Sand filter tank single-bin valve position adaptive control method and system

Abstract

The application discloses a sand filter single-bin valve position adaptive control method and system, and particularly relates to the field of sand filter single-bin valve position adaptive control, which comprises collecting single-bin liquid level signals at the edge end, constructing a first liquid level error, and combining a valve position action amount of the previous moment to generate a first local control factor; receiving first local control factors of multiple single bins at the cloud end, aggregating the first local control factors, counting group action trends within the same time window, identifying single-bin groups with converging actions, and generating a second global scheduling factor based on the identification result. By identifying multiple-bin group action trends at the cloud end and generating a global scheduling factor, the valve position adjustment result is solved at the edge end by combining the local liquid level error and the synchronous suppression factor, forming a closed-loop regulation mechanism of edge-cloud cooperation, so that the unity of single-bin adaptive control and global stability is realized.

Description

Technical Field

[0001] This invention relates to the field of adaptive control technology for single-compartment valve positions in sand filters, and more specifically, to a method and system for adaptive control of single-compartment valve positions in sand filters. Background Technology

[0002] In the current operation of sand filters in water plants, existing technologies generally adopt single-compartment valve position adaptive control. The edge controller will continuously adjust the valve position according to the change of liquid level to ensure the stability of the water intake of the compartment. This method can effectively suppress liquid level fluctuations under single-compartment conditions, which seems reasonable. However, in the actual multi-compartment parallel operation scenario, all compartments share the same clear water pool. When the upstream flow or operating status changes, the edge terminals of different compartments often make similar reactions at the same time, resulting in multiple valve positions acting almost synchronously.

[0003] This situation can cause overall fluctuations in the clear water tank level, frequent valve opening and closing, increased equipment wear and tear, and may also affect the quality of the effluent. Although existing technologies can achieve automatic control at the single-compartment level, they lack global coordination when multiple compartments are in parallel. Local optimization at the edge can trigger global instability.

[0004] With the introduction of the edge-cloud collaborative architecture, the cloud can provide global trend judgment, but there is a difference in rhythm between it and the real-time adjustment of the edge. The information updates in the cloud are slower, while the edge is too fast. The contradiction between the two makes this problem more prominent.

[0005] This analysis reveals that the core problem with existing technologies lies in the fact that the current adaptive control of single-compartment valve positions in sand filters lacks effective global constraints in an end-cloud collaborative environment, leading to a structural conflict between local regulation and global stability. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a single-compartment valve position adaptive control method and system for sand filters. By identifying the movement trends of multiple compartment groups in the cloud and generating a global scheduling factor, and combining local liquid level errors and synchronization suppression factors at the edge end to solve the valve position adjustment results, a closed-loop regulation mechanism of end-cloud collaboration is formed, thereby achieving the unity of single-compartment adaptive control and global stability, thus solving the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for adaptive control of a single-compartment valve position in a sand filter, comprising:

[0008] S1. Collect the liquid level signal of a single compartment at the edge end, construct the first liquid level error, and generate the first local control factor by combining it with the valve position action amount of the previous moment.

[0009] S2. Receive the first local control factors of multiple single warehouses in the cloud, aggregate the first local control factors, statistically analyze the group action trend within the same time window, identify single warehouse groups with similar actions, and generate a second global scheduling factor based on the identification results.

[0010] S3. Receive the second global scheduling factor at the edge end to form the third synchronization suppression factor, and combine the third synchronization suppression factor with the first liquid level error to generate the second local control factor.

[0011] S4. At the edge end, the valve position adjustment command is solved according to the second local control factor. When there is a cross-compartment action tendency, the valve position adjustment command is selectively constrained by the third synchronization inhibition factor by delaying, compressing the amplitude or raising the threshold, and the valve position action amount after execution is output.

[0012] S5. Collect the valve position action quantity and corresponding liquid level response of each single compartment in the cloud, construct the evaluation result of the overall liquid level stability of the system, combine the evaluation result with the execution data of each single compartment, generate the corrected second global scheduling factor, and send it down to the edge again to form a closed-loop adjustment process of end-cloud collaboration.

[0013] In a preferred embodiment, in S1, the generation of the first local control factor includes:

[0014] S1-1. Collect the liquid level signal at the current moment at the edge end, and compare the liquid level signal at the current moment with the preset liquid level value to construct the first liquid level error;

[0015] S1-2. At the edge end, call the valve position action amount of the previous moment, assign the first liquid level error and the valve position action amount to the preset first weight and second weight respectively, and perform a weighted summation operation to generate the first intermediate control amount.

[0016] S1-3. Perform normalization operation on the first intermediate control quantity at the edge end, and perform boundary constraint processing according to the upper and lower limits of the preset valve position action to limit the operation result to the allowable control range.

[0017] S1-4. At the edge end, the first intermediate control quantity after normalization and boundary constraint processing is directly output as the first local control factor.

[0018] In a preferred embodiment, S2 includes:

[0019] S2-1. Receive the first local control factors output from multiple single-warehouse edge terminals in the cloud and store them in a unified control dataset in chronological order.

