Cloud control system and method for inverted umbrella-shaped aeration equipment

Abstract

The invention discloses a cloud control system and method for inverted-umbrella-shaped aeration equipment. The method comprises the steps that S1, state parameters of the inverted-umbrella-shaped aeration equipment and sewage quality parameters are collected; s2, converting the signal into a digital monitoring signal sequence and preprocessing the digital monitoring signal sequence; s3, an improved FEDform network is adopted to execute edge control reasoning, and an operation control instruction is generated; s4, executing the operation control instruction and collecting operation feedback data, and if abnormality is detected, performing abnormality correction; s5, uploading the operation control instruction, the operation feedback data, the current sewage quality parameter and the current equipment state parameter to a cloud processing platform, and generating a structured monitoring data packet; s6, performing control rule reasoning and behavior pattern recognition on the structured monitoring data packet through a Mamdani fuzzy reasoning method, and generating an optimization control strategy; and S7, generating operation state visualization information and pushing the information to the user terminal. According to the invention, the control capability, the energy consumption optimization efficiency and the remote operation and maintenance capability of the inverted umbrella-shaped aeration equipment are improved.

Description

Technical Field

[0001] This invention relates to the field of industrial automation and intelligent control technology, and in particular to a cloud-based control system and method for an inverted umbrella-type aeration device. Background Technology

[0002] With the increasing demands for automation and intelligence in the wastewater treatment industry, inverted umbrella-type aeration equipment has been widely used in typical wastewater treatment processes such as oxidation ditches due to its simple structure, high aeration efficiency, and relatively low energy consumption. However, the following problems commonly exist in practical applications:

[0003] The control methods of existing inverted umbrella-type aeration equipment mainly rely on manual experience to set start-stop times and operating intensity, lacking the ability to dynamically match the wastewater treatment load with perception and regulation capabilities. This results in high energy consumption during the aeration process and large fluctuations in effluent quality. Most existing control systems rely on local PLC logic judgment for timed start-stop control or fixed strategy operation, which cannot perform in-depth analysis of multi-source operating data and lacks anomaly identification and feedback correction mechanisms. Existing control systems generally rely on on-site manual inspection and data recording, with weak communication capabilities and remote manageability, making it difficult to meet the current industrial IoT environment's needs for visualized equipment operating status, remote management, and intelligent diagnosis. In terms of multi-device collaborative operation and energy consumption optimization management, traditional control systems lack edge computing and cloud strategy linkage mechanisms, resulting in delayed response of control strategies and difficulty in achieving global optimization of operating parameters. Although some systems support data uploading to the cloud, they lack the ability to perform structured processing and semantic recognition of operating data, resulting in severely insufficient intelligent optimization and adaptive control capabilities, making it difficult to support large-scale operation and maintenance management needs.

[0004] Therefore, how to provide a cloud-based control system and method for an inverted umbrella-type aeration device is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a cloud-based control system and method for an inverted umbrella-type aeration device. This invention fully integrates edge control inference, improved FEDformer network, fuzzy rule inference, and cloud visualization technology. It describes in detail the operation control algorithm and system architecture for adaptive regulation of wastewater treatment load, and has the advantages of high intelligence, fast control response, strong energy consumption optimization, and good remote manageability.

[0006] A cloud-based control method for an inverted umbrella-type aeration device according to an embodiment of the present invention includes the following steps:

[0007] S1. Collect the status parameters of the inverted umbrella-type aeration equipment and the wastewater quality parameters to generate a simulated monitoring data sequence;

[0008] S2. Convert the analog monitoring data sequence into a digital monitoring signal sequence, and preprocess the digital monitoring signal sequence to form a standardized dataset;

[0009] S3. An improved FEDformer network is used to perform edge control inference on the standardized dataset to generate operation control commands for the inverted umbrella-type aeration equipment.

[0010] S4. Execute the operation control command and collect operation feedback data in real time. If the operation feedback data is found to be abnormal, perform abnormal correction processing.

[0011] S5. Upload the operation control commands, operation feedback data, current wastewater quality parameters, and current equipment status parameters to the cloud processing platform to generate a structured monitoring data package;

[0012] S6. In the cloud processing platform, the Mamdani fuzzy reasoning method is used to perform control rule reasoning and behavior pattern recognition on structured monitoring data packets to generate optimized control strategies for operation control.

[0013] S7. Optimize and classify control strategies and historical data, generate visual information on operating status, and push it to user terminals.

[0014] Optionally, step S1 specifically includes:

[0015] S11. Deploy dissolved oxygen sensor, ammonia nitrogen sensor, total nitrogen sensor, power monitoring module, water level sensor, vibration sensor, temperature sensor and blower status sensor on the inverted umbrella aeration equipment to simultaneously collect the status parameters of the inverted umbrella aeration equipment and the wastewater quality parameters.

[0016] S12. The state parameters include the fan start / stop status, power consumption value, shell vibration amplitude and equipment surface temperature; the wastewater quality parameters include dissolved oxygen concentration, ammonia nitrogen concentration, total nitrogen concentration and water level value.

[0017] S13. Remove abrupt outliers from the state parameters and wastewater quality parameters, correct baseline drift data and complete missing data, and perform time-series alignment based on a unified timestamp to generate a simulated monitoring data sequence.

[0018] Optionally, step S2 specifically includes:

[0019] S21. Perform analog-to-digital conversion on the analog monitoring data sequence using an analog-to-digital converter to generate the corresponding digital monitoring signal sequence;

[0020] S22. The preprocessing includes: performing unit conversion, format regularization and dimension unification on the digital monitoring signal sequence, and performing missing data completion, outlier limiting and noise filtering.

[0021] S23. Add a unified timestamp label to the preprocessed digital monitoring signal sequence and perform time synchronization to form a standardized dataset, which includes numerical variables and categorical variables.

[0022] Optionally, step S3 specifically includes:

[0023] S31, The improved FEDformer network includes a fusion coding module, a trend-period-perturbation decomposition module, a frequency domain attention block, a control command decoding module, and a feedback injection module;

[0024] S32. The fusion encoding module performs feature encoding operations on the standardized dataset, specifically as follows:

[0025] Numerical variables are linearly projected through a fully connected layer to an embedding space with uniform dimensions, generating a numerical embedding tensor. Categorical variables are one-hot encoded and vector-mapped using a pre-defined embedding lookup matrix, generating a categorical embedding tensor.

