or internal states of the machine learning model are initialized using the previously constructed model training dataset. This initialization process can employ random initialization, loading a pre-trained model, or a heuristic initialization method based on a specific algorithm, providing a good starting point for subsequent model training.

[0113] For example, a neural network structure with three long short-term memory layers, each containing 64 units, is designed, using ReLU as the activation function. Before training, all network weights are randomly initialized using a Glorot uniform distribution. Then, the constructed training dataset (including preprocessed features and corresponding water quality change trend labels) is input into this initialized network to prepare for formal model training. Backpropagation and gradient descent are used to optimize its weights and biases, enabling it to learn the spatiotemporal evolution patterns of water quality parameters from the data.

[0114] Optionally, the rule for maximizing energy-saving benefits, which is based on a preset formula, includes:

[0115] A multi-objective optimization function is constructed with water quality monitoring accuracy, energy consumption, and data transmission cost as optimization objectives.

[0116] Specifically, in order to achieve the best balance between water quality monitoring effectiveness and resource consumption, a multi-objective optimization function is constructed. This function will improve the accuracy of water quality monitoring. Energy consumption and data transmission costs As its core optimization objective, water quality monitoring accuracy is... Usually related to sampling frequency There is a positive correlation, meaning that the higher the sampling frequency, the higher the accuracy. Energy consumption. Then related to sampling frequency They are positively correlated, including power consumption during sensor sampling and power consumption during data transmission. Data transmission cost. Also related to sampling frequency There is a positive correlation because a higher sampling frequency means a larger amount of data needs to be transmitted. This function is constructed to quantify the combined benefits and costs at different sampling frequencies, providing a mathematical basis for subsequent decision-making. This multi-objective optimization function... It can be represented as:

[0117] ,

[0118] in, It can be represented by evaluating the reciprocal of the prediction error or directly defined as a binary function that satisfies a specific accuracy threshold. This can be calculated based on the sensor hardware manual and data transmission protocol.

[0119] ,

[0120] It can be calculated based on data transmission volume and operator tariff standards.

[0121] like Figure 4 As shown, this reflects the relationship between sampling frequency and energy consumption.

[0122] For example, for a specific water quality monitoring scenario, an optimization function is constructed with the goal of maximizing monitoring accuracy, such as requiring turbidity prediction error to be less than 0.5 NTU, while minimizing daily total energy consumption and monthly data transmission costs. This function takes different sampling frequencies as input variables and outputs corresponding accuracy, energy consumption, and cost assessment values.

[0123] Based on the water quality fluctuation risk level and energy consumption status, a multi-objective optimization function is solved to minimize energy consumption and data transmission cost while meeting the water quality monitoring accuracy requirements.

[0124] Specifically, this involves obtaining the water quality fluctuation risk level output by the water quality change trend prediction model. And the current energy consumption status of water quality monitoring sensors Then, these parameters are substituted as constraints or weights into the constructed multi-objective optimization function. Solving this optimization function aims to find an optimal sampling frequency. This frequency meets the requirements determined by the risk level. The minimum monitoring accuracy requirement determined At the same time, it increases energy consumption. and data transmission costs The goal is to minimize this. The solution process can be implemented using iterative algorithms, heuristic algorithms, or predefined lookup tables to find the optimal balance between accuracy and resource consumption. The optimization problem can be formalized as:

[0125] ,

[0126] ,

[0127] ,

[0128] in, , Based on energy consumption status The weighting coefficients set by the preset energy-saving strategy indicate the relative importance attached to energy consumption and data transmission costs. It is based on the risk level of water quality fluctuations. The required monitoring accuracy is dynamically set. It is a set of discrete sampling frequencies supported by water quality monitoring sensors.

[0129] For example, if the water quality fluctuation risk level is assessed as "moderate," then the monitoring accuracy must reach a mean absolute error of less than 0.2. In this case, using the current battery power and flow cost as inputs, a multi-objective optimization algorithm is run to search the list of allowed sampling frequencies to find the sampling frequency that meets the MAE requirement while minimizing total energy consumption and transmission costs.

