Water supply strategy calculation method and system coupled with front-end perception and edge computing

Abstract

The application discloses a water supply strategy calculation method and system coupled with front-end perception and edge calculation, relates to the technical field of intelligent control of water supply networks, and comprises the following steps: collecting water supply state data of nodes of a water supply network; calculating state deviation in an edge calculation device and performing time series difference operation to obtain a state change rate; receiving a global water supply scheduling strategy from a central computer room and generating an edge strategy space; predicting future period state deviation based on the state change rate, and performing weighted summation of the current state deviation and the predicted state deviation to obtain a predicted deviation; calculating a water supply adjustment amount under the constraint of the edge strategy space; simultaneously monitoring the load of the edge calculation device, and migrating the calculation task to an adjacent device if necessary; and finally generating a water supply control instruction to perform water supply adjustment. The application realizes the cooperative operation of edge calculation and central control, and improves the accuracy and real-time performance of water supply network control.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for water supply networks, specifically to a method and system for calculating water supply strategies that couples front-end sensing with edge computing. Background Technology

[0002] As the intelligence level of water supply systems continues to improve, the operation and control of intelligent water supply networks increasingly rely on edge computing technology. Traditional water supply network control systems mainly adopt a centralized control method, where all water supply status data needs to be uploaded to the central control room for processing. This method suffers from problems such as large data transmission latency, concentrated computational load, and slow system response speed.

[0003] While existing edge computing technologies can deploy edge computing devices near water supply network nodes for localized computation, the lack of an effective coordination mechanism with the central control system makes it difficult to ensure consistency between local control decisions and global water supply scheduling strategies. Edge computing devices also have limited computing resources, making them prone to computational delays under high loads, thus affecting control performance.

[0004] Current water supply network control methods often only consider current state deviations and fail to fully utilize historical data for trend prediction, resulting in insufficient foresight in control decisions. Furthermore, the failure to effectively incorporate the load status of edge computing devices during water supply strategy calculations can easily lead to overload of local computing resources. Summary of the Invention

[0005] The purpose of this invention is to provide a water supply strategy calculation method and system that couples front-end perception with edge computing, aiming to solve at least one of the technical problems existing in the prior art.

[0006] The technical solution of this invention is: a water supply strategy calculation method coupled with front-end perception and edge computing, comprising the following steps:

[0007] Collect water supply status data from water supply network nodes and transmit it to edge computing devices;

[0008] In edge computing devices, water supply status data is compared with preset water supply target values ​​to calculate the status deviation. Time-series difference operation is performed on the status deviation of multiple consecutive collection cycles in the water supply status data to obtain the status change rate.

[0009] The system receives the global water supply scheduling strategy from the central computer room, determines the deviation range to which the state deviation belongs based on the state deviation, extracts the control amplitude constraint corresponding to the deviation range from the global water supply scheduling strategy, and generates the edge strategy space.

[0010] The state deviation is predicted based on the state change rate to obtain the state deviation for the future period. The state deviation is then weighted and summed with the state deviation for the future period to obtain the predicted deviation.

[0011] Within the control range constraint defined by the edge strategy space, the water supply adjustment amount is obtained by calculating the water supply strategy based on the predicted deviation.

[0012] Simultaneously, the computing load of the edge computing device is detected. When the computing load exceeds the preset load threshold, the predicted deviation and the edge strategy space are sent to the adjacent edge computing device to calculate the water supply adjustment amount.

[0013] Water supply control commands are generated based on the water supply adjustment amount and sent to the execution equipment at the water supply network nodes for water supply regulation.

[0014] In edge computing devices, water supply status data is compared with preset water supply target values ​​to calculate the status deviation. Temporal difference operations are performed on the status deviations from multiple consecutive acquisition cycles in the water supply status data to obtain the status change rate, including:

[0015] The changing trends of water supply status data in different acquisition cycles are used to obtain a status feature sequence;

[0016] Fluctuation analysis is performed on the state feature sequence to obtain the fluctuation intensity value. The fluctuation contribution is calculated based on the fluctuation intensity value. A dynamic weighting factor is generated based on the magnitude of the fluctuation contribution.

[0017] The state feature sequence is weighted by a dynamic weighting factor to obtain a weighted state sequence. The weighted state sequence is then compared with a preset water supply target value to obtain the state deviation.

[0018] The state deviation of the water supply status data in multiple consecutive collection cycles is recursively decomposed to obtain multi-scale features. The numerical distribution of the multi-scale features is calculated to obtain an importance score. The multi-scale features are weighted and combined according to the importance score to obtain a combined feature vector.

[0019] The size of the difference window is determined based on the numerical distribution of the combined feature vectors, and the state change rate is obtained by performing temporal difference operations on the combined feature vectors.

[0020] The system receives the global water supply scheduling strategy from the central computer room, determines the deviation interval to which the state deviation belongs based on the state deviation, extracts the control amplitude constraint corresponding to the deviation interval from the global water supply scheduling strategy, and generates the edge strategy space, including:

[0021] The system receives the global water supply scheduling strategy from the central computer room, extracts the control rules from the global water supply scheduling strategy, and constructs a control mapping structure that includes the mapping relationship between deviation range and control amplitude constraints based on the control rules.

[0022] The state deviation is converted into a feature vector, the numerical distribution of the feature vector is calculated to obtain the distribution density feature, and hierarchical features are constructed based on the distribution density feature;

[0023] Based on hierarchical features, interval matching is performed in the control mapping structure to determine the deviation interval to which the state deviation belongs, and the control amplitude constraint corresponding to the deviation interval is extracted from the global water supply scheduling strategy.

[0024] Calculate the constraint strength value of the control amplitude constraint, and generate constraint subsets of different strength levels according to the magnitude of the constraint strength value;

[0025] The edge policy space is generated by combining subsets of constraints of different strength levels in descending order of constraint strength value.

[0026] The state deviation is predicted based on the rate of change of state to obtain the state deviation for the future time period. The predicted deviation is obtained by weighted summation of the state deviation and the state deviation for the future time period.

[0027] The rate of change of state is constructed into a trend feature sequence according to the time sequence relationship, and the trend feature sequence is decomposed into periodic components and random components by multi-scale decomposition.

[0028] The periodic component is analyzed to obtain the periodic variation law, and the random component is analyzed to obtain the fluctuation variation law. Based on the periodic variation law and the fluctuation variation law, the state deviation is predicted to obtain the state deviation in the future period.

[0029] The state deviation is combined with the state deviation of the future time period to form a time series data sequence, and the fluctuation amplitude of the time series data sequence is calculated to obtain the time series fluctuation intensity value.

[0030] The fluctuation weights of the state deviation and the future time period state deviation are calculated based on the temporal fluctuation intensity value. The state deviation and the future time period state deviation are then weighted and summed according to the fluctuation weights to obtain the predicted deviation.

[0031] Within the control range constraints defined by the edge strategy space, the water supply adjustment amount calculated based on the predicted deviation includes:

[0032] A deviation feature sequence is constructed based on the predicted deviation amount, and the numerical distribution of the deviation feature sequence is calculated to obtain the fluctuation characteristics. The changing trend of the predicted deviation amount is analyzed based on the fluctuation characteristics.

[0033] The control direction is determined based on the changing trend, and the control amplitude constraint range corresponding to the control direction is obtained in the edge strategy space. The predicted deviation is then mapped to the control amplitude constraint range.

[0034] The control amplitude constraint range is divided into multiple control sub-intervals, and the control gain value corresponding to each control sub-interval is calculated based on the fluctuation characteristics.

[0035] The predicted deviation is allocated according to the control sub-interval, and multiplied with the corresponding control gain value to obtain the control calculation result;

[0036] The water supply adjustment amount is obtained by normalizing the control calculation results within the control range constraint.