[0026] Based on the position index at each time step, a position encoding tensor is constructed using sine and cosine functions;

[0027] The numerical embedding tensor, the class embedding tensor, and the position encoding tensor are concatenated along the feature dimension to construct a fused feature tensor;

[0028] S33. The trend-period-perturbation decomposition module performs multi-scale decomposition processing on the fused feature tensor, specifically as follows:

[0029] The fused feature tensor is sliced ​​into sequences based on a sliding time window of fixed length to obtain several local subsequence blocks;

[0030] For each local subsequence block, a one-dimensional convolution operation is performed along the time axis to extract the smooth change trend component and generate a trend component tensor. Based on the bandpass filter and period detection kernel, the main periodic pattern is identified, the periodic change structure is extracted, and a periodic component tensor is generated. The corresponding trend component tensor and periodic component tensor are subtracted from the local subsequence block respectively to obtain the perturbation residual component tensor.

[0031] The trend component tensor, the periodic component tensor, and the perturbation residual tensor are concatenated along the feature dimension to generate the enhanced temporal decomposition tensor.

[0032] S34. The frequency domain attention block performs a fast Fourier transform on the enhanced temporal decomposition tensor to obtain a frequency domain mapping tensor.

[0033] Extract the magnitude of each frequency component in the frequency domain mapping tensor, sort them in descending order of magnitude, select the K frequency indices with the largest magnitude to form the main frequency component set, and extract the corresponding frequency components from the frequency domain mapping tensor based on the main frequency component set to form the main frequency tensor;

[0034] The amplitude of the main frequency tensor is normalized, and a frequency domain attention matrix is ​​generated by the Softmax function. The main frequency tensor and the frequency domain attention matrix are combined element-wise with weighted combination to generate a frequency domain enhancement tensor. The frequency domain enhancement tensor is subjected to inverse fast Fourier transform to generate a time-series reconstruction tensor.

[0035] S35. The control command decoding module includes four parallel control prediction branches, a boundary constraint transformer, and a command encoding unit.

[0036] The four parallel control prediction branches receive the time-series reconstruction tensor respectively, and perform linear compression, nonlinear mapping and activation output within their respective structures to generate wind turbine start-stop control quantity, operation intensity control quantity, power optimization coefficient and target dissolved oxygen setpoint respectively.

[0037] The boundary constraint transformer performs threshold judgment and binarization operations on the wind turbine start-stop control quantity. When the wind turbine start-stop control quantity exceeds the set activation threshold, the wind turbine start-stop control flag is output as 1: activation signal; otherwise, the control flag is output as 0: deactivation signal.

[0038] The boundary constraint transformer performs inverse normalization mapping, upper and lower limit limiting processing, and numerical format conversion operations on the operation intensity control quantity to generate operation intensity control instructions; performs range mapping on the power optimization coefficient to generate power optimization control instructions; and performs linear limiting operation on the target dissolved oxygen setpoint to generate dissolved oxygen setpoint control instructions.

[0039] The instruction encoding unit performs field identification, splicing, and communication protocol encapsulation on the wind turbine start / stop control flag, operation intensity control instruction, power optimization control instruction, and dissolved oxygen setting control instruction to form operation control instructions.

[0040] Optionally, step S4 specifically includes:

[0041] S41. Send the operation control command to the inverted umbrella-type aeration equipment for execution control;

[0042] S42. Real-time collection of operation feedback data of the inverted umbrella aeration device through the feedback injection module of the improved FEDformer network. The operation feedback data includes the start / stop status of the blower, operation intensity, power consumption value and dissolved oxygen concentration.

[0043] S43. When the runtime feedback data contains abrupt outliers, consecutive missing values, or physical inconsistencies, perform anomaly correction processing:

[0044] Based on a fixed-length sliding time window, the mean and standard deviation of each type of operational feedback data are dynamically calculated. If there are operational feedback data that exceed the mean ± 3 times the standard deviation, they are marked as abrupt outliers, and time-series interpolation is used to reconstruct missing points or correct outliers.

[0045] If consecutive missing values ​​exceed the set threshold window length, interpolation will be used to complete the data.

[0046] If there is a logical conflict between the operation feedback data and the physical response of the operation control command, it will be marked as a physical inconsistency and an alarm signal will be generated and reported to the cloud platform.

[0047] S44. Perform preprocessing on the operational feedback data after anomaly correction, and append it to the end of the standardized dataset before continuing to input it into the improved FEDformer network for edge control inference.

[0048] Optionally, step S5 specifically includes:

[0049] S51. Call the communication interface to package the operation control command, operation feedback data, current sewage water quality parameters and current equipment status parameters into a structured monitoring data package;

[0050] S52. The structured monitoring data packets are encrypted using the national cryptographic symmetric encryption algorithm SM4 and sent to the cloud processing platform via a wireless communication network.

[0051] S53. Perform field-level parsing and classification storage on the structured monitoring data packets in the cloud processing platform, and build a mapping table for operation control command fields, operation feedback fields, sewage water quality parameter fields and equipment status parameter fields.

[0052] Optionally, step S6 specifically includes:

[0053] S61. Construct fuzzy input variables, including dissolved oxygen concentration, ammonia nitrogen concentration, fan start / stop status and operating intensity;

[0054] S62. The dissolved oxygen concentration, ammonia nitrogen concentration and operating intensity are expressed using triangular membership functions, and the fan start-stop status is expressed using trapezoidal membership functions.

[0055] S63. Construct fuzzy linguistic values ​​of "low", "medium" and "high" for dissolved oxygen concentration and ammonia nitrogen concentration, construct fuzzy linguistic values ​​of "on" and "off" for fan start-up and shutdown status, and construct fuzzy linguistic values ​​of "weak", "medium" and "strong" for operating intensity.