[0020] S2-2. Perform aggregation operations on the control dataset in the cloud to construct a first aggregation vector within the same time window, and perform statistical operations on the first aggregation vector to obtain the group action trend; the aggregation operation includes performing time alignment and sampling period unification on the control dataset, grouping according to a preset time window, sorting according to the single warehouse identifier, and concatenating the first local control factors of each single warehouse.

[0021] S2-3. Based on the group action trend, identify single warehouse groups with similar actions within the same time window in the cloud, and combine the identification result of the single warehouse group with the group action trend to form a second judgment result;

[0022] S2-4. Generate a second global scheduling factor in the cloud based on the second determination result, and output the second global scheduling factor to each edge controller.

[0023] In a preferred embodiment, S3 includes:

[0024] S3-1. Receive the second global scheduling factor generated and distributed by the cloud at the edge end, perform parsing operation on the second global scheduling factor, extract the time window parameters, action direction parameters and action intensity parameters included therein, and write the extracted parameters into the local control parameter set at the edge end.

[0025] S3-2. At the edge end, perform numerical normalization processing based on the action intensity parameter to obtain a normalized intensity value. Based on the time window parameter, calculate the action direction consistency rate of each single compartment within the same time window. Multiply the normalized intensity value by the action direction consistency rate, perform boundary constraints according to the preset upper and lower limits, and output the third synchronization suppression factor.

[0026] S3-3. Obtain the absolute value of the first liquid level error at the edge end to obtain the error amplitude value; multiply the error amplitude value with the third synchronization suppression factor item by item to obtain the product result; multiply the product result by the preset weight coefficient to output the second intermediate control quantity.

[0027] In a preferred embodiment, S3 further includes:

[0028] S3-4. Perform delay counting processing on the second intermediate control quantity at the edge end to obtain the delay processing result; compress the delay processing result according to the preset compression ratio to obtain the compression processing result; perform boundary constraint processing on the compression processing result according to the upper and lower limits of the valve position action to obtain the second local control factor and output it.

[0029] In a preferred embodiment, in S4, the step of outputting the valve position actuation amount at the edge end includes:

[0030] S4-1. Receive the second local control factor at the edge end, add the second local control factor to the current valve position opening value to obtain the valve position adjustment base value, and use the valve position adjustment base value as the valve position adjustment command.

[0031] S4-2. At the edge, call the cross-warehouse action trend results obtained by the cloud. The cross-warehouse action trend consists of the action direction consistency rate and the average action intensity, which is used to characterize whether there are group actions with the same direction and similar intensity in multiple single warehouses within the same time window.

[0032] S4-3. When the consistency rate of the action direction of the cross-warehouse action trend reaches the preset threshold at the edge end, it is determined that there is a cross-warehouse action trend, and the third synchronization inhibition factor is activated to constrain the valve position adjustment command.

[0033] S4-4. Perform delay processing on the valve position adjustment command at the edge end. The delay processing includes setting a delay counter, maintaining the output of the valve position adjustment command at the previous moment when the count of the delay counter has not reached the preset delay value, and updating the current valve position adjustment command when the count of the delay counter reaches the preset delay value, thereby obtaining the delay processing result.

[0034] In a preferred embodiment, in S4, the step of outputting the valve position actuation amount at the edge end further includes:

[0035] S4-5. Perform amplitude compression processing on the delay processing result at the edge end. The amplitude compression processing includes multiplying the delay processing result by a preset compression coefficient to obtain the compression processing result.

[0036] S4-6. At the edge end, perform threshold boosting processing on the compression processing result. The threshold boosting processing includes boosting the compression processing result to the preset threshold when the compression processing result is less than the preset threshold, maintaining the original value when the compression processing result is greater than or equal to the threshold, and outputting the final valve position action amount.

[0037] In a preferred embodiment, in S5, the step of forming an end-to-cloud collaborative closed-loop regulation process in the cloud includes:

[0038] S5-1. Receive valve position action data output from the edge of each single compartment in the cloud, and synchronously receive liquid level response data of the corresponding time series. Align the data according to the time label to form a single compartment execution dataset.

[0039] S5-2. Perform statistical calculations on the single-compartment execution dataset in the cloud. The statistical calculations include performing root mean square error calculation on the difference between the liquid level response data and the set liquid level value, and performing cumulative calculations on the valve position action quantity, action frequency and action amplitude, to obtain the overall system liquid level stability evaluation result.

[0040] In a preferred embodiment, in S5, the step of forming an end-to-cloud collaborative closed-loop regulation process in the cloud further includes:

[0041] S5-3. Combine the overall system level stability evaluation results with the execution datasets of each individual compartment in the cloud to construct a global evaluation matrix. The global evaluation matrix includes the individual compartment level deviation index, the individual compartment valve position action frequency index, and the overall level stability index.

[0042] S5-4. On the cloud, based on the global evaluation matrix, perform correction calculations: adjust the action intensity parameter in the second global scheduling factor according to the weighted combination of the single-compartment liquid level deviation index and the overall liquid level stability index, adjust the time window parameter in the second global scheduling factor according to the single-compartment valve position action frequency index, and output the corrected second global scheduling factor.

