Multi-meter-position parallel intelligent water meter verification scheduling method and system
By constructing feature vectors and clustering, dynamically mapping verification control parameters, analyzing error data in real time, and performing predictive blocking, the problems of low efficiency and high energy consumption caused by differences in sensor characteristics in parallel verification of multiple positions are solved, and a more efficient verification process is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING ZIFENG WATER EQUIPMENT CO LTD
- Filing Date
- 2025-12-30
- Publication Date
- 2026-04-21
AI Technical Summary
Existing multi-meter parallel verification methods suffer from rigid verification strategies, inefficiency due to the impact of the bottleneck effect, and high energy consumption because they ignore the differences in the micro-sensing characteristics of individual water meters.
By constructing feature vectors and clustering, the set of verification control parameters is dynamically mapped, error data is analyzed in real time, and predictive blocking operations are performed to optimize the verification process.
It improved verification efficiency, reduced energy consumption and time spent on invalid verifications, and increased the throughput and resource utilization rate of the verification production line per unit time.
Smart Images
Figure CN121903579A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of instrument calibration and scheduling, specifically to a method and system for parallel calibration and scheduling of smart water meters with multiple meter positions. Background Technology
[0002] In the production and quality control of smart water meters, meter calibration is a crucial process to ensure product compliance with standards and guarantee metering accuracy. With the widespread application of smart water meters and the continuous growth of market demand, multi-meter parallel calibration technology has emerged. This technology aims to improve calibration efficiency and reduce costs to meet the needs of large-scale production. The development of multi-meter parallel calibration technology allows multiple water meters to be calibrated simultaneously, significantly shortening the overall calibration time, improving production efficiency, and driving the rapid development of the smart water meter industry.
[0003] Existing multi-position smart water meter calibration technology mainly relies on simple physical grouping based on the water meters' physical diameter or accuracy class. Within the same batch of calibrations, a pre-set, fixed calibration procedure is employed. For example, a uniform flow stabilization waiting time is set, regardless of the actual condition of the water meters; a fixed sampling frequency is used to acquire data; and fixed PID control parameters are used for adjustment. This method is relatively simple to operate and easily enables large-scale batch calibration.
[0004] However, existing multi-meter parallel calibration methods have significant drawbacks. Due to the microscopic discreteness of sensor manufacturing processes, even water meters of the same specifications and models can exhibit significant differences in their internal sensor characteristics, such as zero-point drift rate, signal-noise floor, and transient response speed. In traditional serial parallel modes, to ensure the reliability of the entire batch of data, the system sets conservative calibration parameters based on the "worst performing individual." This results in a large number of high-performance water meters having to wait passively and meaninglessly, severely restricting the overall operating cycle of the parallel pipeline due to the "bottleneck effect." Furthermore, the lack of early intervention mechanisms means that unqualified water meters must still complete the entire calibration process, leading to inefficient use of pump energy and time resources. Summary of the Invention
[0005] To address the problems of rigid verification strategies, limited parallel operation efficiency due to the neglect of individual micro-sensing characteristics of water meters, and high energy consumption of ineffective verification in existing multi-meter parallel verification methods, this application provides a multi-meter parallel intelligent water meter verification scheduling method and system.
[0006] Firstly, this application provides a method for scheduling the verification of smart water meters in parallel across multiple meter positions, including: Receive the sensor status parameters of the smart water meter to be tested to construct a feature vector; Clustering is performed on the feature vectors to generate a homogeneous batch queue of test samples; The control parameter set is dynamically mapped based on the cluster center vector of the batch queue; The smart water meter is calibrated according to the mapped set of calibration control parameters. During the calibration process, error data streams are recorded and analyzed in real time, and predictive blocking operations are performed. The predictive blocking operation includes: performing sliding window sampling and least squares linear regression on the real-time error data to calculate the slope and intercept of the current trend line; based on the trend line, the predicted error value at the end of the calibration is estimated and compared with the pass standard threshold. When the predicted error value exceeds the pass standard threshold and the confidence coefficient is greater than the safety threshold, a stop command is sent to the corresponding smart water meter and its data acquisition channel is cut off.
[0007] By adopting the above scheme, feature vectors are constructed and clustered based on the sensing status parameters of smart water meters, realizing dynamic mapping of verification parameters. During the verification process, the verification of unqualified water meters is stopped in time through predictive blocking operations, reducing the energy consumption and time of invalid verification and improving verification efficiency.
[0008] Preferred options also include: A combination of verification conditions is constructed, taking into account the type of smart water meter to be verified, test conditions, and preset flow points. Under each combination of verification conditions, the sensing state parameters of the smart water meter to be verified are received to construct a feature matrix with multi-dimensional feature vectors. Each feature vector is assigned a matching feature weight according to each combination of verification conditions. The multi-dimensional feature vectors include: static noise variance, zero-point drift slope, instantaneous response delay, response time constant, and overshoot.
[0009] By adopting the above scheme, considering the type of smart water meter to be tested, test conditions, and preset flow points to construct a combination of testing conditions, a feature matrix can be constructed for the received sensor state parameters under different conditions. Combined with multi-dimensional feature vectors and matched feature weights, the adjustment of testing parameters can be made to better fit the actual characteristics of the water meter, further improving the precision of testing parameter allocation and the accuracy of testing results.