[0130] The minimum sampling frequency is determined based on the solution results of the multi-objective optimization function.

[0131] Specifically, solving the multi-objective optimization function will yield an optimal sampling frequency value. This value represents the optimal choice that balances monitoring accuracy and resource consumption under the current water quality environmental risks, energy consumption status, and established energy-saving strategies. This frequency will be directly used as the new dynamic energy-saving sampling frequency for the water quality monitoring sensor. This determined frequency will be the ultimate basis for guiding subsequent adjustments and control of the sensor.

[0132] For example, after solving the multi-objective optimization function, the optimal sampling frequency is determined to be once every 4 hours. This frequency value is then identified as the minimum sampling frequency of the current water quality monitoring sensor, and the sensor is subsequently instructed to collect data at this frequency.

[0133] Optionally, the retraining or parameter fine-tuning of the water quality change trend prediction model using the feedback dataset includes:

[0134] By performing pattern recognition analysis on the feedback dataset, new water quality change patterns that have significant statistical differences or clustering anomalies compared to the patterns already learned by the water quality change trend prediction model are identified.

[0135] Specifically, after obtaining the feedback dataset, in-depth pattern recognition analysis is performed on the real-time water quality monitoring data, the corresponding dynamic energy-saving sampling frequency, and the actual energy consumption data. This analysis aims to detect significant differences or anomalies between the data and historical patterns learned and stored by existing water quality change trend prediction models. This identification can be achieved through various data mining and statistical analysis techniques. For example, clustering algorithms can be used to discover new cluster centers in the data that are not covered by existing models; or statistical tests can be used to determine significant differences between the new data and historical data in terms of mean, variance, or distribution. When these differences reach a preset statistical significance level, a new water quality change pattern can be identified.

[0136] For example, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm was used to analyze feedback data (including pH, dissolved oxygen, turbidity, and their rates of change) from the past week. If the analysis results showed that a new data cluster existed whose features did not match any of the pattern clusters previously learned by the model, and the number of data points in this cluster exceeded 5% of the total data points, it was determined that a new and significant water quality change pattern had emerged.

[0137] Based on the characteristics of the new water quality change pattern or the degree of decline in prediction accuracy, an adaptive strategy is selected to perform incremental training or global retraining on the water quality change trend prediction model.

[0138] Specifically, once a new water quality change pattern is identified or the prediction accuracy is assessed to be below the preset accuracy tolerance, the model update strategy is intelligently selected based on the nature of the problem. If the characteristics of the new pattern indicate a gradual or localized change, and the decrease in accuracy is not significant, an incremental training strategy is preferred. This means using only the new feedback dataset, training with small steps and short cycles based on the original model parameters to quickly adapt to new changes and save computational resources. If the new pattern is disruptive, or the prediction accuracy drops sharply, indicating that the existing model can no longer effectively capture the new patterns, a global retraining strategy is chosen. In this case, all historical data is merged with the new feedback data, and the model is trained comprehensively from scratch to build a completely new model that is better adapted to the current environment. This adaptive selection ensures the efficiency and effectiveness of model updates.

[0139] For example, if the identified new water quality change pattern is due to a slow decline in dissolved oxygen caused by seasonally rising temperatures, and the prediction accuracy decreases by only 2%, incremental training is chosen, which involves fine-tuning the model for 10 epochs using feedback data from the most recent month. Conversely, if a large-scale, sudden pollution event occurs, causing unprecedented and drastic fluctuations in water quality parameters, and the prediction accuracy suddenly drops by 15%, global retraining is triggered. This involves retraining the model using all cleaned data accumulated over the past year to address such fundamental environmental changes.

[0140] Based on the same inventive concept, such as Figure 5 As shown, the present invention also provides an energy-saving control system for the sampling frequency of a water quality monitoring sensor, the system comprising:

[0141] The data acquisition module is used to acquire real-time water quality monitoring data and the current energy consumption status of the water quality monitoring sensors;

[0142] The trend prediction module is used to generate a water quality change trend prediction model based on the water quality monitoring data.