[0043] S5-5. The modified second global scheduling factor is sent from the cloud to each edge terminal, so that the edge terminal calls the factor to update the control parameter set, thereby forming a closed-loop adjustment process of end-cloud collaboration.

[0044] An adaptive control system for single-compartment valve position in a sand filter includes a data acquisition module, an identification module, a combination module, an adjustment module, and an end-to-cloud module.

[0045] The acquisition module is used to acquire the liquid level signal of a single compartment at the edge, construct the first liquid level error, and generate the first local control factor by combining the valve position action amount of the previous moment.

[0046] The identification module is used to receive the first local control factors of multiple single warehouses in the cloud, aggregate the first local control factors, statistically analyze the group action trend within the same time window, identify single warehouse groups with similar actions, and generate a second global scheduling factor based on the identification results.

[0047] The joint module is used to receive the second global scheduling factor at the edge, form the third synchronization suppression factor, and combine the third synchronization suppression factor with the first liquid level error to generate the second local control factor.

[0048] The adjustment module is used to solve the valve position adjustment command at the edge based on the second local control factor, and when there is a cross-compartment action tendency, it selectively constrains the valve position adjustment command by delaying, compressing the amplitude or raising the threshold through the third synchronization inhibition factor, and outputs the valve position action amount after execution.

[0049] The edge-cloud module is used to collect the valve position action and corresponding liquid level response output of each single compartment in the cloud, construct the evaluation result of the overall liquid level stability of the system, combine the evaluation result with the execution data of each single compartment to generate a corrected second global scheduling factor, and then send it down to the edge again to form a closed-loop regulation process of edge-cloud collaboration.

[0050] The technical effects and advantages of this invention are as follows:

[0051] 1. In the parallel operation of multiple warehouses, this solution identifies the trend of group actions and generates a global scheduling factor in the cloud, and then introduces a synchronization inhibition mechanism at the edge to constrain valve position adjustment. This fundamentally solves the problem of group synchronous action caused by independent adjustment of a single warehouse in the existing technology, avoids the oscillation of the clear water tank level and the frequent opening and closing of valves, and achieves the unity of local stability and global coordination.

[0052] 2. This solution utilizes the liquid level error and the historical valve position action at the edge to construct a local control factor, and through normalization and boundary constraint processing, ensures that the valve position action result is always within a controllable range, thereby achieving a stable output of the valve position adjustment amount and avoiding execution deviations caused by excessively large or small calculation results.

[0053] 3. This solution can identify features such as the consistency rate of action direction and the average value of action intensity by aggregating and statistically calculating multiple local control factors in the cloud, thereby identifying the phenomenon of cross-warehouse action convergence, providing data support for the construction of global scheduling factors, and making global scheduling more targeted and constrained.

[0054] 4. This solution introduces specific constraint strategies such as delay processing, amplitude compression, and threshold enhancement at the edge, so that valve position adjustment not only depends on local liquid level error, but is also regulated by cross-compartment synchronization inhibition factor, thereby realizing the decentralization and differentiation of valve position execution and avoiding the phenomenon of group resonance.

[0055] 5. This solution constructs a global evaluation matrix in the cloud, which incorporates the single-compartment liquid level deviation index, valve position action frequency index and overall liquid level stability index into a unified structure, and corrects the global scheduling factor parameters based on this matrix, thereby realizing continuous optimization of edge control at the global level, enabling the system to evolve adaptively and maintain long-term stability.

[0056] 6. This solution ensures a dynamic balance between rapid response at the edge and global coordination in the cloud through a closed-loop operation mechanism of edge-cloud collaboration. It leverages the real-time performance of edge control while introducing global optimization capabilities from the cloud, thereby maintaining stable operation of the sand filter and extending the service life of the equipment under different operating conditions. Attached Figure Description

[0057] Figure 1 This is a flowchart of the method steps of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Refer to the instruction manual appendix Figure 1 An embodiment of the present invention provides an adaptive control method for a single-compartment valve position in a sand filter, comprising:

[0060] S1. Collect the liquid level signal of a single compartment at the edge end, construct the first liquid level error, and generate the first local control factor by combining it with the valve position action amount of the previous moment.

[0061] S2. Receive the first local control factors of multiple single warehouses in the cloud, aggregate the first local control factors, statistically analyze the group action trend within the same time window, identify single warehouse groups with similar actions, and generate a second global scheduling factor based on the identification results.

[0062] S3. Receive the second global scheduling factor at the edge end to form the third synchronization suppression factor, and combine the third synchronization suppression factor with the first liquid level error to generate the second local control factor.

[0063] S4. At the edge end, the valve position adjustment command is solved according to the second local control factor. When there is a cross-compartment action tendency, the valve position adjustment command is selectively constrained by the third synchronization inhibition factor by delaying, compressing the amplitude or raising the threshold, and the valve position action amount after execution is output.

[0064] S5. Collect the valve position action quantity and corresponding liquid level response of each single compartment in the cloud, construct the evaluation result of the overall liquid level stability of the system, combine the evaluation result with the execution data of each single compartment, generate the corrected second global scheduling factor, and send it down to the edge again to form a closed-loop adjustment process of end-cloud collaboration.