[0010] Preferred options also include: Density clustering is performed based on the constructed feature matrix to generate a homogeneous batch queue for testing. The density clustering includes calculating the feature space distance between smart water meters to be tested based on Euclidean distance or cosine similarity, and assigning feature vectors smaller than a preset similarity threshold to the same cluster. When the number of smart water meters in a cluster exceeds the maximum number of parallel meters, the batch queue is divided into multiple sub-batches according to a preset upper limit for meter positions. When the number is insufficient, the current batch is marked as not fully loaded. For each smart water meter, when the combination of verification conditions changes, the spatial distance between the newly constructed feature matrix and the cluster center vector of density clustering corresponding to the original feature matrix of the smart water meter is calculated. If the spatial distance is not greater than a preset spatial threshold, the current batch queue of the smart water meter is maintained. If the spatial distance is greater than the preset spatial threshold, the smart water meter is marked as waiting to be reassigned, and the smart water meter to be reassigned is assigned to the batch queue corresponding to the cluster center with the closest spatial distance.
[0011] By adopting the above scheme, density clustering is performed using the constructed feature matrix to generate a homogeneous verification batch queue, which can more accurately classify smart water meters and avoid the limitations of simple clustering. When the combination of verification conditions changes, the clustering results and category assignments are adjusted incrementally and progressively based on the new characteristics of subsequent flow points, based on the initial clustering, thereby improving verification efficiency and resource utilization.
[0012] Preferred options also include: The set of calibration control parameters includes: PID control valve parameters, flow stabilization waiting time, and sampling frequency; and real-time acquisition of calibration environment data. A machine learning-based mapping strategy prediction model is established. The input to this model is a feature matrix and calibration environment data. The output is the optimal mapping calibration control parameter set and expected performance indicators. Reinforcement learning is used for model training. The set states include: current feature matrix, current calibration environment data, and current flow point. Actions include adjusting PID control valve parameters, changing the steady-state waiting time, and sampling frequency. The reward function is a comprehensive score function weighted by calibration time, overshoot, and calibration accuracy. The performance indicators include actual steady-state time, overshoot, and steady-state fluctuation variance. Input the feature matrix corresponding to the cluster center of the current batch queue into the mapping strategy prediction model to obtain the optimal mapping strategy, and complete the dynamic mapping verification control parameter set according to the optimal mapping strategy, instead of the dynamic mapping verification control parameter set based on the cluster center vector of the batch queue.
[0013] By adopting the above scheme, a mapping strategy prediction model is established using machine learning and reinforcement learning. The optimal set of verification control parameters is dynamically mapped by combining the feature matrix and verification environment data. This achieves a shift from static mapping to dynamic mapping, more accurately adapting to different verification situations, improving the mapping accuracy of the set of verification control parameters, and further enhancing verification efficiency and quality.
[0014] Preferred options also include: After each verification, the actual performance index of each smart water meter is recorded; the average performance index of all smart water meters in each cluster is calculated; if the average performance index is lower than the expected performance index, the optimal mapping verification control parameter set of the corresponding batch queue of the cluster is regenerated, and the verification control parameter set of the corresponding batch queue of the current cluster is remapped by optimizing the mapping strategy and predicting the model parameters.
[0015] By adopting the above scheme, the verification control parameter set is dynamically adjusted and remapped based on the comparison between the average actual performance index and the expected performance index of smart water meters in each cluster after each verification. This allows the verification control parameter set to better adapt to the actual situation of different clusters, further improving the accuracy and efficiency of verification.
[0016] Preferred options also include: For each smart water meter whose prediction error value does not exceed the qualified standard threshold, the error level corresponding to the difference range is determined based on the difference between the prediction error value and the qualified standard threshold. During subsequent verification, for each smart water meter whose predicted error value does not exceed the qualified standard threshold, a dynamic adjustment strategy for the verification control parameter set is matched based on whether the determined error level is greater than the preset error level; if the error level is less than the preset error level, the first dynamic adjustment strategy is matched; the first dynamic adjustment strategy includes reducing the number of verification flow points, shortening the stabilization waiting time for each flow point, reducing the collection frequency, and restoring the original verification control parameter set when the actual error level is verified to be greater than the preset error level after the verification of each flow point is completed.
[0017] By adopting the above scheme, the error level is determined based on the difference between the predicted error value and the qualified standard threshold, and the set of verification control parameters for intra-cluster mapping is dynamically adjusted according to the error level. This avoids over-verification of smart water meters on the edge of qualification, improves verification efficiency, and reduces the waste of energy and time.
[0018] Preferred options also include: After each batch of verification is completed, the final verification error value of all smart water meters in that batch is obtained. Calculate the Pearson correlation coefficient between the eigenvectors of each dimension in the feature matrix constructed by the smart water meter and the final verification error value; Based on the absolute value of the calculated Pearson correlation coefficient, adjust the weighting coefficients of the feature vectors of each dimension in the subsequent clustering distance algorithm; including: for feature dimensions whose calculated Pearson correlation coefficient is higher than the preset significance threshold, increase the feature weight based on the preset feature weight; for feature dimensions whose calculated Pearson correlation coefficient is lower than the preset irrelevance threshold, decrease the feature weight based on the preset feature weight.
[0019] By adopting the above scheme, based on the final verification error value of each batch of smart water meters after verification, the Pearson correlation coefficient between the feature vectors of each dimension in the feature matrix and the error value is calculated. Based on this, the weighting coefficients of the feature vectors of each dimension in the subsequent clustering distance algorithm are adjusted, which can make the clustering results more accurately reflect the characteristics of smart water meters, further optimize the allocation of verification parameters, and improve verification efficiency and accuracy.