[0143] The frequency calculation module is used to calculate the dynamic energy-saving sampling frequency of the water quality monitoring sensor based on the water quality change trend prediction model and energy consumption status, combined with the preset energy-saving strategy.

[0144] The frequency adjustment control module is used to control the water quality monitoring sensor to dynamically adjust the sampling frequency according to the dynamic energy-saving sampling frequency.

[0145] The model update module is used to continuously monitor water quality changes during the monitoring period and update the water quality change trend prediction model based on the monitoring results.

[0146] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0147] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for energy-saving control of the sampling frequency of a water quality monitoring sensor, characterized in that, The method includes: Acquire real-time water quality monitoring data and the current energy consumption status of water quality monitoring sensors; Based on the water quality monitoring data, a water quality change trend prediction model is generated. Based on the water quality change trend prediction model and energy consumption status, and combined with the preset energy-saving strategy, the dynamic energy-saving sampling frequency of the water quality monitoring sensor is calculated. Based on the dynamic energy-saving sampling frequency, the water quality monitoring sensor is controlled to dynamically adjust its sampling frequency. This dynamic adjustment includes: determining the difference between the current sampling frequency of the water quality monitoring sensor and the dynamic energy-saving sampling frequency; when the difference exceeds a preset threshold, generating a frequency adjustment command to control the water quality monitoring sensor to adjust its sampling frequency; employing a gradual adjustment strategy to smoothly transition the water quality monitoring sensor's sampling frequency from the current frequency to the dynamic energy-saving sampling frequency; and during the sampling frequency adjustment process, monitoring the working status of the water quality monitoring sensor in real time and triggering an alarm or recovery mechanism when an anomaly occurs. During the monitoring period, water quality changes are continuously monitored, and the water quality change trend prediction model is updated based on the monitoring results. This updating includes: collecting real-time water quality monitoring data, corresponding dynamic energy-saving sampling frequencies, and actual energy consumption data acquired during the monitoring period to form a feedback dataset; evaluating the prediction accuracy of the water quality change trend prediction model within the current monitoring period; and retraining or fine-tuning the parameters of the water quality change trend prediction model using the feedback dataset when the prediction accuracy falls below the accuracy tolerance or a new significant water quality change pattern emerges. The step of retraining or fine-tuning the water quality change trend prediction model using the feedback dataset includes: performing pattern recognition analysis on the feedback dataset to identify new water quality change patterns that have significant statistical differences or clustering anomalies compared to the patterns already learned by the water quality change trend prediction model; and adaptively selecting a strategy for incremental training or global retraining of the water quality change trend prediction model based on the characteristics of the new water quality change pattern or the degree of decline in prediction accuracy.

2. The energy-saving control method for the sampling frequency of a water quality monitoring sensor according to claim 1, characterized in that, The acquisition of real-time water quality monitoring data and the current energy consumption status of the water quality monitoring sensor includes: The system collects water quality parameters and external environmental parameters generated by water quality monitoring sensors to generate real-time water quality monitoring data. The external environmental parameters include air temperature, rainfall, and water flow velocity. The energy consumption information of the water quality monitoring sensor or the power supply module connected to the water quality monitoring sensor is obtained, and an energy consumption status is generated. The energy consumption information includes battery power, data transmission volume, and operating mode.

3. The energy-saving control method for the sampling frequency of a water quality monitoring sensor according to claim 1, characterized in that, The generated water quality change trend prediction model includes: A machine learning model based on the real-time water quality monitoring data and historical water quality monitoring data is constructed to learn the spatiotemporal evolution patterns of water quality parameters; Based on real-time water quality monitoring data and corresponding sampling frequency data, the parameters of the machine learning model are optimized, the prediction accuracy of the machine learning model is verified, and a water quality change trend prediction model is generated.