[0065] In S1, the generation of the first local control factor includes:

[0066] S1-1. Collect the liquid level signal at the current moment at the edge end, and compare the liquid level signal at the current moment with the preset liquid level value to construct the first liquid level error;

[0067] S1-2. At the edge end, call the valve position action amount of the previous moment, assign the first liquid level error and the valve position action amount to the preset first weight and second weight respectively, and perform a weighted summation operation to generate the first intermediate control amount.

[0068] S1-3. Perform normalization operation on the first intermediate control quantity at the edge end, and perform boundary constraint processing according to the upper and lower limits of the preset valve position action to limit the operation result to the allowable control range.

[0069] S1-4. At the edge end, the first intermediate control quantity after normalization and boundary constraint processing is directly output as the first local control factor.

[0070] For S1, it should be noted that the edge device refers to the computing and control node located on the equipment site, close to the sensors and actuators. It can directly collect data locally and perform real-time processing and control without relying on the computing results of a remote cloud. In other words, the edge device is a controller or computing unit that runs on-site in a factory, workshop or water plant to complete rapid response and local decision-making.

[0071] Valve position action refers to the change in valve opening within one control cycle, including the current valve opening value and the change in opening relative to the previous moment, usually measured as a percentage opening or angular displacement.

[0072] Normalization refers to scaling the original calculation results according to a preset scaling factor and benchmark range, so that the values ​​are linearly mapped to a unified standard range, such as 0 to 1 or -1 to 1, thereby ensuring that data from different sources are comparable and computable.

[0073] Boundary constraint processing based on the preset upper and lower limits of valve position action means comparing the calculated valve position action amount with the preset allowable range. If the value exceeds the upper limit, the upper limit is taken; if it is lower than the lower limit, the lower limit is taken, thus ensuring that the output result is always within the allowable control range. The final calculation result is a valve position action amount limited between the preset upper and lower limits, which is used to drive the valve to perform the action.

[0074] S2 includes:

[0075] S2-1. Receive the first local control factors output from multiple single-warehouse edge terminals in the cloud and store them in a unified control dataset in chronological order.

[0076] S2-2. Perform aggregation operations on the control dataset in the cloud to construct a first aggregated vector within the same time window, and perform statistical operations on the first aggregated vector to obtain the group action trend; the aggregation operation includes performing time alignment and sampling period unification on the control dataset, grouping according to a preset time window, sorting according to the single warehouse identifier, and concatenating the first local control factors of each single warehouse to construct the first aggregated vector of the time window; in addition, the statistical operation includes performing mean calculation, variance calculation, and similarity calculation on the first local control factors of each single warehouse within the same time window to reflect the overall action trend, the degree of action fluctuation, and the degree of convergence between single warehouses;

[0077] S2-3. Based on the group action trend, identify single warehouse groups with similar actions within the same time window in the cloud, and combine the identification result of the single warehouse group with the group action trend to form a second judgment result;

[0078] S2-4. Generate a second global scheduling factor in the cloud based on the second determination result, and output the second global scheduling factor to each edge controller;

[0079] The single-compartment edge terminal refers to a local control node deployed near a single filtration unit in the sand filter, directly connected to the sensors and actuators of that single compartment. It is responsible for collecting operating parameters such as liquid level signals, flow data, and valve opening in real time on site, and executing control algorithms locally to generate valve adjustment commands. Compared with the cloud, the single-compartment edge terminal is characterized by fast response speed and short processing cycle, and can complete closed-loop control in milliseconds to seconds, thereby ensuring the immediate stability of the liquid level in the single compartment. At the same time, it also uploads the locally generated control factors and execution data to the cloud for global scheduling analysis and cross-compartment collaboration. In other words, the single-compartment edge terminal is the layer closest to the physical site in the sand filter control system. It is both the main execution body of single-compartment control and the basic node for feeding back real-time status to the cloud in the edge-cloud collaborative architecture.

[0080] S3 includes:

[0081] S3-1. Receive the second global scheduling factor generated and distributed by the cloud at the edge end, perform parsing operation on the second global scheduling factor, extract the time window parameters, action direction parameters and action intensity parameters included therein, and write the extracted parameters into the local control parameter set at the edge end.

[0082] S3-2. At the edge end, perform numerical normalization processing based on the action intensity parameters to obtain a normalized intensity value. Then, based on the time window parameters, calculate the action direction consistency rate of each single warehouse within the same time window. Multiply the normalized intensity value by the action direction consistency rate, apply boundary constraints according to preset upper and lower limits, and output a third synchronization suppression factor. The normalized intensity value refers to the standardized value obtained by proportionally converting the action intensity parameters according to a preset benchmark range, used to ensure the comparability of action intensities among different single warehouses. The action direction consistency rate refers to the proportion of single warehouses whose valve position adjustment direction is the same as the dominant direction of the group within the same time window, out of the total number of single warehouses participating in the statistics within that time window.

[0083] S3-3. Obtain the absolute value of the first liquid level error at the edge end to obtain the error amplitude value; multiply the error amplitude value with the third synchronization suppression factor item by item to obtain the product result; multiply the product result by the preset weight coefficient to output the second intermediate control quantity.