[0020] Secondly, this application provides a multi-position parallel intelligent water meter calibration and scheduling system, comprising: The data acquisition module is used to receive the sensing status parameters of the smart water meter to be tested in order to construct a feature vector; The classification processing module is used to perform clustering on feature vectors to generate a homogeneous batch queue of tests; The parameter mapping module is used to dynamically map the set of verification control parameters based on the cluster center vector of the batch queue. The blocking execution module is used to perform smart water meter verification according to the mapped verification control parameter set. During the verification process, it records and analyzes the error data stream in real time and performs predictive blocking operations. The execution of predictive blocking operations includes: performing sliding window sampling and least squares linear regression on the real-time error data to calculate the slope and intercept of the current trend line; based on the trend line, predicting and calculating the predicted error value at the end of the verification and comparing it with the pass standard threshold; when the predicted error value exceeds the pass standard threshold and the confidence coefficient is greater than the safety threshold, sending a stop command to the corresponding smart water meter and cutting off its data acquisition channel.
[0021] By adopting the above scheme, based on the homogeneous clustering of multidimensional sensing feature vectors and the dynamic adaptive mapping of the inspection strategy, the inspection parameters can be finely allocated on demand for batches with different characteristics, reducing unnecessary waiting redundancy in parallel operations; and the predictive blocking mechanism based on trend regression can be used to avoid ineffective input of defective products.
[0022] Thirdly, this application provides a computer-readable storage medium including a stored computer program, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to perform the method described above.
[0023] Fourthly, this application provides a computer device, the computer device including a memory, a processor and a program stored in the memory and executable thereon, the program being executed by the processor to implement the steps of the method described above.
[0024] In summary, this application has the following beneficial effects: 1. Receive the sensor status parameters of the smart water meters to be tested and construct feature vectors or feature matrices. Perform clustering on these parameters to generate a homogeneous testing batch queue. Dynamically map the testing control parameter set based on the clustering results to achieve fine-grained on-demand allocation of testing parameters for batches with different characteristics, reducing unnecessary waiting redundancy in parallel operations. Analyze the error data stream in real time during testing and perform predictive blocking operations to avoid ineffective input of defective products and improve the throughput per unit time of the testing pipeline. 2. Construct a feature matrix of multi-dimensional feature vectors and set matching feature weights for more accurate clustering and verification; perform density clustering based on the feature matrix and process batch queues to improve the rationality of verification resource allocation; use machine learning mapping strategies to predict the dynamic mapping of the verification control parameter set and optimize verification by combining verification environment data; regenerate and map the verification control parameter set based on actual performance indicators to ensure verification effectiveness. 3. Dynamically adjust the set of verification control parameters according to the error level to improve verification efficiency; adjust the weighting coefficients of feature vectors of each dimension in the subsequent clustering distance algorithm to improve clustering accuracy. Attached Figure Description
[0025] Figure 1 This is a flowchart of the multi-position parallel smart water meter calibration and scheduling method described in a specific embodiment; Figure 2 This is a schematic diagram of the structure of the multi-position parallel smart water meter calibration and scheduling system described in a specific embodiment. Detailed Implementation
[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0027] like Figure 1 As shown in the figure, this application discloses a multi-position parallel smart water meter verification and scheduling method, including: receiving sensor status parameters to construct feature vectors, clustering feature vectors to generate batch queues, mapping verification control parameter sets according to cluster center vectors, verifying according to the verification parameter sets and performing predictive blocking operations, etc. The content of each step will be described in detail below.
[0028] S1. Receive the sensing status parameters of the smart water meter to be tested to construct a feature vector.
[0029] Specifically, the process of constructing a feature vector from the sensor status parameters of the smart water meter to be tested includes: receiving sensor status data packets reported by the smart water meter, parsing the data packets to extract key parameters, and normalizing the data to construct a feature vector. Specifically, a data connection is established beforehand with the smart water meter to be tested, and the sensor status data packets calculated and uploaded by the smart water meter's internal processor are received. The sensor status data packets are parsed to extract key parameters as feature dimensions, such as dynamic response parameters like response time constant, overshoot, and settling time. The extracted parameters are then normalized to construct a dimensional metering feature vector corresponding to each smart water meter.
[0030] In the process of parsing the sensor status data packet and extracting key parameters as feature dimensions, to assist in building a more accurate clustering model, the extraction process progresses from extracting a single dynamic response curve to extracting multi-dimensional feature parameters such as static stability, zero-point characteristics, and dynamic response. This results in the following multi-dimensional feature vectors: static noise variance (extracting the signal amplitude of a preset number of sampling points continuously collected by the sensor in a no-flow state and calculating its statistical variance), zero-point drift slope (obtaining the rate of change of the zero-point reference value of the smart water meter within a preset time window, i.e., establishing the slope coefficient of a linear fitting function for the zero-point value changing with time), transient response delay (extracting the difference in the rising edge response time of the sensor to a preset excitation signal), and response time constant, overshoot, etc. These parameters are then normalized to construct a multi-dimensional metering feature vector corresponding to each smart water meter, which is then aggregated to form a feature matrix to be processed.
[0031] S2. Perform clustering on the feature vectors to generate a homogeneous batch queue of tests.
[0032] Specifically, based on the Euclidean distance calculation of feature vectors, density clustering is performed and a homogeneous batch queue is generated; this includes distance calculation, density clustering processing, and batch mapping.