[0037] The computational load of edge computing devices is detected. When the computational load exceeds a preset load threshold, the predicted deviation and the edge strategy space are sent to adjacent edge computing devices for water supply adjustment calculation, including:

[0038] Obtain resource usage information of edge computing devices, construct a computing resource usage sequence based on the resource usage information, and calculate the computing load based on the computing resource usage sequence;

[0039] The computing load is compared with a preset load threshold. When the computing load exceeds the preset load threshold, the computing resource status of adjacent edge computing devices is obtained.

[0040] Based on the computing resource status, a load balancing index is constructed. Adjacent edge computing devices are sorted according to the load balancing index, and the device with the best load balancing index among the adjacent edge computing devices is selected as the target device.

[0041] The predicted deviation and the edge strategy space are sent to the target device to calculate the water supply adjustment.

[0042] Based on the water supply adjustment amount, a water supply control command is generated and sent to the execution equipment at the water supply network node to perform water supply regulation, including:

[0043] Obtain the operating status information of the water supply network nodes, determine the execution constraints based on the operating status information, and map the water supply adjustment amount to the range of the execution constraints;

[0044] A water supply control command is generated based on the mapped water supply adjustment amount, the communication status of the water supply network node is obtained, and a communication link is selected according to the communication status.

[0045] The water supply control command is segmented and encoded to obtain a control data packet, and the transmission rate of the control data packet is adjusted based on the communication status.

[0046] The control data packets are sent to the execution device via the communication link, and the execution status feedback information returned by the execution device is received.

[0047] Verify the execution status of water supply control commands based on the execution status feedback information, and complete the water supply regulation of water supply network nodes.

[0048] This invention provides a water supply strategy calculation system that couples front-end sensing and edge computing, the system comprising:

[0049] The status acquisition module is used to collect water supply status data of water supply network nodes and transmit it to edge computing devices;

[0050] The deviation calculation module is set in the edge computing device. It is used to compare the water supply status data with the preset water supply target value to calculate the status deviation. It performs time-series difference operation on the status deviation of multiple consecutive collection cycles in the water supply status data to obtain the status change rate.

[0051] The strategy space generation module is used to receive the global water supply scheduling strategy from the central computer room, determine the deviation interval to which the state deviation belongs based on the state deviation, extract the control amplitude constraint corresponding to the deviation interval from the global water supply scheduling strategy, and generate the edge strategy space.

[0052] The deviation prediction module is used to predict the state deviation based on the state change rate to obtain the state deviation in the future time period, and to obtain the predicted deviation by weighted summation of the state deviation and the state deviation in the future time period.

[0053] The strategy calculation module is used to calculate the water supply adjustment amount based on the predicted deviation within the control range limited by the edge strategy space.

[0054] The load monitoring module is used to simultaneously detect the computing load of edge computing devices. When the computing load exceeds the preset load threshold, the predicted deviation and the edge strategy space are sent to the adjacent edge computing devices to calculate the water supply adjustment amount.

[0055] The control execution module is used to generate water supply control commands based on the water supply adjustment amount and send them to the execution equipment at the water supply network nodes for water supply regulation.

[0056] One technical solution provided in this embodiment of the invention is an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0057] One technical solution provided in this embodiment of the invention is a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps in any of the aforementioned methods.

[0058] This invention, by combining historical data with time-series differential calculations and trend predictions in edge computing devices, can detect changes in water supply status in advance, improving the foresight of control decisions. By receiving global water supply scheduling strategies from the central computer room and generating an edge strategy space, it achieves effective coordination between edge control and global scheduling. By monitoring the computing load of edge computing devices and migrating computing tasks to adjacent devices when the load is too high, it solves the problem of limited edge computing resources. By using predicted deviations to calculate water supply strategies, it improves the accuracy and real-time performance of water supply regulation, ultimately achieving precise control of water supply network nodes and ensuring the stable operation of the water supply system. Attached Figure Description

[0059] Figure 1 A flowchart of a water supply strategy calculation method coupled with front-end perception and edge computing provided in an embodiment of the present invention;

[0060] Figure 2 This is a flowchart illustrating the generation process of the edge water supply scheduling strategy in an embodiment of the present invention. Detailed Implementation

[0061] like Figure 1 As shown, Figure 1 A flowchart of a water supply strategy calculation method coupled with front-end perception and edge computing provided in an embodiment of the present invention is shown. The method includes the following steps:

[0062] Collect water supply status data from water supply network nodes and transmit it to edge computing devices;

[0063] In edge computing devices, water supply status data is compared with preset water supply target values ​​to calculate the status deviation. Time-series difference operation is performed on the status deviation of multiple consecutive collection cycles in the water supply status data to obtain the status change rate.

[0064] The system receives the global water supply scheduling strategy from the central computer room, determines the deviation range to which the state deviation belongs based on the state deviation, extracts the control amplitude constraint corresponding to the deviation range from the global water supply scheduling strategy, and generates the edge strategy space.

[0065] The state deviation is predicted based on the state change rate to obtain the state deviation for the future period. The state deviation is then weighted and summed with the state deviation for the future period to obtain the predicted deviation.

[0066] Within the control range constraint defined by the edge strategy space, the water supply adjustment amount is obtained by calculating the water supply strategy based on the predicted deviation.

[0067] Simultaneously, the computing load of the edge computing device is detected. When the computing load exceeds the preset load threshold, the predicted deviation and the edge strategy space are sent to the adjacent edge computing device to calculate the water supply adjustment amount.

[0068] Water supply control commands are generated based on the water supply adjustment amount and sent to the execution equipment at the water supply network nodes for water supply regulation.

[0069] In edge computing devices, water supply status data is compared with preset water supply target values ​​to calculate the status deviation. Temporal difference operations are performed on the status deviations from multiple consecutive acquisition cycles in the water supply status data to obtain the status change rate, including:

[0070] The changing trends of water supply status data in different acquisition cycles are used to obtain a status feature sequence;

[0071] Fluctuation analysis is performed on the state feature sequence to obtain the fluctuation intensity value. The fluctuation contribution is calculated based on the fluctuation intensity value. A dynamic weighting factor is generated based on the magnitude of the fluctuation contribution.

[0072] The state feature sequence is weighted by a dynamic weighting factor to obtain a weighted state sequence. The weighted state sequence is then compared with a preset water supply target value to obtain the state deviation.

[0073] The state deviation of the water supply status data in multiple consecutive collection cycles is recursively decomposed to obtain multi-scale features. The numerical distribution of the multi-scale features is calculated to obtain an importance score. The multi-scale features are weighted and combined according to the importance score to obtain a combined feature vector.

[0074] The size of the difference window is determined based on the numerical distribution of the combined feature vectors, and the state change rate is obtained by performing temporal difference operations on the combined feature vectors.

[0075] The system acquires the changing trends of water supply status data across different acquisition cycles, calculates the data differences between adjacent time points, and marks the direction of change to obtain a status feature sequence. The direction of change is categorized into three types: rising, falling, and stable, determined by comparing the data differences between adjacent time points with a preset threshold. A difference greater than a positive threshold is considered rising, less than a negative threshold is considered falling, and a difference between the two is considered stable. For example, if the water pressure values ​​at two adjacent acquisition points are 0.43 MPa and 0.47 MPa, with a difference of 0.04 MPa, which is greater than the preset positive threshold of 0.02 MPa, then it is marked as an rising trend.

[0076] When performing fluctuation analysis on a state characteristic sequence, the frequency of occurrence of rising, falling, and stable states, as well as the duration of each state, are calculated. The fluctuation intensity value is calculated by combining the state transition frequency and the amplitude of state changes. For example, if the water pressure state undergoes 15 transitions from rising to falling and 12 transitions from falling to rising within a 6-hour period, with an average change amplitude of 0.05 MPa, the calculated fluctuation intensity value is 1.35. Based on the fluctuation intensity value, the fluctuation contribution, i.e., the degree of influence of each parameter fluctuation on the overall water supply stability, is further calculated. The fluctuation contribution is obtained by comparing the fluctuation intensity value with historical statistical reference values, with a value ranging from 0 to 1. The larger the fluctuation contribution, the greater the impact of the parameter fluctuation on water supply stability.