[0084] S3 also includes:

[0085] S3-4. Perform delay counting processing on the second intermediate control quantity at the edge end to obtain the delay processing result; compress the delay processing result according to the preset compression ratio to obtain the compression processing result; perform boundary constraint processing on the compression processing result according to the upper and lower limits of the valve position action to obtain the second local control factor and output it.

[0086] Regarding S3-1, it should be noted that the time window parameter is used to determine the continuous period within which the cloud collects the control factors of each individual warehouse, thereby ensuring the consistency of the time range for the aggregation analysis; the action direction parameter is used to indicate whether the valve position adjustment of each individual warehouse is in the opening or closing direction within the time window, thereby reflecting the trend of the group action; and the action intensity parameter is used to indicate the magnitude of the valve position adjustment of each individual warehouse within the time window, thereby reflecting the strength of the group action.

[0087] In addition, the reason why the second global scheduling factor includes these parameters is that the cloud needs to transmit three key types of information at the same time: time range, action trend, and action intensity, so that the edge can accurately identify and suppress group convergent actions when constructing the synchronization inhibition factor.

[0088] In addition, according to the upper and lower limits of valve position action, after calculating the valve position action amount, it is compared with the preset allowable range, and the result is limited within the range so that the output value does not exceed the upper limit or fall below the lower limit.

[0089] Delay counting processing refers to: inputting the second intermediate control quantity into the delay counter; when the cumulative value of the counter has not reached the preset delay threshold, the output remains the control value of the previous moment; when the cumulative value of the counter reaches the delay threshold, it is updated to the current second intermediate control quantity to obtain the delay processing result.

[0090] Amplitude compression processing refers to multiplying the delay processing result by a preset compression coefficient to obtain a compression processing result, so that the amplitude of a single valve position change is limited to the range after compression.

[0091] Boundary constraint processing refers to comparing the compression processing result with the preset upper and lower limits of valve position action. If the value exceeds the upper limit, the upper limit is taken; if the value is lower than the lower limit, the lower limit is taken, and the second local control factor is obtained and output.

[0092] In S4, the steps for outputting the valve position actuation at the edge end include:

[0093] S4-1. Receive the second local control factor at the edge end, add the second local control factor to the current valve position opening value to obtain the valve position adjustment base value, and use the valve position adjustment base value as the valve position adjustment command. In practical applications, the valve position adjustment base value is input to the arithmetic unit of the edge end controller, written to the valve control register by the arithmetic unit, and then outputs the corresponding opening signal to the valve actuator by the valve control register, thereby driving the valve to move as the valve position adjustment command.

[0094] S4-2. At the edge, call the cross-warehouse action trend results obtained by the cloud. The cross-warehouse action trend consists of the action direction consistency rate and the average action intensity, which is used to characterize whether there are group actions with the same direction and similar intensity in multiple single warehouses within the same time window.

[0095] S4-3. When the consistency rate of the action direction of the cross-warehouse action trend reaches the preset threshold at the edge end, it is determined that there is a cross-warehouse action trend, and the third synchronization inhibition factor is activated to constrain the valve position adjustment command.

[0096] S4-4. Perform delay processing on the valve position adjustment command at the edge end. The delay processing includes setting a delay counter, maintaining the output of the valve position adjustment command at the previous moment when the count of the delay counter has not reached the preset delay value, and updating the current valve position adjustment command when the count of the delay counter reaches the preset delay value, thereby obtaining the delay processing result.

[0097] In S4, the step of outputting the valve position action at the edge end also includes:

[0098] S4-5. Perform amplitude compression processing on the delay processing result at the edge end. The amplitude compression processing includes multiplying the delay processing result by a preset compression coefficient to obtain the compression processing result.

[0099] S4-6. Perform threshold boosting processing on the compression processing result at the edge end. The threshold boosting processing includes boosting the compression processing result to the threshold when it is less than the preset threshold, and keeping the original value when the compression processing result is greater than or equal to the threshold. Output the final valve position action amount.

[0100] In S4-6, "raising to the threshold" means that when the value of the compression processing result is lower than the preset valve position action threshold, the low value will no longer be output directly, but will be replaced by the preset threshold itself as the output value. The increase is equal to the difference between the preset threshold and the current compression processing result. There is no proportional amplification process. Instead, the result is forcibly corrected to the threshold level in one go, so as to ensure that the valve action will not be lower than the set minimum executable action range.

[0101] In S5, the steps to form a closed-loop control process for end-to-end cloud collaboration include:

[0102] S5-1. Receive valve position action data output from the edge of each single compartment in the cloud, and synchronously receive liquid level response data of the corresponding time series. Align the data according to the time label to form a single compartment execution dataset.

[0103] S5-2. Perform statistical calculations on the single-compartment execution dataset in the cloud. The statistical calculations include performing root mean square error calculation on the difference between the liquid level response data and the set liquid level value, and performing cumulative calculations on the valve position action quantity, action frequency and action amplitude, to obtain the overall system liquid level stability evaluation result.

[0104] In S5, the steps for forming a closed-loop control process for end-to-end cloud collaboration also include:

[0105] S5-3. Combine the overall system level stability evaluation results with the execution datasets of each individual compartment in the cloud to construct a global evaluation matrix. The global evaluation matrix includes the individual compartment level deviation index, the individual compartment valve position action frequency index, and the overall level stability index.

