Purifier turbulence fitting system, evaluation system and control system

Abstract

The present invention relates to a purifier turbulence fitting system, an intelligent state assessment system, and an intelligent control system based on an adversarial neural network. The intelligent control system includes a cloud-based purifier turbulence vector pre-training model, a purifier multi-point flow velocity sensor, a neural network turbulence vector correction module, a turbulence vector pre-processing model, a neural network turbulence vector single time slice generation module, a discriminative neural network turbulence vector single time slice assessment module, a neural network turbulence vector multi-time slice generation module, a discriminative neural network turbulence vector multi-time slice assessment module, a neural network turbulence vector fitting flow direction module, a neural network purifier operation status assessment module, and an intelligent control system. The present invention uses joint detection of multiple detection points and the purifier pre-training model to fit and iteratively correct the purifier turbulence output, thereby achieving accurate control of the purifier.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control of purifiers, and in particular to a turbulence fitting system, an intelligent state evaluation system and an intelligent control system for a purifier based on an adversarial neural network. Background Art

[0002] Traditional methods for purifying public space air involve reading fixed-point values ​​and controlling the purifier to purify at a fixed angle and air volume when any value exceeds the standard. More advanced methods also employ intelligent control to automatically adjust purifier parameters. However, these previous intelligent control methods failed to fit the critical output turbulence model, effectively fitting the purifier's actual purification performance and providing effective intelligent control. Furthermore, these methods were unable to determine the current performance of the device through turbulence fitting, prompting maintenance or replacement. Summary of the Invention

[0003] In response to the problems existing in the prior art, the purpose of the present invention is to provide a purifier turbulence fitting system, an intelligent state assessment system and an intelligent control system based on an adversarial neural network to achieve more effective and accurate control of the purifier.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] A turbulence fitting system for purifier based on adversarial neural network, which includes

[0006] The turbulence vector pre-training model of the purifier is used to simulate and output the turbulence vector model based on the operating conditions of the purifier before leaving the factory;

[0007] A neighborhood air quality monitor cluster, consisting of multiple air quality monitors located at different points around the purifier, collects air pollutant concentrations, each with its own location information.

[0008] The purifier multi-point flow rate sensor is composed of multiple flow rate sensors set at different positions of the purifier, which is used to collect the real-time air flow rate at the position point;

[0009] The neural network turbulence vector correction module connects to the cloud-based purifier turbulence vector pre-trained model, the field air quality monitor cluster, and the purifier's multi-point flow rate sensor. It is used to input the turbulence vector, the air pollutant concentration with location information, and the real-time air flow rate at multiple locations. The neural network turbulence vector correction module corrects the input turbulence vector model based on the input air pollutant concentration and real-time air flow rate to obtain the turbulence vector pre-processing model.

[0010] A neural network turbulence vector single time slice generation module is generated, connected to a cluster of neighborhood air quality monitors and a multi-point flow velocity sensor of a purifier. The neural network turbulence vector single time slice generation module is used to input air pollutant concentrations with location information and real-time air flow velocities at multiple locations, and based on the input information, generates multiple single-time slice turbulence vector pre-fitting vectors;

[0011] The discriminant neural network turbulence vector single time slice evaluation module is connected to the turbulence vector preprocessing model and generates the neural network turbulence vector single time slice generation module. It is used to input the turbulence vector preprocessing model and multiple single time slice turbulence vector pre-fitting vectors, and score each turbulence vector pre-fitting vector through the neural network. The highest-scoring turbulence vector is selected as the single time slice turbulence vector and output;

[0012] Generate a neural network turbulence vector multi-time slice generation module, connect it to the discriminant neural network turbulence vector single time slice evaluation module, input the turbulence vector of the current single time slice, and generate a neural network to generate multiple multi-time slice turbulence vector pre-fitting vectors based on the turbulence vector of the current single time slice and the turbulence vectors of multiple previous time slices;

[0013] The discriminant neural network turbulence vector multi-time slice evaluation module is used to input a turbulence vector pre-processing model, multiple multi-time slice turbulence vector pre-fitting vectors, and multiple single-time slice turbulence vector scores corresponding to the multi-time slice turbulence vector pre-fitting vectors. Each multi-time slice turbulence vector pre-fitting vector is scored using the trained neural network, and the highest-scoring one is selected as the multi-time slice corrected turbulence vector for single time slice and output.