[0033] In this embodiment, the feature vector (matrix) to be processed generated from the smart water meter in the above steps is taken as an example. During distance calculation, in the multi-dimensional feature space, a preset distance algorithm, such as Euclidean distance or cosine similarity algorithm, is used to calculate the spatial distance value between any two feature vectors (matrices) in the matrix to be processed, generating a distance vector (matrix). During density clustering, the distance (matrix) is traversed, and feature vectors (matrices) with spatial distance values less than a preset similarity threshold are grouped into the same cluster. The preset similarity threshold can be adjusted according to the actual situation to determine a suitable clustering effect. During batch mapping, smart water meter IDs belonging to the same cluster are mapped to a single batch of verification objects. If the number of water meters in the cluster exceeds the preset maximum number of meter mounting positions per batch, it is divided into multiple sub-batches according to the maximum number; if the number is insufficient, the batch is marked as not fully loaded; finally, a queue of objects to be executed containing several verification batches is output.
[0034] S3. Dynamically map the control parameter set according to the cluster center vector of the batch queue.
[0035] Specifically, the dynamic mapping verification control parameter set based on the cluster center vector of the batch queue includes calculating the cluster center vector, inputting the strategy mapping model output parameter set, and issuing instructions.
[0036] In this embodiment, the output of the above steps includes a queue of objects to be executed containing several batches of verification objects (e.g., batch object numbers are divided into batch 001, batch 002, etc.). When calculating the cluster center vector, for each batch of verification objects (e.g., batch 001), the geometric center vector of the cluster is calculated; wherein, the geometric center vector can be obtained by averaging each dimension of all feature vectors (matrices) within the cluster.
[0037] When inputting the strategy mapping model to output the parameter set, the center vector of the cluster is input into the preset strategy mapping model, which outputs the corresponding verification control parameter set to assist in the subsequent verification control of smart water meters in the same batch. The preset strategy mapping model can be based on neural network training, outputting a verification control parameter set adapted to the input center vector, including: PID control valve parameters, steady-flow waiting time, and sampling frequency; or the preset strategy mapping model can directly perform logical transformation operations on the input center vector according to a preset logical transformation to map and output a verification control parameter set adapted to it. For example, the preset logical transformation process includes PID control valve parameter calculation, steady-flow waiting time calculation, and filter order setting; PID control valve parameter calculation involves performing sliding window statistics on the center vector, calculating the rate of change of each parameter, extracting time-series features and statistical features to generate fused features, establishing a mapping function between kp, ki, and kd and the fused features, and calculating according to the constructed mapping function; steady-flow waiting time calculation involves extracting the mean of the static noise variance in the center vector and calculating the steady-flow waiting time parameter during the verification process through an inverse proportional mapping function. When the mean variance is lower than the preset baseline, the flow stabilization parameter for the first preset duration is output; otherwise, the second preset flow stabilization parameter for the first preset duration is output. The filter order setting is achieved by extracting the mean zero-point drift slope in the center vector and calculating the order parameter of the internal filtering algorithm of the smart water meter through a linear relationship mapping, thereby determining the sampling frequency.
[0038] When the command is issued, a configuration command package containing PID control valve parameters, flow stabilization waiting time parameters, and filter order parameters will be sent in parallel to all smart water meters and calibration device control units corresponding to that batch, completing parameter synchronization before calibration.
[0039] S4. Perform smart water meter verification according to the mapped set of verification control parameters. During the verification process, record and analyze error data streams in real time and perform predictive blocking operations.
[0040] Specifically, smart water meters are calibrated according to the mapped set of calibration control parameters. During the calibration process, the error between the smart water meter reading and the actual flow rate is recorded in real time. These errors are analyzed to predict the final calibration error and prevent the calibration operation of smart water meters with large errors in advance, thus avoiding ineffective investment in defective products. The specific steps include: sliding window sampling, trend fitting, final value prediction and comparison, and blocking trigger.
[0041] Specifically, for each smart water meter, during sliding window sampling, a first-in-first-out (FIFO) data buffer queue of preset length is established. The preset length can be set according to actual conditions, such as a buffer queue of 10 data points. During trend fitting, least squares linear regression is performed on the error data points in the buffer queue to calculate the slope and intercept of the error change trend line at the current moment. During final value prediction and comparison, based on the trend line, the predicted error value at the end of the verification (e.g., the end time of the current Q1 flow point detection) is estimated and calculated. The absolute value of the difference between this predicted error value and the preset pass standard threshold is calculated. When the blocking is triggered, if the predicted error value of a smart water meter exceeds the preset pass standard threshold and the confidence coefficient is greater than the preset safety value, a bypass command for that specific water meter ID is immediately generated. This command drives the verification device to cut off the data acquisition channel of that meter position and sends a stop sampling command to the smart water meter.
[0042] S5. After the verification is completed, calculate the correlation of feature weights and iteratively update the weighting coefficients of the clustering distance algorithm.
[0043] Specifically, after each batch of verification is completed, the final verification error value of all smart water meters in that batch is obtained, which is the error value between the reading of each smart water meter recorded at the end and the actual flow rate.
[0044] Conduct correlation analysis: Calculate the Pearson correlation coefficient between the feature vectors of each dimension in the feature vector (matrix) constructed for each smart water meter and the final verification error value.
[0045] Weight Update: Based on the absolute value of the calculated Pearson correlation coefficient, adjust the weighting coefficients of the feature vectors for each dimension in the subsequent clustering distance algorithm. This includes increasing the feature weights of feature dimensions with calculated Pearson correlation coefficients higher than a preset significance threshold, and decreasing the feature weights of feature dimensions with calculated Pearson correlation coefficients lower than a preset irrelevance threshold. The preset weights can be manually set based on factors such as the type of smart water meter to be tested, test conditions, and specific testing flow points.
[0046] Model saving: Save the updated weighting coefficients for use when clustering the feature vectors (matrices) of the next batch of water meters to be inspected that belong to the same cluster, and add the adjusted weights.