[0056] S64. Based on expert knowledge and operational experience, set fuzzy rules between input and output, and construct a fuzzy rule library. The fuzzy rule library includes at least 27 fuzzy rules for outputting optimized control strategies. The optimized control strategies include increasing the wind turbine operating intensity, decreasing the wind turbine operating intensity, maintaining the current operating state, or triggering an abnormal flag.

[0057] S65. Based on the fuzzy input variables and the fuzzy rule base, the minimum-maximum composite reasoning method is used to perform composite reasoning on all fuzzy rules to obtain the fuzzy membership degree distribution function. The minimum-maximum composite reasoning specifically means that the fuzzy premise membership degree is minimized and the output conclusion is maximized.

[0058] S66. Demodel the fuzzy membership distribution function using the centroid method to obtain and output the corresponding optimized control strategy.

[0059] Optionally, step S7 specifically includes:

[0060] S71. The historical data includes: historical operation control commands, historical operation feedback data, wastewater quality parameters, equipment status parameters, and historical optimization control strategies;

[0061] S72. The operational status visualization information includes operational trend curves, energy consumption analysis charts, and alarm records;

[0062] S73. The visualized operating status information is pushed to the user terminal in real time through APP client, web page or WeChat official account.

[0063] A cloud-based control system for an inverted umbrella-type aeration device according to an embodiment of the present invention includes:

[0064] The data acquisition module is used to collect the status parameters and wastewater quality parameters of the inverted umbrella-type aeration equipment and generate simulated monitoring data sequences.

[0065] The data preprocessing module is used to convert analog monitoring data sequences into digital monitoring signal sequences and perform preprocessing operations to generate standardized datasets.

[0066] The edge control inference module is used to perform edge control inference and generate operation control commands;

[0067] The control execution and feedback module is used to execute operation control commands, collect operation feedback data, detect anomalies, and perform corrective processing.

[0068] The cloud-based data processing module is used to receive operation control commands, operation feedback data, current wastewater quality parameters, and current equipment status parameters, and to construct structured monitoring data packages.

[0069] The fuzzy inference module is used to perform fuzzy rule inference and behavior pattern recognition to generate optimized control strategies.

[0070] The visualization push module is used to generate visual information about the running status and push it to the user terminal in real time.

[0071] The beneficial effects of this invention are:

[0072] This invention combines edge computing and cloud-based intelligent inference to achieve closed-loop regulation and intelligent optimization of equipment operation control. First, by deploying multiple types of sensors, such as dissolved oxygen, ammonia nitrogen, total nitrogen, electrical energy, water level, vibration, temperature, and fan status, the status parameters of the inverted umbrella-type aeration equipment and wastewater quality parameters are collected. Then, through analog-to-digital conversion and multi-level preprocessing, a standardized dataset with consistent time sequence and dimensionality is formed.

[0073] Secondly, an improved FEDformer network is used to generate operation control commands at the edge. The improved FEDformer network introduces fusion coding, trend-period-disturbance decomposition, frequency domain attention and multi-branch control command decoding, which improves the ability to extract key control features from multimodal time series data. It also improves the dynamic response capability to complex operating states through main frequency enhancement and feedback injection, and realizes efficient inference of wind turbine start-up and shutdown, operating intensity, power optimization coefficient and dissolved oxygen setpoint.

[0074] Furthermore, this invention constructs an anomaly detection and correction process for operational feedback data, possessing the capabilities to identify sudden anomalies, complete continuous missing data, and judge physical inconsistencies. It can automatically perform corrections and issue alarms when feedback data is abnormal, ensuring the accuracy and continuity of edge control inference. Simultaneously, after receiving structured monitoring data packets, the cloud processing platform uses the Mamdani fuzzy inference method to construct a fuzzy rule base and identify operational behavior patterns, generating optimized control strategies and linking them with historical data to achieve remote strategy delivery and visualized display of operational status.

[0075] In summary, this invention achieves intelligent, data-driven, and remote management of the entire operation control process of the inverted umbrella-type aeration equipment, significantly improving the energy efficiency and automatic control capabilities of the aeration system, and meeting the comprehensive needs of modern wastewater treatment plants for operational reliability, intelligent diagnosis, and energy consumption optimization. Attached Figure Description

[0076] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0077] Figure 1 This is a schematic diagram of a cloud control method for an inverted umbrella-type aeration device proposed in this invention;

[0078] Figure 2 This is a schematic diagram of the improved FEDformer network structure in this invention;

[0079] Figure 3 This is a schematic diagram of the trend-cycle-perturbation decomposition module structure in this invention;

[0080] Figure 4 This is a flowchart of the execution of operation control commands and the update of feedback data in this invention. Detailed Implementation

[0081] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0082] refer to Figure 1-4 A cloud-based control method for an inverted umbrella-type aeration device includes the following steps:

[0083] S1. Collect the status parameters of the inverted umbrella-type aeration equipment and the wastewater quality parameters to generate a simulated monitoring data sequence;

[0084] S2. Convert the analog monitoring data sequence into a digital monitoring signal sequence, and preprocess the digital monitoring signal sequence to form a standardized dataset;

[0085] S3. An improved FEDformer network is used to perform edge control inference on the standardized dataset to generate operation control commands for the inverted umbrella-type aeration equipment.

[0086] S4. Execute the operation control command and collect operation feedback data in real time. If the operation feedback data is found to be abnormal, perform abnormal correction processing.

[0087] S5. Upload the operation control commands, operation feedback data, current wastewater quality parameters, and current equipment status parameters to the cloud processing platform to generate a structured monitoring data package;

[0088] S6. In the cloud processing platform, the Mamdani fuzzy reasoning method is used to perform control rule reasoning and behavior pattern recognition on structured monitoring data packets to generate optimized control strategies for operation control.

[0089] S7. Optimize and classify control strategies and historical data, generate visual information on operating status, and push it to user terminals.

[0090] In this embodiment, step S1 specifically includes:

[0091] S11. Deploy dissolved oxygen sensor, ammonia nitrogen sensor, total nitrogen sensor, power monitoring module, water level sensor, vibration sensor, temperature sensor and blower status sensor on the inverted umbrella aeration equipment to simultaneously collect the status parameters of the inverted umbrella aeration equipment and the wastewater quality parameters.