[0014] The neural network turbulence vector fitting flow direction module is connected to the discriminant neural network turbulence vector multi-time slice evaluation module, which is used to input the multi-time slice corrected turbulence vector. The neural network corrects the turbulence vector according to the multi-time slice and outputs the neighborhood turbulence vector map.

[0015] The fitting system further includes

[0016] A real-time training module for generating a neural network turbulence vector single time slice generation module is connected to the generating neural network turbulence vector single time slice generation module and the discriminative neural network turbulence vector single time slice evaluation module. The module inputs the current network weight of the generating neural network turbulence vector single time slice generation module at every interval, and the evaluation score of the turbulence vector fitting vector of the corresponding time slice given by all the discriminative neural network turbulence vector single time slice evaluation modules within this time interval, so as to train the network weight of the generating neural network turbulence vector single time slice generation module and overwrite the original network weight.

[0017] The fitting system further includes

[0018] Generate a real-time training module for the neural network turbulence vector multi-time slice generation module, connect it to the discriminant neural network turbulence vector multi-time slice evaluation module, input the current network weight of the generated neural network turbulence vector multi-time slice generation module at every interval, and train the network weight of the generated neural network turbulence vector multi-time slice generation module based on the evaluation score of the corresponding multi-time slice turbulence vector fitting vector given by all the discriminant neural network turbulence vector multi-time slice evaluation modules within this time interval, and overwrite the original network weight.

[0019] The intelligent state evaluation system of purifier based on adversarial neural network includes

[0020] The turbulence vector pre-training model of the purifier is used to simulate and output the turbulence vector model based on the operating conditions of the purifier before leaving the factory;

[0021] A neighborhood air quality monitor cluster, consisting of multiple air quality monitors located at different points around the purifier, collects air pollutant concentrations, each with its own location information.

[0022] The purifier multi-point flow rate sensor is composed of multiple flow rate sensors set at different positions of the purifier, which is used to collect the real-time air flow rate at the position point;

[0023] The neural network turbulence vector correction module connects to the cloud-based purifier turbulence vector pre-trained model, the field air quality monitor cluster, and the purifier's multi-point flow rate sensor. It is used to input the turbulence vector, the air pollutant concentration with location information, and the real-time air flow rate at multiple locations. The neural network turbulence vector correction module corrects the input turbulence vector model based on the input air pollutant concentration and real-time air flow rate to obtain the turbulence vector pre-processing model.

[0024] A neural network turbulence vector single time slice generation module is generated, connected to a cluster of neighborhood air quality monitors and a multi-point flow velocity sensor of a purifier. The neural network turbulence vector single time slice generation module is used to input air pollutant concentrations with location information and real-time air flow velocities at multiple locations, and based on the input information, generates multiple single-time slice turbulence vector pre-fitting vectors;

[0025] The discriminant neural network turbulence vector single time slice evaluation module is connected to the turbulence vector preprocessing model and generates the neural network turbulence vector single time slice generation module. It is used to input the turbulence vector preprocessing model and multiple single time slice turbulence vector pre-fitting vectors, and score each turbulence vector pre-fitting vector through the neural network. The highest-scoring turbulence vector is selected as the single time slice turbulence vector and output;

[0026] Generate a neural network turbulence vector multi-time slice generation module, connect it to the discriminant neural network turbulence vector single time slice evaluation module, input the turbulence vector of the current single time slice, and generate a neural network to generate multiple multi-time slice turbulence vector pre-fitting vectors based on the turbulence vector of the current single time slice and the turbulence vectors of multiple previous time slices;

[0027] The discriminant neural network turbulence vector multi-time slice evaluation module is used to input a turbulence vector pre-processing model, multiple multi-time slice turbulence vector pre-fitting vectors, and multiple single-time slice turbulence vector scores corresponding to the multi-time slice turbulence vector pre-fitting vectors. Each multi-time slice turbulence vector pre-fitting vector is scored using the trained neural network, and the highest-scoring one is selected as the multi-time slice corrected turbulence vector for single time slice and output.

[0028] The neural network purifier operation status evaluation module is connected to the discriminant neural network turbulence vector multi-time slice evaluation module, which is used to input the multi-time slice corrected turbulence vector. The neural network corrects the turbulence vector based on the multi-time slice and outputs the purifier operation status evaluation value.