[0047] In a specific embodiment, further considering the combination of verification conditions, by setting matching feature weights, the characteristics of smart water meters under different conditions can be more accurately reflected, thereby providing a more accurate data foundation for subsequent clustering and parameter mapping, further optimizing the verification scheduling strategy, and improving the accuracy and efficiency of verification. The method also includes: constructing a combination of verification conditions considering the type of smart water meter to be verified, test conditions, and preset flow points, and setting differentiated weights according to the water meter type (e.g., mechanical, ultrasonic, electromagnetic), test conditions (e.g., ambient temperature: low / normal / high temperature), and flow points (zero point, Qmin, Q3, Qmax). Since the verification targets of the same water meter are different at different flow points, the weights of its feature importance should be dynamically adjusted accordingly to reflect the degree of influence of the features on the verification efficiency and accuracy of the current flow point.
[0048] Under each combination of verification conditions, the sensor status parameters of the smart water meter to be verified are received to construct a feature matrix with multi-dimensional feature vectors. Each feature vector is assigned a matching feature weight according to each combination of verification conditions, and a dynamic configuration table of smart water meter verification feature weights is generated.
[0049] Specifically, some contents of the summary table of dynamic configuration of water meter smart verification feature weights are shown in Table 1 below: Table 1
[0050] Regarding the settings in Table 1 above, without considering the influence of temperature, the weighting settings need to reflect the main objectives of the current verification: accuracy is prioritized at low flow rates, and efficiency is prioritized at high flow rates. For example, at Qmin, the weight of noise variance should be high because the signal is weak at low flow rates; at Qmax, the weight of dynamic response time should be increased to achieve rapid stability. For mechanical water meters, the focus is on start-up-related characteristics at low flow rates, while dynamic stability is more important at high flow rates. For ultrasonic and electromagnetic water meters, noise and response are more critical at different flow rates. Furthermore, considering the influence of temperature, temperature changes can lead to changes in the material properties of the water meter and the viscosity of the fluid, thus affecting the characteristics.
[0051] First, for mechanical water meters, at low temperatures, the weight of instantaneous response delay should be appropriately increased for all flow points. The weight of static noise variance at zero and Qmin should be slightly increased, and the weight of time constant at Q3 and Qmax can be appropriately increased. Conversely, at high temperatures (e.g., 35℃), the weight of delay may decrease for mechanical water meters. High temperatures may cause material expansion and gap changes, potentially increasing the static noise variance. The weight of static noise variance at zero and Qmin needs to be increased, while the weight of overshoot at Qmax is slightly reduced due to potential changes in damping at high temperatures. Second, for ultrasonic water meters, the mechanism of temperature's influence is different, mainly manifested in changes in sound velocity and electronic component performance. Regarding static noise variance: temperature changes may affect signal quality, potentially increasing noise at both low and high temperatures, thus increasing its weight. Zero-point drift slope: the zero point of ultrasonic water meters is significantly affected by temperature; therefore, the weight of zero-point drift slope should be significantly increased when temperature changes. Finally, the effects of temperature on electromagnetic water meters mainly lie in the fluid conductivity and coil resistance. Regarding the zero-point drift slope: temperature changes cause zero-point drift, so the weighting needs to be increased. As for the static noise variance: temperature changes may cause uneven conductivity, increasing measurement noise, so the weighting also needs to be increased. Electromagnetic water meters have a fast response, and temperature has a relatively small impact on their dynamic characteristics. However, high temperatures may cause the coil to heat up, affecting stability. Therefore, the weighting of the time constant and overshoot needs to be increased under high temperature and high flow rates.
[0052] In summary, before verification, the water meter type is obtained, the ambient temperature is measured in real time, the flow point for current verification is determined, the feature dimension values of the water meter under the current conditions are calculated to form the original feature vector, and the corresponding matching weight of the main meter is dynamically configured based on the intelligent verification feature weight of the water meter to complete the weighted processing.
[0053] In one specific embodiment, dynamically adjusting the batch queue of water meters according to changes in verification conditions can better adapt to different verification situations, further optimize the verification process, improve the efficiency of parallel operations, and reduce unnecessary waiting and resource waste caused by changes in verification conditions; the method also includes: Specifically, choosing to perform density clustering based on the constructed feature matrix to generate a homogeneous verification batch queue, and optimizing the clustering of feature vectors to generate a homogeneous verification batch queue, can achieve more accurate generation of homogeneous verification batch queues. However, during a multi-flow point verification process, the combination of verification conditions may change. If clustering is re-performed every time the combination of verification conditions changes, the computational load is large, especially when there are many water meters and high feature dimensions. The grouping may change frequently, leading to instability in the control strategy.
[0054] Therefore, based on the initial clustering, subsequent flow points can be compared with the distance between the new feature matrix and the existing cluster centers to determine whether the water meter should be regrouped. For each smart water meter, when the combination of verification conditions changes, such as when entering the next preset flow point (e.g., Q3) for verification, the spatial distance between the newly constructed feature matrix and the cluster center vector of the original feature matrix corresponding to the smart water meter for density clustering is calculated. If the spatial distance is not greater than the preset spatial threshold, the current batch queue of the smart water meter is maintained. If the spatial distance is greater than the preset spatial threshold, the smart water meter is marked as to be reassigned. The smart water meter to be reassigned is assigned to the batch queue corresponding to the cluster center with the closest spatial distance. For the reassigned water meter, its verification control parameters (PID, steady flow waiting time, etc.) are also switched to the verification control parameter set of the batch queue corresponding to the newly assigned cluster center.
[0055] At the same time, the cluster centers can be updated after each change in the combination of verification conditions to adapt to the slow changes in water meter characteristics.