[0092] S12. The state parameters include the fan start / stop status, power consumption value, shell vibration amplitude and equipment surface temperature; the wastewater quality parameters include dissolved oxygen concentration, ammonia nitrogen concentration, total nitrogen concentration and water level value.

[0093] S13. Remove abrupt outliers from the state parameters and wastewater quality parameters, correct baseline drift data and complete missing data, and perform time-series alignment based on a unified timestamp to generate a simulated monitoring data sequence.

[0094] In this embodiment, step S2 specifically includes:

[0095] S21. Perform analog-to-digital conversion on the analog monitoring data sequence using an analog-to-digital converter to generate the corresponding digital monitoring signal sequence;

[0096] S22. The preprocessing includes: performing unit conversion, format regularization and dimension unification on the digital monitoring signal sequence, and performing missing data completion, outlier amplitude limiting and noise filtering.

[0097] S23. Add a unified timestamp label to the preprocessed digital monitoring signal sequence and perform time synchronization to form a standardized dataset, which includes numerical variables and categorical variables.

[0098] In this embodiment, step S3 specifically includes:

[0099] S31, The improved FEDformer network includes a fusion coding module, a trend-period-perturbation decomposition module, a frequency domain attention block, a control command decoding module, and a feedback injection module;

[0100] S32. The fusion encoding module performs feature encoding operations on the standardized dataset, specifically as follows:

[0101] Numerical variables are linearly projected through a fully connected layer to an embedding space with uniform dimensions, generating a numerical embedding tensor.

[0102] One-hot encoding is performed on categorical variables, and vector mapping is performed through a preset embedding lookup matrix to generate a categorical embedding tensor;

[0103] Based on the position index at each time step, a position encoding tensor is constructed using sine and cosine functions;

[0104] The numerical embedding tensor, the class embedding tensor, and the position encoding tensor are concatenated along the feature dimension to construct a fused feature tensor;

[0105] S33. The trend-period-perturbation decomposition module performs multi-scale decomposition processing on the fused feature tensor, specifically as follows:

[0106] The fused feature tensor is sliced ​​into sequences based on a sliding time window of fixed length to obtain several local subsequence blocks;

[0107] For each local subsequence block, a one-dimensional convolution operation is performed along the time axis to extract the smooth change trend components and generate trend component tensors;

[0108] For each local subsequence block, the main periodic pattern is identified based on the bandpass filter and periodic detection kernel, the periodic change structure is extracted, and the periodic component tensor is generated;

[0109] Subtract the corresponding trend component tensor and period component tensor from the local subsequence block to obtain the perturbation residual component tensor.

[0110] The trend component tensor, the periodic component tensor, and the perturbation residual tensor are concatenated along the feature dimension to generate an enhanced temporal decomposition tensor.

[0111] S34. The frequency domain attention block performs a fast Fourier transform on the enhanced temporal decomposition tensor to obtain a frequency domain mapping tensor.

[0112] Extract the magnitude of each frequency component in the frequency domain mapping tensor, sort them in descending order of magnitude, select the K frequency indices with the largest magnitude to form the main frequency component set, and extract the corresponding frequency components from the frequency domain mapping tensor based on the main frequency component set to form the main frequency tensor;

[0113] The amplitude of the main frequency tensor is normalized, and a frequency domain attention matrix is ​​generated by the Softmax function. The main frequency tensor and the frequency domain attention matrix are combined element-wise with weighted combination to generate a frequency domain enhancement tensor. The frequency domain enhancement tensor is subjected to inverse fast Fourier transform to generate a time-series reconstruction tensor.

[0114] S35. The control command decoding module includes four parallel control prediction branches, a boundary constraint transformer and a command encoding unit. Each control prediction branch includes a feature compression layer, a nonlinear mapping layer and a control output layer.

[0115] The feature compression layer adopts a fully connected structure, the nonlinear mapping layer adopts a multilayer perceptron structure, and the control output layer adopts the Sigmoid activation function, ReLU activation function, Tanh activation function and linear activation function respectively;

[0116] The four parallel control prediction branches receive the time-series reconstruction tensor respectively, and perform linear compression, nonlinear mapping and activation output within their respective structures to generate wind turbine start-stop control quantity, operation intensity control quantity, power optimization coefficient and target dissolved oxygen setpoint respectively.

[0117] The boundary constraint transformer performs threshold judgment and binarization operations on the wind turbine start-stop control quantity. When the wind turbine start-stop control quantity exceeds the set activation threshold, the wind turbine start-stop control flag is output as 1: activation signal; otherwise, the control flag is output as 0: deactivation signal.

[0118] The boundary constraint transformer performs inverse normalization mapping, upper and lower limit amplitude processing, and numerical format conversion operations on the running intensity control quantity to generate running intensity control instructions.

[0119] The boundary constraint transformer performs range mapping on the power optimization coefficients and generates power optimization control commands.

[0120] The boundary constraint transformer performs a linear limiting operation on the target dissolved oxygen setpoint and generates a dissolved oxygen setpoint control command.

[0121] The instruction encoding unit performs field identification, splicing, and communication protocol encapsulation on the wind turbine start / stop control flag, operation intensity control instruction, power optimization control instruction, and dissolved oxygen setting control instruction to form operation control instructions.

[0122] In this invention, the improved FEDformer network combines fusion coding, trend-period-perturbation decomposition, and frequency domain attention mechanism to enhance the modeling capability of non-stationary time-series data in wastewater treatment operations. It can accurately identify the trend of operating status changes and the main periodic law, while effectively suppressing disturbance noise. By constructing a multi-branch output structure through the control command decoding module, it realizes parallel prediction of blower start-up and shutdown, operating intensity, power optimization coefficient, and target dissolved oxygen setpoint, which significantly improves the accuracy and real-time performance of control decisions and effectively enhances the energy saving and stability of the aeration process.