[0029] The state assessment system further includes

[0030] A real-time training module for generating a neural network turbulence vector single time slice generation module is connected to the generating neural network turbulence vector single time slice generation module and the discriminative neural network turbulence vector single time slice evaluation module. The module inputs the current network weight of the generating neural network turbulence vector single time slice generation module at every interval, and the evaluation score of the turbulence vector fitting vector of the corresponding time slice given by all the discriminative neural network turbulence vector single time slice evaluation modules within this time interval, so as to train the network weight of the generating neural network turbulence vector single time slice generation module and overwrite the original network weight.

[0031] The state assessment system further includes

[0032] Generate a real-time training module for the neural network turbulence vector multi-time slice generation module, connect it to the discriminant neural network turbulence vector multi-time slice evaluation module, input the current network weight of the generated neural network turbulence vector multi-time slice generation module at every interval, and train the network weight of the generated neural network turbulence vector multi-time slice generation module based on the evaluation score of the corresponding multi-time slice turbulence vector fitting vector given by all the discriminant neural network turbulence vector multi-time slice evaluation modules within this time interval, and overwrite the original network weight.

[0033] A purifier intelligent control system based on an adversarial neural network, comprising the turbulence fitting system and the intelligent state evaluation system as described above;

[0034] The intelligent control system also includes an intelligent control module, which is connected to the neural network purifier operation status evaluation module and the neural network turbulence vector flow direction module, and is used to input the purifier operation status evaluation value and the neighborhood turbulence vector diagram, and adjust the purifier operation mode according to the purifier operation status evaluation value and the neighborhood turbulence vector diagram.

[0035] After adopting the above scheme, the present invention uses the joint detection of multiple detection points and the purifier pre-training model to fit and iteratively correct the turbulence output of the purifier, thereby achieving accurate control of the purifier. The present invention not only takes into account the turbulence vector of a single time slice, but also considers other single time slices before and after the single time slice on this basis, so as to form turbulence vectors of multiple time slices, so that the formed turbulence vector is more accurate, and thus the purifier can be accurately controlled. Specifically, the present invention uses the concept of a single time slice to fit the turbulence vector considered by the neural network, and then uses the concept of multiple time slices to select from these generated vectors the situation that the turbulence vector changes with time. In simple terms, the turbulence vector is actually generated by a single time slice, and multiple time slices are a selection process, and then the result of this selection is compared with the sensor data finally obtained to obtain an accurate turbulence vector. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a principle block diagram of the present invention. DETAILED DESCRIPTION

[0037] like Figure 1 As shown, the present invention discloses an intelligent control system for a purifier based on an adversarial neural network, which includes:

[0038] The purifier turbulence vector pre-training model is used to simulate and output a turbulence vector model based on the purifier's pre-factory operating conditions. This purifier turbulence vector pre-training model is currently hosted in the cloud but can also be hosted locally.

[0039] The neighborhood air quality monitor cluster consists of multiple air quality monitors set at different locations of the purifier. It is used to collect air pollutant concentrations, and each air pollutant concentration is accompanied by location information.

[0040] The purifier multi-point flow rate sensor consists of multiple flow rate sensors set at different positions of the purifier, which is used to collect the real-time air flow rate at the position points.

[0041] The neural network turbulence vector correction module is connected to the cloud purifier turbulence vector pre-training model, the field air quality monitor cluster, and the purifier multi-point flow rate sensor. It is used to input the turbulence vector, the air pollutant concentration with location information, and the real-time air flow rate at multiple locations; the neural network turbulence vector correction module corrects the input turbulence vector model according to the input air pollutant concentration and real-time air flow rate to obtain a turbulence vector preprocessing model.

[0042] Generate a neural network turbulence vector single time slice generation module, connect to the neighborhood air quality monitor cluster and the purifier multi-point flow rate sensor, and use it to input the air pollutant concentration with location information and the real-time air flow rate at multiple locations. Based on the input information, the neural network generates multiple single time slice turbulence vector pre-fitting vectors.

[0043] The discriminant neural network turbulence vector single time slice evaluation module is connected to the turbulence vector preprocessing model and generates the neural network turbulence vector single time slice generation module, which is used to input the turbulence vector preprocessing model and multiple single time slice turbulence vector pre-fitting vectors, and score each turbulence vector pre-fitting vector through the neural network. The one with the highest score is selected as the turbulence vector of the single time slice and output.