[0056] In one specific embodiment, a machine learning mapping strategy prediction model is introduced, taking into account the verification environment data. By training the model through reinforcement learning, the optimal set of verification control parameters is dynamically mapped more intelligently for different batches of water meters, better adapting to different environmental conditions and water meter characteristics, further improving the accuracy and efficiency of verification, and reducing energy consumption. The method also includes: real-time acquisition of verification environment data, such as temperature data.
[0057] A machine learning-based mapping strategy prediction model is established. The inputs to this model are a feature matrix and calibration environment data, while the outputs are the optimal mapping calibration control parameter set and expected performance indicators. Initially, a preset mapping strategy is used. If different input feature matrices and calibration environment data are set based on expert experience, the optimal mapping calibration control parameter set in the corresponding mapping strategy is adopted. Information for each calibration is recorded: input feature matrix, mapping strategy used, calibration temperature, and performance indicators. These performance indicators include actual steady-state current consumption time, overshoot, and steady-state fluctuation variance.
[0058] Specifically, the mapping strategy prediction model is trained using reinforcement learning. The settings include: current feature matrix, current calibration environment data, and current flow point. Actions include adjusting PID control valve parameters, changing the flow stabilization waiting time, and sampling frequency. The reward function is a weighted score function calculated based on calibration time (a negative reward; the longer the calibration time, the higher the penalty coefficient), overshoot (a negative reward; the higher the overshoot value, the higher the penalty coefficient), and calibration accuracy (a positive reward; the higher the calibration accuracy, the higher the penalty coefficient). The feature matrix corresponding to the cluster center of the current batch queue is input into the mapping strategy prediction model to obtain the optimal mapping strategy. This completes the dynamic mapping of the calibration control parameter set according to the optimal mapping strategy, replacing the dynamic mapping of the calibration control parameter set based on the cluster center vector of the batch queue.
[0059] Furthermore, during the verification process of dynamically mapping the verification control parameter set according to the optimal mapping strategy, the performance indicators of the mapping strategy itself are recorded, thereby constructing a feedback loop. The effectiveness of the mapped verification control parameter set is verified through the performance indicators. Specifically, after each verification (preset flow point), the actual performance indicators of each smart water meter are recorded; the average performance indicator of all smart water meters within each cluster is calculated. If the average performance indicator is lower than the expected performance indicator, the optimal mapping verification control parameter set for the corresponding batch queue of that cluster is regenerated. By optimizing the parameters of the mapping strategy prediction model, the verification control parameter set for the corresponding batch queue of the current cluster is remapped, completing the dynamic mapping of the verification control parameter set according to the remapped optimal mapping strategy, and thus re-verifying.
[0060] In one specific embodiment, besides determining whether feedback adjustment is needed based on the performance metrics of the mapping strategy itself, feedback adjustment can also be performed by obtaining the error between the recorded traffic and the actual traffic through the actual prediction difference; the method further includes: For each smart water meter whose prediction error value does not exceed the pass standard threshold, it indicates that the smart water meter prediction is qualified. For smart water meters with a high probability of passing the test, we further distinguish between water meters that are qualified in prediction and have small errors and water meters that are qualified in prediction and have moderate errors. Based on the difference between the prediction error value and the pass standard threshold, we determine the error level corresponding to the difference range (e.g., small error and moderate error).
[0061] During subsequent verification (flow point change verification), for each smart water meter whose predicted error value does not exceed the qualified standard threshold, a dynamic adjustment strategy for the verification control parameter set is applied based on whether the determined error level is greater than the preset error level. If the error level is less than the preset error level, the first dynamic adjustment strategy is applied. An error level less than the preset error level indicates a small error, potentially allowing for simpler verification, i.e., the first dynamic adjustment strategy is adopted: reducing the number of verification flow points, shortening the stabilization waiting time for each flow point, and reducing the sampling frequency, etc. To ensure accuracy, after each flow point verification is completed, if it is verified that the actual error level is greater than the preset error level, indicating a large error in the simple measurement method, the original verification control parameter set is restored. This approach improves verification efficiency while meeting verification requirements, and restores the original parameter set to ensure verification accuracy when the actual error exceeds the preset error level.
[0062] like Figure 2 As shown in the figure, this embodiment discloses a multi-position parallel intelligent water meter calibration and scheduling system, specifically including: The data acquisition module 100 is used to receive the sensing status parameters of the smart water meter to be tested in order to construct a feature vector; The classification processing module 200 is used to perform clustering on the feature vectors to generate a homogeneous test batch queue; The parameter mapping module 300 is used to dynamically map the set of verification control parameters according to the cluster center vector of the batch queue. The blocking execution module 400 is used to perform smart water meter verification according to the mapped verification control parameter set. During the verification process, it records and analyzes the error data stream in real time and performs predictive blocking operations. The execution of predictive blocking operations includes: performing sliding window sampling and least squares linear regression on the real-time error data to calculate the slope and intercept of the current trend line; based on the trend line, predicting and calculating the predicted error value at the end of the verification and comparing it with the pass standard threshold; when the predicted error value exceeds the pass standard threshold and the confidence coefficient is greater than the safety threshold, sending a stop command to the corresponding smart water meter and cutting off its data acquisition channel.