[0123] In this embodiment, step S4 specifically includes:

[0124] S41. Send the operation control command to the inverted umbrella-type aeration equipment for execution control;

[0125] S42. Real-time collection of operation feedback data of the inverted umbrella aeration device through the feedback injection module of the improved FEDformer network. The operation feedback data includes the start / stop status of the blower, operation intensity, power consumption value and dissolved oxygen concentration.

[0126] S43. When the runtime feedback data contains abrupt outliers, consecutive missing values, or physical inconsistencies, perform anomaly correction processing:

[0127] Based on a fixed-length sliding time window, the mean and standard deviation of each type of operational feedback data are dynamically calculated. If there are operational feedback data that exceed the mean ± 3 times the standard deviation, they are marked as abrupt outliers, and time-series interpolation is used to reconstruct missing points or correct outliers.

[0128] If consecutive missing values ​​exceed the set threshold window length, interpolation will be used to complete the data.

[0129] If there is a logical conflict between the operation feedback data and the physical response of the operation control command, it will be marked as a physical inconsistency and an alarm signal will be generated and reported to the cloud platform.

[0130] S44. Perform preprocessing on the operational feedback data after anomaly correction, and append it to the end of the standardized dataset before continuing to input it into the improved FEDformer network for edge control inference.

[0131] In this embodiment, step S5 specifically includes:

[0132] S51. Call the communication interface to package the operation control command, operation feedback data, current sewage water quality parameters and current equipment status parameters into a structured monitoring data package;

[0133] S52. The structured monitoring data packets are encrypted using the national cryptographic symmetric encryption algorithm SM4 and sent to the cloud processing platform via a wireless communication network.

[0134] S53. Perform field-level parsing and classification storage on the structured monitoring data packets in the cloud processing platform, and build a mapping table for operation control command fields, operation feedback fields, sewage water quality parameter fields and equipment status parameter fields.

[0135] In this embodiment, step S6 specifically includes:

[0136] S61. Construct fuzzy input variables, including dissolved oxygen concentration, ammonia nitrogen concentration, fan start / stop status and operating intensity;

[0137] S62. The dissolved oxygen concentration, ammonia nitrogen concentration and operating intensity are expressed using triangular membership functions, and the fan start-stop status is expressed using trapezoidal membership functions.

[0138] S63. Construct fuzzy linguistic values ​​of "low", "medium" and "high" for dissolved oxygen concentration and ammonia nitrogen concentration, construct fuzzy linguistic values ​​of "on" and "off" for fan start-up and shutdown status, and construct fuzzy linguistic values ​​of "weak", "medium" and "strong" for operating intensity.

[0139] S64. Based on expert knowledge and operational experience, set fuzzy rules between input and output, and construct a fuzzy rule library. The fuzzy rule library includes at least 27 fuzzy rules for outputting optimized control strategies. The optimized control strategies include increasing the wind turbine operating intensity, decreasing the wind turbine operating intensity, maintaining the current operating state, or triggering an abnormal flag.

[0140] In the fuzzy rule base, some of the fuzzy rules are:

[0141] If the dissolved oxygen concentration is "low", the ammonia nitrogen concentration is "high", the fan start / stop status is "on" and the operating intensity is "weak", the output optimization control strategy is: increase the fan operating intensity;

[0142] If the dissolved oxygen concentration is "medium", the ammonia nitrogen concentration is "high", the fan start / stop status is "on" and the operating intensity is "medium", the output optimization control strategy is: increase the fan operating intensity;

[0143] If the dissolved oxygen concentration is "high", the ammonia nitrogen concentration is "low", the fan start / stop status is "on" and the operating intensity is "strong", the output optimization control strategy is: reduce the fan operating intensity;

[0144] If the dissolved oxygen concentration is "high", the ammonia nitrogen concentration is "medium", the fan start / stop status is "on" and the operating intensity is "medium", the output optimization control strategy is: reduce the fan operating intensity;

[0145] If the dissolved oxygen concentration is "medium", the ammonia nitrogen concentration is "medium", the fan start / stop status is "on" and the operating intensity is "medium", the output optimization control strategy is: maintain the current operating status;

[0146] If the dissolved oxygen concentration is "low", the ammonia nitrogen concentration is "medium", the fan start / stop status is "on", and the operating intensity is "medium", the output optimization control strategy is: maintain the current operating status;

[0147] If the dissolved oxygen concentration is "low", the ammonia nitrogen concentration is "high", the fan start / stop status is "off" and the operating intensity is "strong", the output optimization control strategy is: trigger the abnormal flag;

[0148] If the dissolved oxygen concentration is "high", the ammonia nitrogen concentration is "low", the fan start / stop status is "off" and the operating intensity is "strong", the output optimization control strategy is: trigger the abnormal flag.

[0149] In this invention, the requirement of "covering at least 27 fuzzy rules" is a logical derivation based on the number of input variables and the combination of linguistic values. Since the fuzzy input variables include dissolved oxygen concentration, ammonia nitrogen concentration, fan start / stop status, and operating intensity, which have 3, 3, 2, and 3 fuzzy linguistic values ​​respectively, the total number of possible combinations is 3 × 3 × 2 × 3 = 54. Considering that some input combinations may have logical conflicts, lack control significance, or be redundant in actual operation, it is necessary to select representative rules through expert experience to construct a subset of at least 27 effective fuzzy control rules. This ensures both strategy integrity and improved control efficiency. This rule quantity setting reflects the engineering adaptability and flexible scalability of fuzzy control.

[0150] S65. Based on the fuzzy input variables and the fuzzy rule base, the minimum-maximum composite reasoning method is used to perform composite reasoning on all fuzzy rules to obtain the fuzzy membership degree distribution function. The minimum-maximum composite reasoning specifically means that the fuzzy premise membership degree is minimized and the output conclusion is maximized.

[0151] For example, the input variable values ​​for a structured monitoring data package at a certain moment are set as follows: dissolved oxygen concentration is 3.0 mg / L, ammonia nitrogen concentration is 5.5 mg / L, fan start / stop status is "on", and operating intensity is 1.8 kW.