4. The energy-saving control method for the sampling frequency of a water quality monitoring sensor according to claim 1, characterized in that, The dynamic energy-saving sampling frequency of the calculated water quality monitoring sensor includes: Based on the output of the water quality change trend prediction model, the risk level of water quality fluctuations is comprehensively assessed. Based on the water quality fluctuation risk level and energy consumption status, and combined with the preset rules for maximizing energy-saving benefits, the minimum sampling frequency that meets the monitoring accuracy requirements is calculated, and a dynamic energy-saving sampling frequency is generated.

5. The energy-saving control method for the sampling frequency of a water quality monitoring sensor according to claim 3, characterized in that, Constructing a machine learning model based on the real-time water quality monitoring data and historical water quality monitoring data includes: The real-time water quality monitoring data and historical water quality monitoring data are preprocessed to generate preprocessed data. The data preprocessing includes outlier removal, missing value filling, and data standardization. Extract time-series features, statistical features, or correlation features related to water quality change trends from the preprocessed data to construct a model training dataset; Design the network structure or algorithm parameters of the machine learning model, and initialize the weights or states of the machine learning model based on the training dataset.

6. The energy-saving control method for the sampling frequency of a water quality monitoring sensor according to claim 4, characterized in that, The pre-defined rules for maximizing energy-saving benefits include: A multi-objective optimization function is constructed with water quality monitoring accuracy, energy consumption, and data transmission cost as optimization objectives. Based on the water quality fluctuation risk level and energy consumption status, a multi-objective optimization function is solved to minimize energy consumption and data transmission cost while meeting the water quality monitoring accuracy requirements. The minimum sampling frequency is determined based on the solution results of the multi-objective optimization function.

7. A sampling frequency energy-saving control system for a water quality monitoring sensor, applied to the sampling frequency energy-saving control method for a water quality monitoring sensor as described in any one of claims 1-6, characterized in that, The system includes: The data acquisition module is used to acquire real-time water quality monitoring data and the current energy consumption status of the water quality monitoring sensors; The trend prediction module is used to generate a water quality change trend prediction model based on the water quality monitoring data. The frequency calculation module is used to calculate the dynamic energy-saving sampling frequency of the water quality monitoring sensor based on the water quality change trend prediction model and energy consumption status, combined with the preset energy-saving strategy. The frequency adjustment control module is used to control the water quality monitoring sensor to dynamically adjust the sampling frequency according to the dynamic energy-saving sampling frequency. The dynamic adjustment of the sampling frequency includes: determining the difference between the current sampling frequency of the water quality monitoring sensor and the dynamic energy-saving sampling frequency; when the difference exceeds a preset threshold, generating a frequency adjustment command to control the water quality monitoring sensor to adjust the sampling frequency; employing a gradual adjustment strategy to smoothly transition the sampling frequency of the water quality monitoring sensor from the current frequency to the dynamic energy-saving sampling frequency; and monitoring the working status of the water quality monitoring sensor in real time during the sampling frequency adjustment process, triggering an alarm or recovery mechanism when an abnormality occurs. The model update module is used to continuously monitor water quality changes during the monitoring period and update the water quality change trend prediction model based on the monitoring results. This update includes: collecting real-time water quality monitoring data, corresponding dynamic energy-saving sampling frequencies, and actual energy consumption data acquired during the monitoring period to form a feedback dataset; evaluating the prediction accuracy of the water quality change trend prediction model within the current monitoring period; and retraining or fine-tuning the parameters of the water quality change trend prediction model using the feedback dataset when the prediction accuracy falls below the accuracy tolerance or a new significant water quality change pattern emerges. The step of retraining or fine-tuning the water quality change trend prediction model using the feedback dataset includes: performing pattern recognition analysis on the feedback dataset to identify new water quality change patterns that have significant statistical differences or clustering anomalies compared to the patterns already learned by the water quality change trend prediction model; and adaptively selecting a strategy for incremental training or global retraining of the water quality change trend prediction model based on the characteristics of the new water quality change pattern or the degree of decline in prediction accuracy.

Images