[0063] In one specific embodiment, the system further includes: The weight update module 500 is used to obtain the final verification error value of all smart water meters in each batch after the verification is completed; calculate the Pearson correlation coefficient between the feature vector of each dimension in the feature matrix constructed by the smart water meter and the final verification error value; and adjust the weighting coefficient of each dimension feature vector in the subsequent clustering distance algorithm according to the absolute value of the calculated Pearson correlation coefficient. This includes: increasing the feature weight of feature dimensions with a calculated Pearson correlation coefficient higher than a preset significance threshold, and decreasing the feature weight of feature dimensions with a calculated Pearson correlation coefficient lower than a preset irrelevance threshold.
[0064] In one specific embodiment, the system further includes: The data acquisition module 100 is also used to construct a combination of verification conditions that considers the type of smart water meter to be verified, test conditions, and preset flow points. Under each combination of verification conditions, it receives the sensing state parameters of the smart water meter to be verified to construct a feature matrix with multi-dimensional feature vectors. Each feature vector is assigned a matching feature weight according to each combination of verification conditions. The multi-dimensional feature vectors include: static noise variance, zero-point drift slope, instantaneous response delay, response time constant, and overshoot.
[0065] The classification processing module 200 is further configured to select and perform density clustering based on the constructed feature matrix to generate a homogeneous verification batch queue; for each smart water meter, when the combination of verification conditions changes, the spatial distance between the newly constructed feature matrix and the cluster center vector of the original feature matrix corresponding to the smart water meter is calculated; if the spatial distance is not greater than a preset spatial threshold, the current batch queue of the smart water meter is maintained; if the spatial distance is greater than the preset spatial threshold, the smart water meter is marked as to be reassigned, and the smart water meter to be reassigned is assigned to the batch queue corresponding to the cluster center with the closest spatial distance.
[0066] In one specific embodiment, the system further includes: The parameter mapping module 300 is also used to set the verification control parameter set, including: PID control valve parameters, steady flow waiting time, and sampling frequency; acquire verification environment data in real time; establish a machine learning-based mapping strategy prediction model, the input of which is a feature matrix and verification environment data, and the output is the optimal mapping verification control parameter set and expected performance indicators. Reinforcement learning is used for model training. The settings include: current feature matrix, current verification environment data, and current flow point; actions include adjusting PID control valve parameters, changing the steady flow waiting time, and sampling frequency; the reward function is set as a comprehensive score function weighted by verification time, overshoot, and verification accuracy; the performance indicators include actual steady flow time, overshoot, and steady-state fluctuation variance; input the feature matrix corresponding to the cluster center of the current batch queue into the mapping strategy prediction model to obtain the optimal mapping strategy, and complete the dynamic mapping of the verification control parameter set according to the optimal mapping strategy, replacing the dynamic mapping of the verification control parameter set based on the cluster center vector of the batch queue.
[0067] The parameter mapping module 300 is also used to record the actual performance index of each smart water meter after each verification; calculate the average performance index of all smart water meters in each cluster; if the average performance index is lower than the expected performance index, trigger the regeneration of the optimal mapping verification control parameter set of the corresponding batch queue of the cluster; and remap the verification control parameter set of the corresponding batch queue of the current cluster by optimizing the mapping strategy to predict the model parameters.
[0068] In one specific embodiment, the system further includes: The parameter mapping module 300 is further configured to, for each smart water meter whose predicted error value does not exceed the qualified standard threshold, determine the error level corresponding to the difference range between the predicted error value and the qualified standard threshold; in the subsequent verification process, for each smart water meter whose predicted error value does not exceed the qualified standard threshold, match the dynamic adjustment strategy of the verification control parameter set according to whether the determined error level is greater than the preset error level; if the error level is less than the preset error level, match the first dynamic adjustment strategy; the first dynamic adjustment strategy includes reducing the number of verification flow points, shortening the stabilization waiting time of each flow point, reducing the acquisition frequency, and restoring the original verification control parameter set after verifying that the actual error level is greater than the preset error level after the verification of each flow point is completed.
[0069] This application also discloses a computer-readable storage medium.
[0070] Specifically, the computer-readable storage medium stores a computer program that can be loaded by a processor and executed, such as the above-described multi-position parallel smart water meter verification and scheduling method. The computer-readable storage medium includes, for example, various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0071] This application also discloses a computer device.
[0072] Specifically, the computer device includes a memory and a processor. The memory stores a computer program that can be loaded by the processor and executed to perform the aforementioned multi-position parallel smart water meter verification and scheduling method.
[0073] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.
Claims
1. A method for parallel calibration and scheduling of smart water meters with multiple positions, characterized in that, include: Receive the sensor status parameters of the smart water meter to be tested to construct a feature vector; Clustering is performed on the feature vectors to generate a homogeneous batch queue of test samples; The control parameter set is dynamically mapped based on the cluster center vector of the batch queue; The smart water meter is calibrated according to the mapped set of calibration control parameters. During the calibration process, error data streams are recorded and analyzed in real time, and predictive blocking operations are performed. The execution of the predictive blocking operation includes: performing sliding window sampling and least squares linear regression on the real-time error data to calculate the slope and intercept of the current trend line; Based on the trend line, the predicted error value at the end of the verification is estimated and compared with the pass standard threshold. When the predicted error value exceeds the pass standard threshold and the confidence coefficient is greater than the safety threshold, a stop command is sent to the corresponding smart water meter and its data acquisition channel is cut off.
2. The multi-position parallel smart water meter calibration and scheduling method according to claim 1, characterized in that, Also includes: A combination of verification conditions is constructed, taking into account the type of smart water meter to be verified, test conditions, and preset flow points. Under each combination of verification conditions, the sensing state parameters of the smart water meter to be verified are received to construct a feature matrix with multi-dimensional feature vectors. Each feature vector is assigned a matching feature weight according to each combination of verification conditions. The multi-dimensional feature vectors include: static noise variance, zero-point drift slope, instantaneous response delay, response time constant, and overshoot.