[0077] The dynamic weighting factor is generated based on the contribution of fluctuations, using a non-linear mapping method. The numerical range of the weighting factor is 0.1 to 2.0; the greater the contribution of fluctuations, the larger the corresponding dynamic weighting factor. For example, if the contribution of fluctuations is 0.8, the dynamic weighting factor obtained through mapping is 1.6. The state feature sequence is then weighted with the dynamic weighting factor to obtain a weighted state sequence. The weighting calculation process involves multiplying each value in the state feature sequence by the corresponding dynamic weighting factor. The weighted state sequence better reflects the actual impact of parameter fluctuations on water supply.

[0078] The weighted state sequence is compared with the preset water supply target value to calculate the state deviation, which represents the difference between the current water supply state and the target state. For water pressure parameters, if the weighted state value is 0.50 MPa and the target value is 0.45 MPa, then the state deviation is 0.05 MPa.

[0079] A recursive decomposition of the state deviation quantities across multiple consecutive acquisition cycles yields multi-scale features. The recursive decomposition process employs wavelet decomposition to separate the state deviation sequence into different frequency components, including high-frequency fluctuation components and low-frequency trend components. The decomposition depth is three levels, resulting in three sets of frequency components. Statistical characteristics such as variance, kurtosis, and entropy of each frequency component are calculated to form a multi-scale feature representation.

[0080] Importance scores are obtained by calculating the numerical distribution of multi-scale features. The importance score reflects the contribution of each frequency component to the state change. The calculation considers the energy proportion and information gain of the features, with scores ranging from 0 to 10. High-frequency components typically reflect noise and short-term fluctuations, while low-frequency components reflect changes in water supply trends. The multi-scale features are then weighted and combined according to their importance scores to obtain a combined feature vector. Features with higher importance scores have greater weights in this weighted combination.

[0081] The difference window size is determined based on the numerical distribution of the combined feature vectors. A larger standard deviation indicates greater data volatility, requiring a larger window for smoothing; a smaller standard deviation indicates relatively stable data, allowing for a smaller window. The difference window size ranges from 3 to 12 acquisition cycles. For example, if the standard deviation of the combined feature vectors is 0.12, the resulting difference window size is 6 acquisition cycles.

[0082] A temporal difference operation is performed on the combined feature vectors to calculate the ratio of the change within a specific window size to time, yielding the rate of change. The rate of change reflects the speed at which water supply state parameters change and is an important indicator of water supply stability and responsiveness. A positive rate of change indicates an increase in parameter value, while a negative value indicates a decrease; the absolute value reflects the rate of change.

[0083] This invention achieves deep integration of front-end sensing and computing through edge computing devices, enabling efficient processing and analysis of water supply status data. By performing multi-dimensional analysis and processing of water supply status data, fluctuation characteristics can be accurately captured and the rate of change of status can be calculated, providing an important basis for optimizing water supply strategies. This method can quickly respond to changes in water supply status, effectively improving the stability and energy efficiency of water supply operation, reducing response time to abnormal events, and reducing energy consumption. The implementation of the method does not rely on a centralized server, reducing network dependence, improving robustness, and making it suitable for water supply management in various complex environments.

[0084] like Figure 2 As shown, the system receives the global water supply scheduling strategy from the central computer room, determines the deviation interval to which the state deviation belongs based on the state deviation, extracts the control amplitude constraint corresponding to the deviation interval from the global water supply scheduling strategy, and generates the edge strategy space, including:

[0085] The system receives the global water supply scheduling strategy from the central computer room, extracts the control rules from the global water supply scheduling strategy, and constructs a control mapping structure that includes the mapping relationship between deviation range and control amplitude constraints based on the control rules.

[0086] The state deviation is converted into a feature vector, the numerical distribution of the feature vector is calculated to obtain the distribution density feature, and hierarchical features are constructed based on the distribution density feature;

[0087] Based on hierarchical features, interval matching is performed in the control mapping structure to determine the deviation interval to which the state deviation belongs, and the control amplitude constraint corresponding to the deviation interval is extracted from the global water supply scheduling strategy.

[0088] Calculate the constraint strength value of the control amplitude constraint, and generate constraint subsets of different strength levels according to the magnitude of the constraint strength value;

[0089] The edge policy space is generated by combining subsets of constraints of different strength levels in descending order of constraint strength value.

[0090] The system receives the global water supply scheduling strategy from the central data center. This strategy is a comprehensive control plan based on the overall operation of the water supply network, containing various control rules, deviation range definitions, and corresponding control amplitude constraints. Control rules are typically stored in JSON or XML format, containing a mapping between deviation ranges and control amplitude constraints. Edge computing devices parse the received global water supply scheduling strategy data and extract the control rule information. The control rules in the global water supply scheduling strategy may include multiple conditional branches, such as a fine-tuning strategy for water pressure deviations within ±0.05MPa, a medium-level control strategy for deviations between ±0.05 and 0.15MPa, and a strong control strategy for deviations exceeding ±0.15MPa.

[0091] Edge computing devices construct a control mapping structure based on extracted control rules, containing a mapping relationship between deviation ranges and control amplitude constraints. This control mapping structure is implemented using hash tables or multi-level index trees, enabling rapid lookup of control amplitude constraints corresponding to specific deviation ranges. In the control mapping structure, each deviation range corresponds to a set of control amplitude constraints, which define the upper and lower limits for adjusting control parameters such as valve opening and pump frequency. Deviation ranges can be continuous numerical ranges or discrete state categories. Taking water pressure control as an example, deviation ranges can be divided into five levels: extremely low (-∞, -0.15MPa), slightly low (-0.15MPa, -0.05MPa), normal (-0.05MPa, 0.05MPa), slightly high (0.05MPa, 0.15MPa), and extremely high (0.15MPa, +∞).

[0092] In the process of converting state deviation quantities into feature vectors, edge computing devices perform feature extraction and transformation on the original state deviation quantities. Feature extraction includes calculating statistical features such as the mean, maximum, minimum, and rate of change of the state deviation quantities to form feature vectors. For multi-parameter water supply conditions, such as considering deviations in parameters like water pressure, flow rate, and water quality simultaneously, the feature extraction results for each parameter are combined to form a multi-dimensional feature vector. The dimension of the feature vector is typically between 10 and 20, which can comprehensively reflect the characteristics of the state deviation.

[0093] When calculating the numerical distribution of feature vectors to obtain distribution density features, edge computing devices use kernel density estimation methods to analyze the numerical distribution of each dimension of the feature vectors. Distribution density features reflect the degree of clustering of state deviations across numerical intervals and can be used to determine the central tendency and dispersion of deviations. Distribution density features typically include statistics such as mean, variance, kurtosis, and skewness, as well as key point location information of the distribution curve. Based on distribution density features, hierarchical features are constructed, organizing distribution characteristics into a hierarchical structure according to importance. The construction of hierarchical features adopts a bottom-up aggregation approach, with the bottom layer containing detailed distribution characteristics and the top layer containing generalized distribution characteristics, facilitating rapid location of relevant intervals in the control mapping structure.

[0094] When performing interval matching based on hierarchical features in the control mapping structure, edge computing devices employ a multi-level matching strategy. First, high-level information from the hierarchical features is used to quickly locate possible deviation intervals, and then detailed information from the lower levels is used for precise matching. The matching process uses a similarity calculation method to find the predefined deviation interval that is closest to the current hierarchical feature. After determining the deviation interval to which the state deviation belongs, the corresponding control amplitude constraint is extracted from the global water supply scheduling strategy. The control amplitude constraint typically includes information such as the adjustment direction of the control parameters, the adjustment step size range, and the maximum adjustment amplitude. For example, for a water pressure range of 0.05MPa-0.15MPa, the control amplitude constraint might specify a valve opening increase range of 5%-15% and a pump frequency decrease range of 3Hz-8Hz.