[0106] S5-4. On the cloud, based on the global evaluation matrix, perform correction calculations: adjust the action intensity parameter in the second global scheduling factor according to the weighted combination of the single-compartment liquid level deviation index and the overall liquid level stability index, adjust the time window parameter in the second global scheduling factor according to the single-compartment valve position action frequency index, and output the corrected second global scheduling factor.

[0107] S5-5. The modified second global scheduling factor is sent from the cloud to each edge terminal, so that the edge terminal calls the factor to update the control parameter set, thereby forming a closed-loop adjustment process of end-cloud collaboration.

[0108] It should be noted in S5-3 that the reason why the global evaluation matrix can include single-compartment liquid level deviation index, single-compartment valve position action frequency index and overall liquid level stability index is that these three types of indexes describe the system operation status from three dimensions: local accuracy, local load and global coordination, respectively, and can jointly reflect the control effect of single compartment and the system as a whole under the end-cloud collaboration.

[0109] The process of constructing the global evaluation matrix is ​​as follows: First, the execution dataset of each single compartment is decomposed in the cloud. The mean square error between the liquid level response and the set value of each single compartment is calculated to obtain the liquid level deviation index of the single compartment. The number of actions and the cumulative value of the action amplitude of the valve position in the sampling window of the single compartment are counted to obtain the valve position action frequency index of the single compartment. Then, all the liquid level deviation indices of the single compartment are summarized and combined with the overall mean square error result to obtain the overall liquid level stability index.

[0110] The final global evaluation matrix uses individual warehouses as rows and indicator types as columns, storing the above three types of indicators uniformly in the matrix cells, thus forming a comprehensive evaluation structure that can simultaneously cover the operational status of individual warehouses and the overall stability level.

[0111] In S5-3, it should be noted that the global evaluation matrix is ​​read from the cloud, and the single-compartment liquid level deviation index and single-compartment valve position action frequency index are retrieved sequentially according to the single compartment order, while the overall liquid level stability index is retrieved simultaneously. For each single compartment, the weighted combination value of deviation and stability is calculated according to the preset weight coefficient, and the correction increment of the action intensity parameter is generated. This correction increment is added to the current action intensity parameter, and boundary constraints are applied according to the upper and lower limits of the action intensity parameter to form the corrected action intensity parameter. Then, based on the single compartment valve position action frequency index, the target value of the time window parameter is found in the preset mapping table, and the correction increment of the time window parameter is generated. This correction increment is added to the current time window parameter, and boundary constraints are applied according to the upper and lower limits of the time window parameter to form the corrected time window parameter. The corrected action intensity parameters and time window parameters of each single compartment are written into a new parameter set to generate the corrected second global scheduling factor. After being assigned a timestamp and version number, it is output and distributed to each edge terminal.

[0112] An adaptive control system for single-compartment valve position in a sand filter includes a data acquisition module, an identification module, a combination module, an adjustment module, and an end-to-cloud module.

[0113] The acquisition module is used to acquire the liquid level signal of a single compartment at the edge, construct the first liquid level error, and generate the first local control factor by combining the valve position action amount of the previous moment.

[0114] The identification module is used to receive the first local control factors of multiple single warehouses in the cloud, aggregate the first local control factors, statistically analyze the group action trend within the same time window, identify single warehouse groups with similar actions, and generate a second global scheduling factor based on the identification results.

[0115] The joint module is used to receive the second global scheduling factor at the edge, form the third synchronization suppression factor, and combine the third synchronization suppression factor with the first liquid level error to generate the second local control factor.

[0116] The adjustment module is used to solve the valve position adjustment command at the edge based on the second local control factor, and when there is a cross-compartment action tendency, it selectively constrains the valve position adjustment command by delaying, compressing the amplitude or raising the threshold through the third synchronization inhibition factor, and outputs the valve position action amount after execution.

[0117] The edge-cloud module is used to collect the valve position action and corresponding liquid level response output of each single compartment in the cloud, construct the evaluation result of the overall liquid level stability of the system, combine the evaluation result with the execution data of each single compartment to generate a corrected second global scheduling factor, and then send it down to the edge again to form a closed-loop regulation process of edge-cloud collaboration.

[0118] Working principle: This solution operates continuously in a closed-loop link under the edge-cloud collaborative architecture: The edge first collects the liquid level signal of a single compartment, compares the current liquid level with the preset liquid level value to construct the first liquid level error, calls the valve position action quantity of the previous moment, performs weighted summation according to the preset first weight and second weight to generate the first intermediate control quantity, performs normalization and upper and lower limit boundary constraints on the first intermediate control quantity, and outputs the first local control factor;

[0119] The cloud receives the first local control factor of each single warehouse in parallel, completes time alignment and window grouping, constructs the first aggregation vector of the same time window, and sequentially calculates the mean, variance and similarity to form a group action trend, identifies single warehouse groups with similar actions, generates the second global scheduling factor containing time window parameters, action direction parameters and action intensity parameters and sends it down.