[0044] A real-time training module for generating a neural network turbulence vector single time slice generation module is connected to the generating neural network turbulence vector single time slice generation module and the discriminative neural network turbulence vector single time slice evaluation module. The module inputs the current network weight of the generating neural network turbulence vector single time slice generation module at every interval, and the evaluation score of the turbulence vector fitting vector of the corresponding time slice given by all the discriminative neural network turbulence vector single time slice evaluation modules within this time interval, so as to train the network weight of the generating neural network turbulence vector single time slice generation module and overwrite the original network weight.

[0045] Generate a neural network turbulence vector multi-time slice generation module, connect it to the discriminant neural network turbulence vector single time slice evaluation module, input the turbulence vector of the current time slice, and generate a neural network based on the turbulence vector of the current time slice and the turbulence vectors of multiple previous time slices to generate multiple multi-time slice turbulence vector pre-fitting vectors.

[0046] The discriminant neural network turbulence vector multi-time slice evaluation module is characterized by inputting a cloud-based turbulence vector preprocessing model, multiple multi-time slice turbulence vector pre-fitting vectors, and multiple single-time slice turbulence vector scores corresponding to the multi-time slice turbulence vector pre-fitting vectors. Each multi-time slice turbulence vector pre-fitting vector is scored through a trained neural network, and the one with the highest score is selected as the multi-time slice corrected turbulence vector for a single time slice and output.

[0047] Generate a real-time training module for the neural network turbulence vector multi-time slice generation module, connect it to the discriminant neural network turbulence vector multi-time slice evaluation module, input the current network weight of the generated neural network turbulence vector multi-time slice generation module at every interval, and train the network weight of the generated neural network turbulence vector multi-time slice generation module based on the evaluation score of the corresponding multi-time slice turbulence vector fitting vector given by all the discriminant neural network turbulence vector multi-time slice evaluation modules within this time interval, and overwrite the original network weight.

[0048] The neural network turbulence vector fitting flow direction module is connected to the discriminant neural network turbulence vector multi-time slice evaluation module, which is used to input the multi-time slice corrected turbulence vector. The neural network corrects the turbulence vector according to the multi-time slice and outputs the neighborhood turbulence vector map.

[0049] The neural network purifier health assessment module, connected to the discriminant neural network turbulence vector multi-time slice assessment module, takes as input the multi-time slice corrected turbulence vector. The neural network uses this multi-time slice corrected turbulence vector to output a purifier health assessment. The neural network purifier health assessment module compares the multi-time slice corrected turbulence vector with historical multi-time vectors to determine if there have been significant changes in the specific output under similar conditions, as significant changes generally indicate problems.

[0050] The intelligent control module is connected to the neural network purifier operation status evaluation module and the neural network turbulence vector flow direction module, and is used to input the purifier operation status evaluation value and the neighborhood turbulence vector diagram, and adjust the purifier operation mode according to the purifier operation status evaluation value and the neighborhood turbulence vector diagram.

[0051] The above-mentioned cloud purifier turbulence vector pre-training model, purifier multi-point flow velocity sensor, neural network turbulence vector correction module, turbulence vector preprocessing model, generation neural network turbulence vector single time slice generation module, discrimination neural network turbulence vector single time slice evaluation module, generation neural network turbulence vector single time slice generation module real-time training module, generation neural network turbulence vector multi-time slice generation module, discrimination neural network turbulence vector multi-time slice evaluation module, generation neural network turbulence vector multi-time slice generation module real-time training module, neural network turbulence vector fitting flow direction module can constitute a purifier turbulence fitting system based on adversarial neural network, which is used to obtain the neighborhood turbulence vector diagram of the purifier.

[0052] The above-mentioned cloud purifier turbulence vector pre-training model, purifier multi-point flow velocity sensor, neural network turbulence vector correction module, turbulence vector preprocessing model, generation neural network turbulence vector single time slice generation module, discrimination neural network turbulence vector single time slice evaluation module, generation neural network turbulence vector single time slice generation module real-time training module, generation neural network turbulence vector multiple time slice generation module, discrimination neural network turbulence vector multiple time slice evaluation module, generation neural network turbulence vector multiple time slice generation module real-time training module, neural network purifier operation status evaluation module can constitute a purifier intelligent state evaluation system based on adversarial neural network, which is used to obtain the purifier operation status evaluation value.