[0095] When calculating the constraint strength value of the control amplitude constraint, the edge computing device comprehensively considers the strictness of the constraint, the scope of influence, and the control effect. The calculation of the constraint strength value involves factors such as the number of constraint parameters, the size of the constraint interval, and the degree of influence of the constraint on the water supply status. The constraint strength value is usually normalized to between 0 and 100, with a larger value indicating a stronger constraint. The control amplitude constraint is graded according to the magnitude of the constraint strength value, generating constraint subsets of different strength levels. The division of constraint subsets usually uses uniform partitioning or clustering-based methods, grouping constraints of similar strength together. Common grading may include three levels: mild constraint (0-30), moderate constraint (30-70), and strong constraint (70-100).

[0096] The edge policy space is generated by combining subsets of constraints of different strength levels in descending order of constraint strength. The edge policy space is a multi-level constraint structure containing a complete constraint spectrum from strict to lenient constraints. Edge computing devices prioritize strong constraints, and only consider weak constraints after satisfying strong constraints, thus ensuring the satisfaction of critical constraints while also considering other constraints. During the generation of the edge policy space, the compatibility between constraints is also considered, and conflicting constraint combinations are removed.

[0097] This invention achieves fine-grained decomposition and execution of water supply scheduling strategies from global to local levels through the collaborative work of edge computing devices and a central computer room. Based on the received global strategy and combined with locally perceived water supply status deviations, the edge devices dynamically generate an edge strategy space adapted to the current scenario. This effectively improves the real-time performance and adaptability of water supply strategy execution, reduces network communication burden, and enhances the robustness of water supply regulation. Through the construction of a regulation mapping structure and the organization of hierarchical constraints, fine-grained management of the regulation strategy is achieved, enabling the implementation of regulation measures of appropriate intensity for different degrees of status deviations.

[0098] The state deviation is predicted based on the rate of change of state to obtain the state deviation for the future time period. The predicted deviation is obtained by weighted summation of the state deviation and the state deviation for the future time period.

[0099] The rate of change of state is constructed into a trend feature sequence according to the time sequence relationship, and the trend feature sequence is decomposed into periodic components and random components by multi-scale decomposition.

[0100] The periodic component is analyzed to obtain the periodic variation law, and the random component is analyzed to obtain the fluctuation variation law. Based on the periodic variation law and the fluctuation variation law, the state deviation is predicted to obtain the state deviation in the future period.

[0101] The state deviation is combined with the state deviation of the future time period to form a time series data sequence, and the fluctuation amplitude of the time series data sequence is calculated to obtain the time series fluctuation intensity value.

[0102] The fluctuation weights of the state deviation and the future time period state deviation are calculated based on the temporal fluctuation intensity value. The state deviation and the future time period state deviation are then weighted and summed according to the fluctuation weights to obtain the predicted deviation.

[0103] The obtained state change rates are used to construct a trend feature sequence based on time-series relationships. The state change rate is an indicator reflecting the rate of change of water supply parameters, such as a water pressure change of 0.02 MPa per hour or a flow rate change of 5 m³ per hour. 3 / h, etc. A trend feature sequence is a data sequence formed by arranging the rate of change of state at multiple consecutive time points in chronological order. Edge computing devices typically collect state change rate data every 5 minutes over the past 24 hours, totaling 288 data points to form a trend feature sequence.

[0104] When performing multi-scale decomposition on trend feature sequences to obtain periodic and random components, edge computing devices employ the empirical mode decomposition (EMD) method. This method does not require pre-defined basis functions and is suitable for nonlinear and non-stationary time series analysis. The decomposition process splits the trend feature sequence into several intrinsic mode functions (IMFs) and a residue term. The first few IMFs represent high-frequency random fluctuations, the middle IMFs represent periodic changes, and the final residue term represents the long-term trend. Typically, the trend feature sequence is decomposed into 5-8 IMFs and 1 residue term.

[0105] Periodic components reflect the regular changes in water supply conditions, such as the alternation of peak and off-peak water usage periods. Edge computing devices analyze these periodic components using autocorrelation and spectral analysis methods to identify key periodic characteristics. For urban water supply, typical periods include the 24-hour daily variation cycle and the weekly cycle caused by differences between weekdays and weekends. The analyzed periodic variation patterns include cycle length, peak time, and amplitude characteristics.

[0106] Random components reflect the irregular fluctuations in water supply conditions, influenced by various random factors. Edge computing devices analyze random components using statistical analysis methods, calculating their statistical characteristics such as variance, kurtosis, and skewness, and analyzing their temporal correlation through conditional entropy. The fluctuation variation patterns describe the intensity, duration, and transition probability of random fluctuations.

[0107] When predicting state deviations based on periodic and fluctuating patterns, edge computing devices employ deterministic and probabilistic prediction methods to handle periodic and random components, respectively. For periodic components, Fourier analysis is used to extract the main periodic terms, and future periodic changes are predicted through frequency domain feature reconstruction. For random components, a probability distribution model is used for modeling and prediction. This probability distribution model is based on kernel density estimation, utilizing the distribution characteristics of historical random components to construct a conditional probability density function. Probability density estimation is performed on historical data of the random components to obtain their conditional distribution characteristics under different states. The historical random components are divided into several intervals according to their magnitude, such as -0.03MPa to -0.02MPa, -0.02MPa to -0.01MPa, etc., and the probability of the current state transitioning to each interval is calculated. Based on these conditional probabilities and combined with the current state of the random components, a probability distribution for future random components is generated. During the prediction process, samples are drawn from this probability distribution through Monte Carlo sampling as predicted values ​​for the random components.

[0108] During the prediction process, the edge computing device first determines a baseline value based on the current state deviation and state change rate, and then superimposes the influence of periodic and random predictions to obtain the predicted state deviation value for the future period. For example, if the current water pressure deviation is 0.08 MPa and the state change rate is -0.01 MPa / hour, and the predicted influence of periodic factors after 1 hour is -0.005 MPa, and the influence of random fluctuations is ±0.003 MPa, then the predicted water pressure deviation after 1 hour is approximately 0.065 MPa.

[0109] The current state deviation is combined with the predicted future state deviation to form a time series data sequence. This sequence contains the current value and multiple predicted values, forming a continuous data sequence spanning from now to the predicted end point. The volatility amplitude of this time series data sequence is calculated to obtain the time series volatility intensity value. The volatility amplitude is characterized by calculating the difference between the maximum and minimum values ​​in the sequence, as well as the standard deviation. The time series volatility intensity value is a normalized representation of the volatility amplitude, ranging from 0 to 1; a larger value indicates stronger volatility.

[0110] The fluctuation weights of state deviation and future state deviation are calculated based on the intensity of time-series fluctuations. These fluctuation weights determine the relative importance of the current and predicted values ​​when calculating the predicted deviation. A non-linear mapping method is used to calculate the fluctuation weights: when the fluctuation intensity is small, the future predicted value receives a higher weight; when the fluctuation intensity is large, the current measured value receives a higher weight. Under stable conditions, the predicted value has high reliability and should be given a higher weight; under conditions of severe fluctuation, the prediction uncertainty increases, and the prediction should rely more heavily on the current measured value.

[0111] The predicted deviation is obtained by weighting and summing the current state deviation with the state deviation for each future period according to fluctuation weights. In the weighted summation process, the current state deviation and the state deviation for each prediction period are multiplied by their respective weights, and then summed to obtain the predicted deviation. The predicted deviation comprehensively considers both the current state and future trends, thus providing a more complete reflection of the development trend of the water supply status.