[0152] Based on the membership function mapping results, the following are obtained: dissolved oxygen concentration corresponds to a membership degree of 0.8 for "low" and 0.2 for "medium"; ammonia nitrogen concentration corresponds to a membership degree of 0.9 for "high" and 0.1 for "medium"; the fan start / stop status is "on", corresponding to a membership degree of 1.0 for "on" and 0 for "off"; operating intensity corresponds to a membership degree of 0.6 for "medium" and 0.4 for "strong". In this case, matching the fuzzy rule base "if the dissolved oxygen concentration is 'low', the ammonia nitrogen concentration is 'high', the fan start / stop status is 'on' and the operating intensity is 'weak', the output optimized control strategy is: increase the fan operating intensity".

[0153] Performing min-max composite inference, the membership degrees of the fuzzy premises are as follows: dissolved oxygen concentration "low": 0.8; ammonia nitrogen concentration "high": 0.9; fan start / stop status "on": 1.0; operating intensity "medium": 0.6. Taking the minimum value among the above premise membership degrees, the activation intensity of the rule is: min(0.8, 0.9, 1.0, 0.6) = 0.6. Therefore, "increase fan operating intensity" is included in the fuzzy membership degree distribution function with a membership degree of 0.6.

[0154] S66. Demodel the fuzzy membership distribution function using the centroid method to obtain and output the corresponding optimized control strategy.

[0155] In this embodiment, step S7 specifically includes:

[0156] S71. The historical data includes: historical operation control commands, historical operation feedback data, wastewater quality parameters, equipment status parameters, and historical optimization control strategies;

[0157] S72. The visualized operating status information includes: generating an operating trend curve by performing time series statistics and moving average processing on historical operating feedback data and historical optimized control strategies; generating an energy consumption analysis chart by performing regression modeling and energy efficiency index calculation on the relationship between historical power consumption values ​​and wind turbine operating intensity; and generating alarm records by filtering and time-marking historical alarm signals and anomaly detection results.

[0158] S73. The visualized operating status information is pushed to the user terminal in real time through APP client, web page or WeChat official account.

[0159] A cloud-based control system for an inverted umbrella-type aeration device includes:

[0160] The data acquisition module is used to collect the status parameters and wastewater quality parameters of the inverted umbrella-type aeration equipment and generate simulated monitoring data sequences.

[0161] The data preprocessing module is used to convert analog monitoring data sequences into digital monitoring signal sequences and perform preprocessing operations to generate standardized datasets.

[0162] The edge control inference module is used to perform edge control inference and generate operation control commands;

[0163] The control execution and feedback module is used to execute operation control commands, collect operation feedback data, detect anomalies, and perform corrective processing.

[0164] The cloud-based data processing module is used to receive operation control commands, operation feedback data, current wastewater quality parameters, and current equipment status parameters, and to construct structured monitoring data packages.

[0165] The fuzzy inference module is used to perform fuzzy rule inference and behavior pattern recognition to generate optimized control strategies.

[0166] The visualization push module is used to generate visual information about the running status and push it to the user terminal in real time.

[0167] Example 1:

[0168] To verify the feasibility of this invention in practice, it was applied to the oxidation ditch process section of a large-scale wastewater treatment plant. This process section is equipped with 12 sets of inverted umbrella-type aeration devices, with a designed treatment capacity of 15,000 m³. 3 / d, mainly treating domestic sewage and some industrial wastewater, the traditional control method uses a PLC timed start-stop mechanism and fixed fan operation intensity setting, which cannot be dynamically adjusted according to water quality changes, resulting in low operating efficiency and frequent manual intervention.

[0169] In implementation, this invention is deployed in the upstream data acquisition module of each aeration device, connecting to dissolved oxygen sensors, ammonia nitrogen sensors, total nitrogen sensors, power monitoring modules, water level sensors, vibration sensors, temperature sensors, and blower status sensors. This allows for real-time acquisition of the aeration device's status parameters and wastewater quality parameters, forming a simulated monitoring data sequence. The analog-to-digital converter in the edge node performs analog-to-digital conversion to generate a digital monitoring signal sequence. The data preprocessing module performs unit conversion, format standardization, dimension unification, missing data completion, outlier amplitude limiting, and noise filtering. A unified timestamp label is added to complete time-series synchronization, generating a standardized dataset.

[0170] The standardized dataset is input into the control inference module deployed at the edge, where trend-cycle-perturbation decomposition and frequency domain attention enhancement processing are performed. The output includes the wind turbine start / stop status, operating intensity, power optimization coefficient, and dissolved oxygen target value. These are then used by the control execution module to generate equipment control commands for real-time adjustment. During equipment command execution, this invention simultaneously collects operational feedback data, identifies and corrects outliers and missing data using a sliding window mechanism. Simultaneously, the operational data, control commands, and feedback information are encrypted with SM4 and uploaded to the cloud platform. Based on the Mamdani fuzzy inference method and historical data, an optimized control strategy is generated and visualized in charts and reports, then pushed to the user's app for remote monitoring and intelligent alarms.

[0171] To verify the actual performance of the present invention, the implementation scheme of the present invention was compared with two comparative schemes: Comparative scheme A is a traditional timed start-stop control scheme, and comparative scheme B is a PLC+DO feedback closed-loop control scheme. The continuous operation cycle was 45 days. During the test, the influent water quality fluctuated significantly, with ammonia nitrogen concentration ranging from 8.3 mg / L to 18.1 mg / L and water temperature ranging from 9.8 to 17.6℃, ensuring that the comparative test environment was representative. The evaluation indicators included energy consumption per unit air volume, effluent dissolved oxygen stability, ammonia nitrogen removal rate, equipment failure rate, frequency of manual intervention, and control response delay. The experimental results are shown in Table 1.