3. The method for parallel verification and scheduling of smart water meters with multiple meter positions according to claim 2, characterized in that, Also includes: Density clustering is performed based on the constructed feature matrix to generate a homogeneous batch queue for testing. The density clustering includes calculating the feature space distance between smart water meters to be tested based on Euclidean distance or cosine similarity, and assigning feature vectors smaller than a preset similarity threshold to the same cluster. When the number of smart water meters in a cluster exceeds the maximum number of parallel meters, the batch queue is divided into multiple sub-batches according to a preset upper limit for meter positions. When the number is insufficient, the current batch is marked as not fully loaded. For each smart water meter, when the combination of verification conditions changes, the spatial distance between the newly constructed feature matrix and the cluster center vector of density clustering corresponding to the original feature matrix of the smart water meter is calculated. If the spatial distance is not greater than a preset spatial threshold, the current batch queue of the smart water meter is maintained. If the spatial distance is greater than the preset spatial threshold, the smart water meter is marked as waiting to be reassigned, and the smart water meter to be reassigned is assigned to the batch queue corresponding to the cluster center with the closest spatial distance.
4. The multi-position parallel smart water meter calibration and scheduling method according to claim 2, characterized in that, Also includes: The set of calibration control parameters includes: PID control valve parameters, flow stabilization waiting time, and sampling frequency; and real-time acquisition of calibration environment data. A machine learning-based mapping strategy prediction model is established. The input to this model is a feature matrix and calibration environment data. The output is the optimal mapping calibration control parameter set and expected performance indicators. Reinforcement learning is used for model training. The set states include: current feature matrix, current calibration environment data, and current flow point. Actions include adjusting PID control valve parameters, changing the steady-state waiting time, and sampling frequency. The reward function is a comprehensive score function weighted by calibration time, overshoot, and calibration accuracy. The performance indicators include actual steady-state time, overshoot, and steady-state fluctuation variance. The feature matrix corresponding to the cluster center of the current batch queue is input into the mapping strategy prediction model to obtain the optimal mapping strategy. The dynamic mapping verification control parameter set is then completed according to the optimal mapping strategy, replacing the dynamic mapping verification control parameter set based on the cluster center vector of the batch queue.
5. The multi-position parallel intelligent water meter calibration and scheduling method according to claim 4, characterized in that, Also includes: After each calibration, record the actual performance indicators of each smart water meter; Calculate the average performance index of all smart water meters in each cluster. If the average performance index is lower than the expected performance index, trigger the regeneration of the optimal mapping verification control parameter set for the corresponding batch queue of that cluster. By optimizing the mapping strategy and predicting the model parameters, the verification control parameter set for the corresponding batch queue of the current cluster is remapped.
6. The method for parallel verification and scheduling of smart water meters with multiple meter positions according to claim 1, characterized in that, Also includes: For each smart water meter whose prediction error value does not exceed the qualified standard threshold, the error level corresponding to the difference range is determined based on the difference between the prediction error value and the qualified standard threshold. In the subsequent verification process, for each smart water meter whose predicted error value does not exceed the qualified standard threshold, a dynamic adjustment strategy for matching the verification control parameter set is adopted based on whether the determined error level is greater than the preset error level. If the error level is less than the preset error level, the first dynamic adjustment strategy is matched; the first dynamic adjustment strategy includes reducing the number of calibration flow points, shortening the stabilization waiting time for each flow point, reducing the acquisition frequency, and restoring the original calibration control parameter set when the actual error level is verified to be greater than the preset error level after the calibration of each flow point is completed.
7. The method for parallel verification and scheduling of smart water meters with multiple meter positions according to claim 2, characterized in that, Also includes: After each batch of verification is completed, the final verification error value of all smart water meters in that batch is obtained. Calculate the Pearson correlation coefficient between the eigenvectors of each dimension in the feature matrix constructed by the smart water meter and the final verification error value; Based on the absolute value of the calculated Pearson correlation coefficient, adjust the weighting coefficients of the feature vectors of each dimension in the subsequent clustering distance algorithm; including: for feature dimensions whose calculated Pearson correlation coefficient is higher than the preset significance threshold, increase the feature weight based on the preset feature weight; for feature dimensions whose calculated Pearson correlation coefficient is lower than the preset irrelevance threshold, decrease the feature weight based on the preset feature weight.
8. A multi-position parallel intelligent water meter calibration and scheduling system, characterized in that, include: The data acquisition module is used to receive the sensing status parameters of the smart water meter to be tested in order to construct a feature vector; The classification processing module is used to perform clustering on feature vectors to generate a homogeneous batch queue of tests; The parameter mapping module is used to dynamically map the set of verification control parameters based on the cluster center vector of the batch queue. The blocking execution module is used to perform smart water meter verification according to the mapped set of verification control parameters. During the verification process, it records and analyzes error data streams in real time and performs predictive blocking operations. The execution of the predictive blocking operation includes: performing sliding window sampling and least squares linear regression on the real-time error data to calculate the slope and intercept of the current trend line; Based on the trend line, the predicted error value at the end of the verification is estimated and compared with the pass standard threshold. When the predicted error value exceeds the pass standard threshold and the confidence coefficient is greater than the safety threshold, a stop command is sent to the corresponding smart water meter and its data acquisition channel is cut off.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the method as described in any one of claims 1 to 7.
10. A computer device, characterized in that, The computer device includes a memory, a processor, and a program stored in and executable on the memory, the program being executed by the processor to implement the steps of the method as described in any one of claims 1 to 7.