[0112] For example, if the current water pressure deviation at a water supply station is 0.06 MPa, and the predicted deviations for the next 30, 60, and 90 minutes are 0.08 MPa, 0.09 MPa, and 0.08 MPa, respectively, the calculated time-series fluctuation intensity value is 0.3. Based on the fluctuation intensity value, the current value is weighted at 0.4, and the three predicted values ​​are weighted at 0.25, 0.2, and 0.15, respectively. Through weighted summation, the predicted deviation is obtained as 0.074 MPa. This predicted deviation will be used in the formulation of subsequent water supply strategies to guide the generation of control decisions.

[0113] This invention achieves accurate prediction of future trends in water supply status through predictive analysis based on the rate of change of state, providing a forward-looking basis for water supply strategy calculations. A dynamic weight allocation mechanism based on the intensity of temporal fluctuations makes the prediction results more robust under conditions of severe fluctuations. Implemented on edge computing devices, it does not rely on cloud resources, offering fast response speeds and strong adaptability, making it particularly suitable for water supply control scenarios with high real-time requirements. Accurate prediction of future states allows for the early detection of potential problems and timely adjustment of strategies, effectively reducing water supply fluctuations and energy waste, and improving water supply quality and efficiency.

[0114] Within the control range constraints defined by the edge strategy space, the water supply adjustment amount calculated based on the predicted deviation includes:

[0115] A deviation feature sequence is constructed based on the predicted deviation amount, and the numerical distribution of the deviation feature sequence is calculated to obtain the fluctuation characteristics. The changing trend of the predicted deviation amount is analyzed based on the fluctuation characteristics.

[0116] The control direction is determined based on the changing trend, and the control amplitude constraint range corresponding to the control direction is obtained in the edge strategy space. The predicted deviation is then mapped to the control amplitude constraint range.

[0117] The control amplitude constraint range is divided into multiple control sub-intervals, and the control gain value corresponding to each control sub-interval is calculated based on the fluctuation characteristics.

[0118] The predicted deviation is allocated according to the control sub-interval, and multiplied with the corresponding control gain value to obtain the control calculation result;

[0119] The water supply adjustment amount is obtained by normalizing the control calculation results within the control range constraint.

[0120] A deviation feature sequence is constructed based on the predicted deviation amount. This sequence includes the current predicted deviation amount and its historical values, selecting predicted deviation amounts from the most recent period, such as hourly predicted deviation data over the past 24 hours, for a total of 24 data points. The deviation feature sequence reflects the temporal evolution of the predicted deviation amount and is an important basis for analyzing the stability and trends of water supply conditions.

[0121] When calculating the numerical distribution of the deviation feature sequence to obtain the fluctuation characteristics, the edge computing device performs statistical analysis on the deviation feature sequence. The fluctuation characteristics include statistics such as mean, variance, skewness, and kurtosis, reflecting the central tendency, dispersion, and distribution pattern of the predicted deviation. The edge computing device sorts the data in the deviation feature sequence, statistically analyzes the frequency distribution of each numerical interval, and calculates relevant statistics. For example, if the predicted deviation sequence for a water supply station has a mean of 0.05 MPa, a variance of 0.0004, a skewness of 1.2, and a kurtosis of 3.5, it indicates that the overall deviation values ​​are high and the distribution is not concentrated, showing strong fluctuations.

[0122] When predicting the trend of deviation based on fluctuation characteristics, edge computing devices employ trend extraction algorithms to smooth the deviation feature sequence, filtering out short-term fluctuation interference. By calculating the first-order difference and cumulative sum of the sequence, they determine the increasing or decreasing trend and acceleration characteristics of the sequence. Change trends are generally classified into three basic types: rising, falling, and oscillating. Each type can be further subdivided into acceleration and deceleration states. By analyzing the rate of change and acceleration of deviation in the most recent period, edge computing devices predict the direction and rate of change of deviation in the near future.

[0123] When determining the control direction based on the changing trend, the edge computing device adopts a reverse control strategy. When the predicted deviation shows an upward trend, the control direction is negative, i.e., reducing the water supply parameters; when the predicted deviation shows a downward trend, the control direction is positive, i.e., increasing the water supply parameters; when the predicted deviation shows an oscillating trend, the control direction is stable, i.e., maintaining the current water supply parameters and making fine adjustments. The determination of the control direction takes into account the current value and changing trend of the predicted deviation, aiming to proactively address the risk of deviation expansion.

[0124] When obtaining the control amplitude constraint range corresponding to the control direction in the edge strategy space, the edge computing device calls the preset strategy constraint table. The strategy constraint table defines the control amplitude limits under different water supply parameters and different control directions. For example, for water pressure parameters, the amplitude constraint range for positive adjustment is 0 to 0.2 MPa, the amplitude constraint range for negative adjustment is -0.15 to 0 MPa, and the amplitude constraint range for stable adjustment is -0.05 to 0.05 MPa.

[0125] When mapping the predicted deviation to the control amplitude constraint range, the edge computing device uses a nonlinear mapping function. This mapping function considers both the absolute magnitude and relative rate of change of the predicted deviation, uniformly converting deviations of different dimensions and ranges to the control amplitude constraint range. The mapping process employs a piecewise function; linear mapping is used when the deviation is small to maintain sensitivity, while logarithmic mapping is used when the deviation is large to avoid over-control. For example, when the predicted deviation is 0.08 MPa and the control direction is negative, the initial control amount calculated through the mapping function is -0.12 MPa.

[0126] When dividing the control amplitude constraint range into multiple control sub-intervals, edge computing devices typically employ equal-interval or equal-frequency segmentation methods. Equal-interval segmentation evenly divides the control amplitude constraint range into several sub-intervals, suitable for scenarios with consistent control accuracy requirements. Equal-frequency segmentation, based on the distribution of historical control data, ensures that each sub-interval contains approximately the same number of historical data points, suitable for scenarios with uneven control sensitivity. Typically, the control amplitude constraint range is divided into 3-5 sub-intervals. For example, the negative control range of -0.15 to 0 MPa can be divided into three sub-intervals: -0.15 to -0.1 MPa, -0.1 to -0.05 MPa, and -0.05 to 0 MPa.

[0127] When calculating the control gain value corresponding to each control sub-interval based on fluctuation characteristics, the edge computing device adjusts the control sensitivity of each sub-interval according to the fluctuation characteristics of the deviation feature sequence. When fluctuations are strong, the gain value near the boundary interval is reduced to improve control stability; when fluctuations are small, the gain value in the middle interval is increased to improve the control response speed. The calculation of the control gain value takes into account the combined effects of fluctuation variance, skewness, and kurtosis, and is usually an adjustment coefficient between 0.7 and 1.3. For example, if the gain values ​​of the three sub-intervals in a certain calculation are 0.8, 1.0, and 1.2, it indicates that a differentiated control strategy is adopted for different degrees of deviation.

[0128] When allocating the predicted deviation amount according to the control sub-intervals, the edge computing device uses a segmented mapping method to decompose the mapped initial control amount into each control sub-interval and calculate the control amount that each sub-interval should bear. The allocation process follows the principle of "filling the low threshold interval before entering the high threshold interval" to ensure the gradualness and stability of the control. For example, if the initial control amount is -0.12MPa, after being allocated to three sub-intervals, the first interval receives -0.05MPa, the second interval receives -0.05MPa, and the third interval receives -0.02MPa.

[0129] When multiplying the corresponding control gain value to obtain the control calculation result, the edge computing device multiplies the control amount allocated to each sub-interval with the corresponding gain value, and then sums them to obtain the control calculation result. This step introduces a dynamic control mechanism based on fluctuation characteristics, enabling the control intensity to adaptively adjust according to the stability of the water supply status. For example, the calculation results for the three sub-intervals are -0.04MPa, -0.05MPa, and -0.024MPa, respectively, and the total calculation result is -0.114MPa.