[0172] Table 1 Comparison of Operational Effects under Different Control Schemes

[0173]

[0174]

[0175] As shown in Table 1, the present invention outperforms both comparative scheme A and comparative scheme B in all key performance indicators. Regarding energy consumption per unit air volume, the present invention achieves 0.76 kWh / 1000 Nm³. 3 The lowest energy consumption level is reduced by 37.2% compared to Comparative Solution A and 22.4% compared to Comparative Solution B, significantly improving the energy efficiency of the aeration process. Regarding the stability of dissolved oxygen in the effluent, the standard deviation of this invention is 0.19 mg / L, superior to 0.43 mg / L in Solution A and 0.29 mg / L in Solution B, effectively controlling dissolved oxygen fluctuations and improving the stability of the treatment process. In terms of ammonia nitrogen removal rate, this invention achieves 88.7%, an improvement of 10.8 percentage points compared to Comparative Solution A, demonstrating excellent pollutant removal capabilities even under highly fluctuating influent water quality conditions. Regarding equipment maintenance, this invention reduces the equipment failure rate to 0.5%, far lower than the comparative solutions, and reduces the frequency of manual intervention to 0.4 times / day, effectively reducing the burden on operators. In terms of control response, the response delay of this invention is only 6.3 seconds, shortening it by 11.9 seconds and 5.2 seconds compared to Comparative Solution A and Solution B, respectively, achieving faster strategy closed-loop and adaptive adjustment capabilities.

[0176] This invention constructs an edge-cloud collaborative control architecture that integrates and improves the FEDformer network. This architecture not only enables intelligent and efficient operation control of aeration equipment, but also provides excellent energy efficiency, stability, and maintainability. It is suitable for the actual needs of energy consumption optimization and intelligent operation and maintenance in modern wastewater treatment processes.

[0177] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An inverting-umbrella-type aeration device cloud control method, characterized by, The method comprises the following steps: S1, collecting state parameters and sewage quality parameters of the inverted umbrella type aeration equipment to generate a simulation monitoring data sequence; S2, converting the simulation monitoring data sequence into a digital monitoring signal sequence, and preprocessing the digital monitoring signal sequence to form a standardized data set; S3, using an improved FEDformer network to perform edge control reasoning on the standardized data set to generate an operation control instruction of the inverted umbrella type aeration equipment; S4, executing the operation control instruction and collecting real-time operation feedback data, and performing abnormal correction processing if the operation feedback data is detected to be abnormal; S5, uploading the operation control instruction, the operation feedback data, the current sewage quality parameters and the current equipment state parameters to a cloud processing platform to generate a structured monitoring data package; S6, in the cloud processing platform, performing control rule reasoning and behavior pattern recognition on the structured monitoring data package by using a Mamdani fuzzy reasoning method to generate an optimized control strategy for operation control; S7, classifying the optimized control strategy and historical data to generate operation state visual information and push the information to a user terminal.

2. The cloud control method of an inverted-umbrella type aeration apparatus according to claim 1, characterized by, The step S1 specifically comprises: S11, deploying dissolved oxygen sensors, ammonia nitrogen sensors, total nitrogen sensors, electric energy monitoring modules, water level sensors, vibration sensors, temperature sensors and fan state sensors on the inverted umbrella type aeration equipment to synchronously collect state parameters and sewage quality parameters of the inverted umbrella type aeration equipment; S12, the state parameters include fan start-stop state, electric energy consumption value, shell vibration amplitude and equipment surface temperature, and the sewage quality parameters include dissolved oxygen concentration, ammonia nitrogen concentration, total nitrogen concentration and water level value; S13, removing mutation type abnormal values of the state parameters and the sewage quality parameters, correcting baseline drift data and supplementing missing data, and performing time sequence alignment based on a unified time stamp to generate a simulation monitoring data sequence.

3. The cloud control method of an inverted-umbrella type aeration apparatus according to claim 1, characterized by, The step S2 specifically comprises: S21, performing an analog-digital conversion operation on the simulation monitoring data sequence by using an analog-digital converter to generate a corresponding digital monitoring signal sequence; S22, the preprocessing includes: performing unit conversion, format regularization and dimension unification on the digital monitoring signal sequence, and performing missing data completion, abnormal value limiting processing and noise filtering; S23, adding a unified time stamp label to the preprocessed digital monitoring signal sequence and performing time sequence synchronization to form a standardized data set, wherein the standardized data set includes numerical variables and categorical variables.

4. The inverted-umbrella type aeration device cloud control method according to claim 1, characterized by, The step S3 specifically comprises: S31, the improved FEDformer network includes a fusion encoding module, a trend-period-disturbance decomposition module, a frequency domain attention block, a control instruction decoding module and a feedback injection module; S32, the fusion encoding module performs feature encoding operation on the standardized data set, specifically: performing linear projection on the numerical variables through a fully connected layer to map them to an embedded space with unified dimensions to generate a numerical embedded tensor, performing one-hot encoding on the categorical variables, and mapping the vectors through a pre-set embedded lookup matrix to generate a categorical embedded tensor; based on the position index of each time step, constructing a position encoding tensor through a sine function and a cosine function; The numerical embedding tensor, the category embedding tensor and the position coding tensor are spliced in the feature dimension to construct a fusion feature tensor; S33, the trend-period-disturbance decomposition module performs multi-scale decomposition processing on the fusion feature tensor, specifically: The fusion feature tensor is sequentially sliced based on a fixed-length sliding time window to obtain a plurality of local subsequence blocks; For each local subsequence block, a one-dimensional convolution is performed along the time axis direction to extract a smooth change trend component to generate a trend component tensor, a main periodic pattern is identified based on a band-pass filter and a period detection kernel to extract a periodic change structure to generate a period component tensor, and the local subsequence block is subtracted by the corresponding trend component tensor and period component tensor to obtain a disturbance residual component tensor; The trend component tensor, the period component tensor and the disturbance residual tensor are spliced in the feature dimension to generate an enhanced time series decomposition tensor; S34, the frequency domain attention block performs a fast Fourier transform on the enhanced time series decomposition tensor to obtain a frequency domain mapping tensor; The modulus values of each frequency component in the frequency domain mapping tensor are extracted and sorted in descending order according to the modulus values, the first K frequency indexes with the largest modulus values are selected to form a main frequency component set, and the corresponding frequency components are extracted from the frequency domain mapping tensor according to the main frequency component set to form a main frequency tensor; The main frequency tensor is subjected to amplitude normalization processing to generate a frequency domain attention matrix through a Softmax function, the main frequency tensor and the frequency domain attention matrix are combined by element-wise weighting to generate a frequency domain enhanced tensor, and the frequency domain enhanced tensor is subjected to an inverse fast Fourier transform operation to generate a time series reconstruction tensor; S35, the control instruction decoding module includes four parallel control prediction branches, a boundary constraint transformer and an instruction encoding unit; The four parallel control prediction branches respectively receive the time series reconstruction tensor and perform linear compression, nonlinear mapping and activation output in their respective structures to respectively generate a fan start-stop control quantity, a running intensity control quantity, an electric energy optimization coefficient and a target dissolved oxygen set value; The boundary constraint transformer performs threshold judgment and binary operation on the fan start-stop control quantity, and when the fan start-stop control quantity exceeds a set start threshold, outputs a fan start-stop control flag as 1: start signal, otherwise outputs the control flag as 0: stop signal; The boundary constraint transformer performs reverse normalization mapping, upper and lower limit amplitude limiting and numerical format conversion operations on the running intensity control quantity to generate a running intensity control instruction, performs range mapping on the electric energy optimization coefficient to generate an electric energy optimization control instruction, and performs linear limiting operation on the target dissolved oxygen set value to generate a dissolved oxygen set control instruction; The instruction encoding unit performs field identification, splicing and communication protocol packaging on the fan start-stop control flag, the running intensity control instruction, the electric energy optimization control instruction and the dissolved oxygen set control instruction to form a running control instruction.