[0120] After receiving the factor, the edge end analyzes three types of parameters, obtains the normalized intensity value based on the action intensity parameter, calculates the action direction consistency rate based on the time window parameter, multiplies the normalized intensity value with the action direction consistency rate and performs boundary constraints to form the third synchronization suppression factor, then takes the absolute value of the first liquid level error and multiplies it with the third synchronization suppression factor item by item and processes it according to the preset weight to obtain the second intermediate control quantity, and then sequentially performs delay counting, amplitude compression and boundary constraints to output the second local control factor;

[0121] At the edge, the valve position adjustment base value is obtained by adding the second local control factor to the current valve position opening. This value is written into the valve control register to form a valve position adjustment command. If there is a cross-compartment action trend, the third synchronization suppression factor is used to trigger delay processing, amplitude compression, or threshold increase, and the final valve position action amount is output. The cloud continuously collects the valve position action amount and liquid level response of each single compartment. After alignment, the root mean square error between the liquid level and the set value, as well as the cumulative frequency and amplitude of valve position action, are calculated. A global evaluation matrix containing single compartment liquid level deviation index, single compartment valve position action frequency index, and overall liquid level stability index is constructed. Based on this, the action intensity parameter and time window parameter in the second global scheduling factor are weighted and corrected and upper and lower limit constraints are applied. The corrected second global scheduling factor is generated and issued again, thereby realizing closed-loop operation of single compartment valve position self-adaptation and multi-compartment global stability under the collaboration of the edge and cloud.

[0122] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A sand filter single bin valve position adaptive control method, characterized by, include: S1. Collect the liquid level signal of a single compartment at the edge end, construct the first liquid level error, and generate the first local control factor by combining it with the valve position action amount of the previous moment. S2. Receive the first local control factors of multiple single warehouses in the cloud, aggregate the first local control factors, statistically analyze the group action trend within the same time window, identify single warehouse groups with similar actions, and generate a second global scheduling factor based on the identification results. S3. Receive the second global scheduling factor at the edge end to form the third synchronization suppression factor, and combine the third synchronization suppression factor with the first liquid level error to generate the second local control factor. S3 includes: S3-1. Receive the second global scheduling factor generated and distributed by the cloud at the edge end, perform parsing operation on the second global scheduling factor, extract the time window parameters, action direction parameters and action intensity parameters included therein, and write the extracted parameters into the local control parameter set at the edge end. S3-2. At the edge end, perform numerical normalization processing based on the action intensity parameter to obtain a normalized intensity value. Based on the time window parameter, calculate the action direction consistency rate of each single compartment within the same time window. Multiply the normalized intensity value by the action direction consistency rate, perform boundary constraints according to the preset upper and lower limits, and output the third synchronization suppression factor. S3-3. Obtain the absolute value of the first liquid level error at the edge end to obtain the error amplitude value; multiply the error amplitude value with the third synchronization suppression factor item by item to obtain the product result; multiply the product result by a preset weighting coefficient to output the second intermediate control quantity. S3-4. Perform delay counting processing on the second intermediate control quantity at the edge end to obtain the delay processing result; compress the delay processing result according to the preset compression ratio to obtain the compression processing result; perform boundary constraint processing on the compression processing result according to the upper and lower limits of the valve position action to obtain the second local control factor and output it. S4. At the edge end, the valve position adjustment command is solved according to the second local control factor. When there is a cross-compartment action tendency, the valve position adjustment command is selectively constrained by the third synchronization inhibition factor by delaying, compressing the amplitude or raising the threshold, and the valve position action amount after execution is output. S5. Collect the valve position action quantity and corresponding liquid level response of each single compartment in the cloud, construct the evaluation result of the overall liquid level stability of the system, combine the evaluation result with the execution data of each single compartment, generate the corrected second global scheduling factor, and send it down to the edge again to form a closed-loop adjustment process of end-cloud collaboration.

2. The adaptive control method for single-compartment valve position of a sand filter according to claim 1, characterized in that: In S1, the generation of the first local control factor includes: S1-1. Collect the liquid level signal at the current moment at the edge end, and compare the liquid level signal at the current moment with the preset liquid level value to construct the first liquid level error; S1-2. At the edge end, call the valve position action amount of the previous moment, assign the first liquid level error and the valve position action amount to the preset first weight and second weight respectively, and perform a weighted summation operation to generate the first intermediate control amount. S1-3. Perform normalization operation on the first intermediate control quantity at the edge end, and perform boundary constraint processing according to the upper and lower limits of the preset valve position action to limit the operation result to the allowable control range. S1-4. At the edge end, the first intermediate control quantity after normalization and boundary constraint processing is directly output as the first local control factor.

3. The adaptive control method for single-compartment valve position of a sand filter according to claim 2, characterized in that: S2 includes: S2-1. Receive the first local control factors output from multiple single-warehouse edge terminals in the cloud and store them in a unified control dataset in chronological order. S2-2. Perform aggregation operations on the control dataset in the cloud to construct a first aggregation vector within the same time window, and perform statistical operations on the first aggregation vector to obtain the group action trend; the aggregation operation includes performing time alignment and sampling period unification on the control dataset, grouping according to a preset time window, sorting according to the single warehouse identifier, and concatenating the first local control factors of each single warehouse. S2-3. Based on the group action trend, identify single warehouse groups with similar actions within the same time window in the cloud, and combine the identification result of the single warehouse group with the group action trend to form a second judgment result; S2-4. Generate a second global scheduling factor in the cloud based on the second determination result, and output the second global scheduling factor to each edge controller.