[0130] When normalizing the calculated control results within the control range to obtain the water supply adjustment amount, the edge computing device checks whether the calculation result exceeds the constraint range and performs necessary truncation and smoothing. The normalization process ensures that the final water supply adjustment amount is strictly within the safety constraint range and has an appropriate control strength. For example, if the calculated result is -0.114 MPa and the constraint range is -0.15 to 0 MPa, the normalized water supply adjustment amount is -0.114 MPa, and no truncation is required; if the calculated result exceeds the constraint range, the constraint boundary value is taken as the final adjustment amount.

[0131] The water supply adjustment quantity, as a control signal output by the edge computing device, is directly used to guide the parameter adjustment of the water supply equipment. The edge computing device sends the calculated water supply adjustment quantity to the execution unit, which then adjusts control parameters such as pump speed and valve opening accordingly based on the received adjustment quantity, thereby achieving closed-loop control of the water supply status.

[0132] This invention employs a water supply strategy calculation method based on predicted deviation, fully leveraging the synergistic advantages of front-end sensing and edge computing. By constructing a deviation feature sequence analysis to predict the fluctuation characteristics and trends of the deviation, it achieves forward-looking judgment of future changes in the water supply status. The segmented control and dynamic gain mechanism based on fluctuation characteristics make the control strategy more adaptive, automatically adjusting the control intensity according to the stability of the water supply status, avoiding over-control and under-control problems. The constraint mechanism of the edge strategy space ensures the safety of the control, preventing water supply fluctuations caused by extreme control.

[0133] The computational load of edge computing devices is detected. When the computational load exceeds a preset load threshold, the predicted deviation and the edge strategy space are sent to adjacent edge computing devices for water supply adjustment calculation, including:

[0134] Obtain resource usage information of edge computing devices, construct a computing resource usage sequence based on the resource usage information, and calculate the computing load based on the computing resource usage sequence;

[0135] The computing load is compared with a preset load threshold. When the computing load exceeds the preset load threshold, the computing resource status of adjacent edge computing devices is obtained.

[0136] Based on the computing resource status, a load balancing index is constructed. Adjacent edge computing devices are sorted according to the load balancing index, and the device with the best load balancing index among the adjacent edge computing devices is selected as the target device.

[0137] The predicted deviation and the edge strategy space are sent to the target device to calculate the water supply adjustment.

[0138] Resource usage information is acquired, including key metrics such as processor utilization, memory utilization, storage utilization, network bandwidth utilization, and task queue length. This data is collected every 5 seconds, reflecting the workload of computing resources, data processing capacity limitations, data transfer constraints, and task backlog.

[0139] Edge computing devices arrange resource usage data from multiple consecutive time points in chronological order to form a computing resource usage sequence. This sequence typically contains resource usage data from 120 sampling points within the last 10 minutes. Each sequence element is a multi-dimensional vector containing multiple indicator values, organized into a time-series matrix by timestamps, serving as input for computing load assessment.

[0140] The computing device employs a comprehensive weighted evaluation method to calculate its current load. Weights are assigned to different resource metrics: processor utilization 0.4, memory utilization 0.3, storage utilization 0.1, network bandwidth utilization 0.1, and task queue length 0.1. The weighted average of these metrics is calculated to obtain the comprehensive load index. Furthermore, the time-weighted trend of resource usage is considered, with more recent data receiving higher weight. The final load value is a combination of the comprehensive load index and the time-weighted average load, ranging from 0 to 100, with higher values ​​indicating higher load.

[0141] Edge computing devices are set with a preset load threshold of 75. When the computing load exceeds this threshold, the device is considered to be under high load, and a load balancing mechanism needs to be activated. The device not only compares the absolute value of the computing load but also considers the load change trend. If the load does not exceed the threshold but shows a rapid upward trend (the growth rate exceeds 20% in the last 3 minutes), the load balancing mechanism will also be triggered.

[0142] When the load exceeds the threshold, the device obtains the resource status of adjacent edge computing devices through the communication interface. Adjacent devices refer to other edge devices within the same water supply network area that are physically close and have stable network connections. Resource status includes information such as current computing load, number of available processor cores, available memory capacity, task queue length, and network connection quality.

[0143] Edge computing devices construct load balancing metrics based on the resource status of adjacent devices, comprehensively considering factors such as computing power, current load, load stability, network connectivity, and historical performance. Each factor is normalized to a value between 0 and 1, and its weight is adjusted according to the characteristics of the application scenario. The resulting comprehensive value serves as the load balancing metric.

[0144] The system employs a multi-objective optimization method to rank adjacent devices, considering load balancing metrics, task migration costs, and risks. Migration costs include data transmission time, device switching latency, and coordination overhead; risks include communication interruption probability, device failure rate, and the likelihood of sudden load fluctuations. Devices are calculated with a comprehensive score, ranked from highest to lowest; a higher score indicates that the device is more suitable for receiving computing tasks.

[0145] When selecting a target device, the edge computing device sends confirmation requests to the top three ranked devices to verify whether the resource status has changed. If the status of the top-ranked device deteriorates, the second or third-ranked device is considered. During the confirmation process, the matching degree between task characteristics and device features is also evaluated, and the device most suitable for the current task is selected as the target device.

[0146] When the predicted deviation and edge strategy space are sent to the target device, the data is packaged. The data includes the current predicted deviation, historical deviation sequences, edge strategy space parameters, and computational context information. The edge strategy space includes the control direction, control amplitude constraints, sub-interval division rules, and gain calculation parameters, ensuring consistency of calculation results across different devices. The packaged data is transmitted through a secure communication channel, supporting data compression and breakpoint resumption.

[0147] After receiving the data, the target device verifies its integrity and timeliness, analyzes the predicted deviation and edge strategy space, calculates the water supply adjustment, and returns the results to the originating device. The originating device verifies and adapts the results to ensure they meet the water supply control requirements and are applied to actual control. If the results are not received on time or verification fails, the originating device activates its local emergency calculation mechanism to ensure control continuity.

[0148] For example, if an edge computing device in a certain area detects CPU utilization of 85%, memory utilization of 70%, storage utilization of 65%, network bandwidth utilization of 60%, and a task queue length of 10, the calculated overall load is 78, exceeding the threshold of 75. The resource status of three adjacent devices is obtained, with load balancing metrics of 0.72, 0.85, and 0.63 respectively. The device with a metric of 0.85 is identified as the target device. A predicted deviation of 0.06 MPa and policy space data are sent to this device. A returned water supply adjustment of -0.04 MPa is received, completing the load balancing calculation.

[0149] The edge computing load balancing technology proposed in this invention enables dynamic scheduling and collaborative computing of water supply strategy calculation tasks among edge computing devices. Through multi-dimensional resource monitoring and comprehensive load assessment, the load status of edge devices is accurately grasped, and resource bottlenecks are identified in a timely manner. Based on multi-factor load balancing indicators and multi-objective optimization ranking, intelligent allocation of computing tasks is achieved, effectively balancing the resource utilization of the edge computing network. Data packaging and transmission and result verification mechanisms during task migration ensure the consistency and reliability of the calculation results.

[0150] Based on the water supply adjustment amount, a water supply control command is generated and sent to the execution equipment at the water supply network node to perform water supply regulation, including:

[0151] Obtain the operating status information of the water supply network nodes, determine the execution constraints based on the operating status information, and map the water supply adjustment amount to the range of the execution constraints;

[0152] A water supply control command is generated based on the mapped water supply adjustment amount, the communication status of the water supply network node is obtained, and a communication link is selected according to the communication status.

[0153] The water supply control command is segmented and encoded to obtain a control data packet, and the transmission rate of the control data packet is adjusted based on the communication status.