5. The inverted-umbrella type aeration device cloud control method according to claim 1, characterized by, The step S4 specifically includes: S41, the running control instruction is issued to the inverted umbrella type aeration equipment for execution control; S42, collecting the operation feedback data of the inverted-umbrella aeration equipment in real time through the feedback injection module of the improved FEDformer network, the operation feedback data including fan start-stop state, operation intensity, power consumption value and dissolved oxygen concentration; S43, when there are mutation abnormal values, continuous missing values or physical inconsistencies in the operation feedback data, performing abnormal correction processing: based on a fixed-length sliding time window, dynamically calculating the mean and standard deviation of each type of operation feedback data, if there is operation feedback data exceeding the range of mean ± 3 times standard deviation, it is marked as a mutation abnormal value, and the missing points or abnormal values are reconstructed by using time series interpolation method; if there are continuous missing values exceeding the set threshold window length, the data is completed by using interpolation method; if there is a logical conflict between the operation feedback data and the physical response of the operation control instruction, it is marked as physical inconsistency, and an alarm signal is generated and reported to the cloud platform; S44, pre-processing the operation feedback data after abnormal correction, and splicing to the end of the standardized data set, and continuing to input to the improved FEDformer network for edge control reasoning.

6. The inverted-umbrella type aeration device cloud control method according to claim 1, characterized by, The step S5 specifically includes: S51, calling the communication interface to package the operation control instruction, the operation feedback data, the current sewage water quality parameter and the current equipment state parameter to form a structured monitoring data packet; S52, encrypting the structured monitoring data packet using the SM4 symmetric encryption algorithm, and sending it to the cloud processing platform through the wireless communication network; S53, performing field-level analysis and classified storage on the structured monitoring data packet in the cloud processing platform, and constructing a mapping table of the operation control instruction field, the operation feedback field, the sewage water quality parameter field and the equipment state parameter field.

7. The inverted-umbrella type aeration device cloud control method according to claim 1, characterized by, The step S6 specifically includes: S61, constructing fuzzy input variables, the fuzzy input variables including dissolved oxygen concentration, ammonia nitrogen concentration, fan start-stop state and operation intensity; S62, the dissolved oxygen concentration, ammonia nitrogen concentration and operation intensity adopt triangular membership functions, and the fan start-stop state adopts trapezoidal membership function; S63, constructing "low", "medium", "high" fuzzy language values for dissolved oxygen concentration and ammonia nitrogen concentration, constructing "on" and "off" fuzzy language values for fan start-stop state, and constructing "weak", "medium", "strong" fuzzy language values for operation intensity; S64, setting the fuzzy rules between input and output based on expert knowledge and operation experience, constructing a fuzzy rule base, the fuzzy rule base including at least 27 fuzzy rules for outputting an optimized control strategy, the optimized control strategy including improving fan operation intensity, reducing fan operation intensity, maintaining current operation state or triggering abnormal flag; S65, synthesizing all fuzzy rules according to the fuzzy input variables and the fuzzy rule base by using the minimum-maximum synthesis reasoning method to obtain a fuzzy membership distribution function, the minimum-maximum synthesis reasoning specifically being that the fuzzy premise membership degree takes the minimum and the output conclusion synthesis takes the maximum; S66, performing de-fuzzing operation on the fuzzy membership distribution function by using the barycenter method to obtain and output the corresponding optimized control strategy.

8. The inverted-umbrella type aeration device cloud control method according to claim 1, characterized by, The step S7 specifically includes: S71, the historical data includes: historical operation control instructions, historical operation feedback data, sewage water quality parameters, equipment state parameters and historical optimization control strategies; S72, the operation state visualization information includes operation trend curve, energy consumption analysis chart and alarm record; S73, the operation state visualization information is pushed to the user terminal in real time through APP client, web page or WeChat public number.

9. An inverting-umbrella-shaped aeration device cloud control system, which executes the inverting-umbrella-shaped aeration device cloud control method according to any one of claims 1 to 8, characterized by, Comprise: A data acquisition module for acquiring state parameters and sewage water quality parameters of the inverted-umbrella-shaped aeration equipment and generating a simulated monitoring data sequence; A data preprocessing module for converting the simulated monitoring data sequence into a digital monitoring signal sequence and performing a preprocessing operation to generate a standardized data set; An edge control reasoning module for performing edge control reasoning to generate operation control instructions; A control execution and feedback module for executing operation control instructions, collecting operation feedback data, detecting abnormalities and performing correction processing; A cloud data processing module for receiving operation control instructions, operation feedback data, current sewage water quality parameters and current equipment state parameters, and constructing a structured monitoring data package; A fuzzy reasoning module for fuzzy rule reasoning and behavior pattern recognition to generate an optimization control strategy; A visualization pushing module for generating operation state visualization information and pushing to the user terminal in real time.

Images