4. The adaptive control method for single-compartment valve position of a sand filter according to claim 3, characterized in that: In S4, the steps for outputting the valve position actuation at the edge end include: S4-1. Receive the second local control factor at the edge end, add the second local control factor to the current valve position opening value to obtain the valve position adjustment base value, and use the valve position adjustment base value as the valve position adjustment command. S4-2. At the edge, call the cross-warehouse action trend results obtained by the cloud. The cross-warehouse action trend results consist of the action direction consistency rate and the average action intensity, which are used to characterize whether there are group actions with the same direction and similar intensity in multiple single warehouses within the same time window. S4-3. When the consistency rate of the cross-position action trend results reaches a preset threshold at the edge, it is determined that there are cross-position action convergence results, and the third synchronization inhibition factor is activated to constrain the valve position adjustment command. S4-4. Perform delay processing on the valve position adjustment command at the edge end. The delay processing includes setting a delay counter, maintaining the output of the valve position adjustment command at the previous moment when the count of the delay counter has not reached the preset delay value, and updating the current valve position adjustment command when the count of the delay counter reaches the preset delay value, thereby obtaining the delay processing result.

5. The adaptive control method for single-compartment valve position of a sand filter according to claim 4, characterized in that: In S4, the step of outputting the valve position action at the edge end also includes: S4-5. Perform amplitude compression processing on the delay processing result at the edge end. The amplitude compression processing includes multiplying the delay processing result by a preset compression coefficient to obtain the compression processing result. S4-6. At the edge end, perform threshold boosting processing on the compression processing result. The threshold boosting processing includes boosting the compression processing result to the preset threshold when the compression processing result is less than the preset threshold, maintaining the original value when the compression processing result is greater than or equal to the threshold, and outputting the final valve position action amount.

6. The adaptive control method for single-compartment valve position of a sand filter according to claim 5, characterized in that: In S5, the steps to form a closed-loop control process for end-to-end cloud collaboration include: S5-1. Receive valve position action data output from the edge of each single compartment in the cloud, and synchronously receive liquid level response data of the corresponding time series. Align the data according to the time label to form a single compartment execution dataset. S5-2. Perform statistical calculations on the single-compartment execution dataset in the cloud. The statistical calculations include performing root mean square error calculation on the difference between the liquid level response data and the set liquid level value, and performing cumulative calculations on the valve position action quantity, action frequency and action amplitude, to obtain the overall system liquid level stability evaluation result.

7. The adaptive control method for single-compartment valve position of a sand filter according to claim 6, characterized in that: In S5, the steps for forming a closed-loop control process for end-to-end cloud collaboration also include: S5-3. Combine the overall system level stability evaluation results with the execution datasets of each individual compartment in the cloud to construct a global evaluation matrix. The global evaluation matrix includes the individual compartment level deviation index, the individual compartment valve position action frequency index, and the overall level stability index. S5-4. On the cloud, based on the global evaluation matrix, perform correction calculations: adjust the action intensity parameter in the second global scheduling factor according to the weighted combination of the single-compartment liquid level deviation index and the overall liquid level stability index, adjust the time window parameter in the second global scheduling factor according to the single-compartment valve position action frequency index, and output the corrected second global scheduling factor. S5-5. The modified second global scheduling factor is sent from the cloud to each edge terminal, so that the edge terminal calls the second global scheduling factor to update the control parameter set, thereby forming a closed-loop adjustment process of end-cloud collaboration.

8. A single-compartment valve position adaptive control system for a sand filter, used to implement the single-compartment valve position adaptive control method for a sand filter as described in claim 1, comprising a data acquisition module, an identification module, a joint module, an adjustment module, and an end-to-end cloud module, characterized in that: The acquisition module is used to acquire the liquid level signal of a single compartment at the edge, construct the first liquid level error, and generate the first local control factor by combining the valve position action amount of the previous moment. The identification module is used to receive the first local control factors of multiple single warehouses in the cloud, aggregate the first local control factors, statistically analyze the group action trend within the same time window, identify single warehouse groups with similar actions, and generate a second global scheduling factor based on the identification results. The joint module is used to receive the second global scheduling factor at the edge, form the third synchronization suppression factor, and combine the third synchronization suppression factor with the first liquid level error to generate the second local control factor. The adjustment module is used to solve the valve position adjustment command at the edge based on the second local control factor, and when there is a cross-compartment action tendency, it selectively constrains the valve position adjustment command by delaying, compressing the amplitude or raising the threshold through the third synchronization inhibition factor, and outputs the valve position action amount after execution. The edge-cloud module is used to collect the valve position action and corresponding liquid level response output of each single compartment in the cloud, construct the evaluation result of the overall liquid level stability of the system, combine the evaluation result with the execution data of each single compartment to generate a corrected second global scheduling factor, and then send it down to the edge again to form a closed-loop regulation process of edge-cloud collaboration.

Images