[0154] The control data packets are sent to the execution device via the communication link, and the execution status feedback information returned by the execution device is received.

[0155] Verify the execution status of water supply control commands based on the execution status feedback information, and complete the water supply regulation of water supply network nodes.

[0156] The system acquires operational status information from water supply network nodes, including key data such as water pressure, flow rate, valve opening, pump speed, water quality parameters, and energy consumption. This data is collected every minute by node sensors, filtered and calibrated by a data preprocessing module to remove outliers and supplement missing data, ensuring data accuracy and completeness.

[0157] The execution constraints determined based on operational status information include physical constraints and operational constraints. Physical constraints refer to the characteristic limitations of the executing equipment, such as valve opening range of 0-100% and pump speed range of 600-1800 rpm. Operational constraints refer to the safe operation requirements of the water supply network, such as the safe water pressure range of nodes of 0.2-0.8 MPa and the flow rate change rate not exceeding 30% / hour. Edge computing devices consider the characteristics of the nodes, such as pipe material, diameter, and service life, to dynamically calculate precise constraints.

[0158] A piecewise linear mapping method is used to map water supply adjustments to the range of execution constraints, converting changes in pressure or flow rate into control parameters for the actuators. For adjustments exceeding the constraints, boundary truncation and gradient decay strategies are employed to ensure the mapping result satisfies the constraints while approximating the original adjustment target as closely as possible. The mapping process considers the response characteristics of the actuators, appropriately amplifying the adjustment for actuators with slow responses.

[0159] Control commands are generated based on the mapped water supply adjustment. Each command includes a command header, execution object identifier, operation type, parameter values, execution time window, and checksum. Different execution devices correspond to different operation types; for example, valves are adjusted for opening degree, and water pumps are adjusted for speed. Edge computing devices select control precision and execution time window based on device characteristics to generate structured commands. For multiple devices requiring coordinated control, command priorities and timing relationships are set.

[0160] The communication status of water supply network nodes is acquired, including signal strength, communication latency, packet loss rate, bandwidth utilization, and communication quality stability. Edge computing devices assess link status by sending probe packets or analyzing communication logs, with results categorized into four levels: excellent, good, average, and poor, corresponding to different communication strategies.

[0161] When selecting a communication link based on communication status, multiple communication methods supported by the water supply network nodes are considered, including wired Ethernet, wireless LAN, mobile cellular network, and low-power wide area network. Edge computing devices select the optimal link based on communication evaluation results and the urgency of the command. The selection process comprehensively considers reliability, latency, and energy consumption; urgent commands prioritize reliability, while routine adjustments take into account both energy consumption and latency.

[0162] Water supply control commands are segmented and encoded according to the maximum transmission unit size of the communication link. Each segment constitutes a data packet, containing a sequence number, total number of packets, payload, and checksum. An adaptive encoding scheme is adopted, selecting an appropriate encoding method based on the link characteristics to improve transmission efficiency and anti-interference capability.

[0163] The transmission rate is adjusted based on communication status. A high-rate mode is used when communication is good, and the rate is reduced and a retransmission mechanism is added when communication is poor. The transmission rate adopts a dynamic adaptive strategy, optimizing transmission parameters based on real-time feedback to balance reliability and efficiency.

[0164] Control data packets are sent via a selected communication link, employing an acknowledgment and retransmission mechanism. The executing device is required to acknowledge each packet; if no acknowledgment is received within a timeout period, the packet is retransmitted. Important instructions may be sent simultaneously through multiple links, with the executing device merging and deduplicating them. Communication quality is dynamically monitored during transmission, and backup links are switched on when necessary.

[0165] The system receives execution status feedback from the execution device, including reception status, parsing results, execution progress, current parameters, and any exceptions. The execution device acknowledges receipt of the instruction immediately, periodically returns progress updates during execution, and returns the final result upon completion. Feedback information uses a similar encoding and transmission mechanism as the instructions.

[0166] The execution status feedback verifies the control command execution, compares the original command with the execution result, assesses deviations, and analyzes the causes. Successfully executed commands have their parameters recorded for strategy optimization; commands that fail to execute are handled according to their type, such as resending, parameter adjustment, or strategy switching. After verification, the node status record is updated, completing the water supply regulation operation.

[0167] For example, if the water pressure at a certain water supply network node is 0.65 MPa, which is higher than the target of 0.55 MPa, the calculated water supply adjustment is -0.1 MPa. The node's executing device is an electric regulating valve with an opening range of 0-100%, currently at 80%. The edge computing device maps the pressure adjustment to an opening adjustment of -15%, considering a minimum accuracy of 1%, with a target opening of 65%. After generating the control command, it selects the best wireless LAN signal for transmission, encodes it into 3 data packets, and sends them. The executing device returns confirmation and progress, and the final actual opening is 66%. The edge computing device verifies that the deviation is within the allowable range, completing the water supply adjustment.

[0168] The water supply control command generation and execution method provided by this invention achieves accurate conversion and reliable execution of water supply adjustment quantities into control commands. By acquiring the operating status and determining execution constraints, the rationality and feasibility of the control commands are ensured; by communication status awareness and link selection, the reliability of command transmission is improved; by segmented encoding and rate adjustment, data transmission efficiency is optimized; and by execution status feedback and verification, closed-loop control of water supply regulation is achieved. This effectively solves the problems of inaccurate commands, unreliable execution, and untimely feedback in traditional water supply control, improving the regulation accuracy and response speed of the water supply network, and providing technical support for the refined management of intelligent water supply networks.

[0169] The water supply strategy calculation system coupled with front-end sensing and edge computing provided in this embodiment of the invention includes:

[0170] The status acquisition module is used to collect water supply status data of water supply network nodes and transmit it to edge computing devices;

[0171] The deviation calculation module is set in the edge computing device. It is used to compare the water supply status data with the preset water supply target value to calculate the status deviation. It performs time-series difference operation on the status deviation of multiple consecutive collection cycles in the water supply status data to obtain the status change rate.

[0172] The strategy space generation module is used to receive the global water supply scheduling strategy from the central computer room, determine the deviation interval to which the state deviation belongs based on the state deviation, extract the control amplitude constraint corresponding to the deviation interval from the global water supply scheduling strategy, and generate the edge strategy space.

[0173] The deviation prediction module is used to predict the state deviation based on the state change rate to obtain the state deviation in the future time period, and to obtain the predicted deviation by weighted summation of the state deviation and the state deviation in the future time period.

[0174] The strategy calculation module is used to calculate the water supply adjustment amount based on the predicted deviation within the control range limited by the edge strategy space.

[0175] The load monitoring module is used to simultaneously detect the computing load of edge computing devices. When the computing load exceeds the preset load threshold, the predicted deviation and the edge strategy space are sent to the adjacent edge computing devices to calculate the water supply adjustment amount.

[0176] The control execution module is used to generate water supply control commands based on the water supply adjustment amount and send them to the execution equipment at the water supply network nodes for water supply regulation.

[0177] One technical solution provided in this embodiment of the invention is an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0178] One technical solution provided in this embodiment of the invention is a computer-readable storage medium storing a computer program, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0179] The specific embodiments described above are preferred embodiments of the present invention and are not intended to limit the specific scope of the present invention. The scope of the present invention includes, but is not limited to, these specific embodiments. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A water supply strategy calculation method coupled with front-end sensing and edge computing, characterized in that, Includes the following steps: Collect water supply status data from water supply network nodes and transmit it to edge computing devices; In edge computing devices, water supply status data is compared with preset water supply target values ​​to calculate the status deviation. Time-series difference operation is performed on the status deviation of multiple consecutive collection cycles in the water supply status data to obtain the status change rate. The system receives the global water supply scheduling strategy from the central computer room, determines the deviation range to which the state deviation belongs based on the state deviation, extracts the control amplitude constraint corresponding to the deviation range from the global water supply scheduling strategy, and generates the edge strategy space. The state deviation is predicted based on the state change rate to obtain the state deviation for the future period. The state deviation is then weighted and summed with the state deviation for the future period to obtain the predicted deviation. Within the control range constraint defined by the edge strategy space, the water supply adjustment amount is obtained by calculating the water supply strategy based on the predicted deviation. Simultaneously, the computing load of the edge computing device is detected. When the computing load exceeds the preset load threshold, the predicted deviation and the edge strategy space are sent to the adjacent edge computing device to calculate the water supply adjustment amount. Water supply control commands are generated based on the water supply adjustment amount and sent to the execution equipment at the water supply network nodes for water supply regulation.

2. The method according to claim 1, characterized in that, In edge computing devices, water supply status data is compared with preset water supply target values ​​to calculate the status deviation. Temporal difference operations are performed on the status deviations from multiple consecutive acquisition cycles in the water supply status data to obtain the status change rate, including: The changing trends of water supply status data in different acquisition cycles are used to obtain a status feature sequence; Fluctuation analysis is performed on the state feature sequence to obtain the fluctuation intensity value. The fluctuation contribution is calculated based on the fluctuation intensity value. A dynamic weighting factor is generated based on the magnitude of the fluctuation contribution. The state feature sequence is weighted by a dynamic weighting factor to obtain a weighted state sequence. The weighted state sequence is then compared with a preset water supply target value to obtain the state deviation. The state deviation of the water supply status data in multiple consecutive collection cycles is recursively decomposed to obtain multi-scale features. The numerical distribution of the multi-scale features is calculated to obtain an importance score. The multi-scale features are weighted and combined according to the importance score to obtain a combined feature vector. The size of the difference window is determined based on the numerical distribution of the combined feature vectors, and the state change rate is obtained by performing temporal difference operations on the combined feature vectors.

3. The method according to claim 1, characterized in that, The system receives the global water supply scheduling strategy from the central computer room, determines the deviation interval to which the state deviation belongs based on the state deviation, extracts the control amplitude constraint corresponding to the deviation interval from the global water supply scheduling strategy, and generates the edge strategy space, including: The system receives the global water supply scheduling strategy from the central computer room, extracts the control rules from the global water supply scheduling strategy, and constructs a control mapping structure that includes the mapping relationship between deviation range and control amplitude constraints based on the control rules. The state deviation is converted into a feature vector, the numerical distribution of the feature vector is calculated to obtain the distribution density feature, and hierarchical features are constructed based on the distribution density feature; Based on hierarchical features, interval matching is performed in the control mapping structure to determine the deviation interval to which the state deviation belongs, and the control amplitude constraint corresponding to the deviation interval is extracted from the global water supply scheduling strategy. Calculate the constraint strength value of the control amplitude constraint, and generate constraint subsets of different strength levels according to the magnitude of the constraint strength value; The edge policy space is generated by combining subsets of constraints of different strength levels in descending order of constraint strength value.

4. The method according to claim 1, characterized in that, The state deviation is predicted based on the rate of change of state to obtain the state deviation for the future time period. The predicted deviation is obtained by weighted summation of the state deviation and the state deviation for the future time period. The rate of change of state is constructed into a trend feature sequence according to the time sequence relationship, and the trend feature sequence is decomposed into periodic components and random components by multi-scale decomposition. The periodic component is analyzed to obtain the periodic variation law, and the random component is analyzed to obtain the fluctuation variation law. Based on the periodic variation law and the fluctuation variation law, the state deviation is predicted to obtain the state deviation in the future period. The state deviation is combined with the state deviation of the future time period to form a time series data sequence, and the fluctuation amplitude of the time series data sequence is calculated to obtain the time series fluctuation intensity value. The fluctuation weights of the state deviation and the future time period state deviation are calculated based on the temporal fluctuation intensity value. The state deviation and the future time period state deviation are then weighted and summed according to the fluctuation weights to obtain the predicted deviation.

5. The method according to claim 1, characterized in that, Within the control range constraints defined by the edge strategy space, the water supply adjustment amount calculated based on the predicted deviation includes: A deviation feature sequence is constructed based on the predicted deviation amount, and the numerical distribution of the deviation feature sequence is calculated to obtain the fluctuation characteristics. The changing trend of the predicted deviation amount is analyzed based on the fluctuation characteristics. The control direction is determined based on the changing trend, and the control amplitude constraint range corresponding to the control direction is obtained in the edge strategy space. The predicted deviation is then mapped to the control amplitude constraint range. The control amplitude constraint range is divided into multiple control sub-intervals, and the control gain value corresponding to each control sub-interval is calculated based on the fluctuation characteristics. The predicted deviation is allocated according to the control sub-interval, and multiplied with the corresponding control gain value to obtain the control calculation result; The water supply adjustment amount is obtained by normalizing the control calculation results within the control range constraint.

6. The method according to claim 1, characterized in that, The computational load of edge computing devices is detected. When the computational load exceeds a preset load threshold, the predicted deviation and the edge strategy space are sent to adjacent edge computing devices for water supply adjustment calculation, including: Obtain resource usage information of edge computing devices, construct a computing resource usage sequence based on the resource usage information, and calculate the computing load based on the computing resource usage sequence; The computing load is compared with a preset load threshold. When the computing load exceeds the preset load threshold, the computing resource status of adjacent edge computing devices is obtained. Based on the computing resource status, a load balancing index is constructed. Adjacent edge computing devices are sorted according to the load balancing index, and the device with the best load balancing index among the adjacent edge computing devices is selected as the target device. The predicted deviation and the edge strategy space are sent to the target device to calculate the water supply adjustment.

7. The method according to claim 1, characterized in that, Based on the water supply adjustment amount, a water supply control command is generated and sent to the execution equipment at the water supply network node to perform water supply regulation, including: Obtain the operating status information of the water supply network nodes, determine the execution constraints based on the operating status information, and map the water supply adjustment amount to the range of the execution constraints; A water supply control command is generated based on the mapped water supply adjustment amount, the communication status of the water supply network node is obtained, and a communication link is selected according to the communication status. The water supply control command is segmented and encoded to obtain a control data packet, and the transmission rate of the control data packet is adjusted based on the communication status. The control data packets are sent to the execution device via the communication link, and the execution status feedback information returned by the execution device is received. Verify the execution status of water supply control commands based on the execution status feedback information, and complete the water supply regulation of water supply network nodes.

8. A water supply strategy calculation system coupled with front-end sensing and edge computing, used to implement the method described in any one of claims 1-7, characterized in that, The system includes: The status acquisition module is used to collect water supply status data of water supply network nodes and transmit it to edge computing devices; The deviation calculation module is set in the edge computing device. It is used to compare the water supply status data with the preset water supply target value to calculate the status deviation. It performs time-series difference operation on the status deviation of multiple consecutive collection cycles in the water supply status data to obtain the status change rate. The strategy space generation module is used to receive the global water supply scheduling strategy from the central computer room, determine the deviation interval to which the state deviation belongs based on the state deviation, extract the control amplitude constraint corresponding to the deviation interval from the global water supply scheduling strategy, and generate the edge strategy space. The deviation prediction module is used to predict the state deviation based on the state change rate to obtain the state deviation in the future time period, and to obtain the predicted deviation by weighted summation of the state deviation and the state deviation in the future time period. The strategy calculation module is used to calculate the water supply adjustment amount based on the predicted deviation within the control range limited by the edge strategy space. The load monitoring module is used to simultaneously detect the computing load of edge computing devices. When the computing load exceeds the preset load threshold, the predicted deviation and the edge strategy space are sent to the adjacent edge computing devices to calculate the water supply adjustment amount. The control execution module is used to generate water supply control commands based on the water supply adjustment amount and send them to the execution equipment at the water supply network nodes for water supply regulation.

9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 7.

Images