[0053] The above description is merely an embodiment of the present invention and does not limit the technical scope of the present invention. Therefore, any minor modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A turbulent flow fitting system for purifiers based on adversarial neural networks, characterized by: The system includes The turbulence vector pre-training model of the purifier is used to simulate and output the turbulence vector model based on the operating conditions of the purifier before leaving the factory; A neighborhood air quality monitor cluster, consisting of multiple air quality monitors located at different points around the purifier, collects air pollutant concentrations, each with its own location information. The purifier multi-point flow rate sensor is composed of multiple flow rate sensors set at different positions of the purifier, which is used to collect the real-time air flow rate at the position point; The neural network turbulence vector correction module connects to the cloud-based purifier turbulence vector pre-trained model, the field air quality monitor cluster, and the purifier's multi-point flow rate sensor. It is used to input the turbulence vector, the air pollutant concentration with location information, and the real-time air flow rate at multiple locations. The neural network turbulence vector correction module corrects the input turbulence vector model based on the input air pollutant concentration and real-time air flow rate to obtain the turbulence vector pre-processing model. A neural network turbulence vector single time slice generation module is generated, connected to a cluster of neighborhood air quality monitors and a multi-point flow velocity sensor of a purifier. The neural network turbulence vector single time slice generation module is used to input air pollutant concentrations with location information and real-time air flow velocities at multiple locations, and based on the input information, generates multiple single-time slice turbulence vector pre-fitting vectors; The discriminant neural network turbulence vector single time slice evaluation module is connected to the turbulence vector preprocessing model and generates the neural network turbulence vector single time slice generation module. It is used to input the turbulence vector preprocessing model and multiple single time slice turbulence vector pre-fitting vectors, and score each turbulence vector pre-fitting vector through the neural network. The highest-scoring turbulence vector is selected as the single time slice turbulence vector and output; Generate a neural network turbulence vector multi-time slice generation module, connect it to the discriminant neural network turbulence vector single time slice evaluation module, input the turbulence vector of the current single time slice, and generate a neural network to generate multiple multi-time slice turbulence vector pre-fitting vectors based on the turbulence vector of the current single time slice and the turbulence vectors of multiple previous time slices; The discriminant neural network turbulence vector multi-time slice evaluation module is used to input a turbulence vector pre-processing model, multiple multi-time slice turbulence vector pre-fitting vectors, and multiple single-time slice turbulence vector scores corresponding to the multi-time slice turbulence vector pre-fitting vectors. Each multi-time slice turbulence vector pre-fitting vector is scored using the trained neural network, and the highest-scoring one is selected as the multi-time slice corrected turbulence vector for single time slice and output. The neural network turbulence vector fitting flow direction module is connected to the discriminant neural network turbulence vector multi-time slice evaluation module, which is used to input the multi-time slice corrected turbulence vector. The neural network corrects the turbulence vector according to the multi-time slice and outputs the neighborhood turbulence vector map.

2. The turbulence fitting system for purifiers based on adversarial neural networks according to claim 1, characterized in that: The fitting system further includes A real-time training module for generating a neural network turbulence vector single time slice generation module is connected to the generating neural network turbulence vector single time slice generation module and the discriminative neural network turbulence vector single time slice evaluation module. The module inputs the current network weight of the generating neural network turbulence vector single time slice generation module at every interval, and the evaluation score of the turbulence vector fitting vector of the corresponding time slice given by all the discriminative neural network turbulence vector single time slice evaluation modules within this time interval, so as to train the network weight of the generating neural network turbulence vector single time slice generation module and overwrite the original network weight.

3. The turbulence fitting system for purifiers based on adversarial neural networks according to claim 1, characterized in that: The fitting system further includes Generate a real-time training module for the neural network turbulence vector multi-time slice generation module, connect it to the discriminant neural network turbulence vector multi-time slice evaluation module, input the current network weight of the generated neural network turbulence vector multi-time slice generation module at every interval, and train the network weight of the generated neural network turbulence vector multi-time slice generation module based on the evaluation score of the corresponding multi-time slice turbulence vector fitting vector given by all the discriminant neural network turbulence vector multi-time slice evaluation modules within this time interval, and overwrite the original network weight.

4. The intelligent status assessment system for purifiers based on adversarial neural networks is characterized by: The system includes The turbulence vector pre-training model of the purifier is used to simulate and output the turbulence vector model based on the operating conditions of the purifier before leaving the factory; A neighborhood air quality monitor cluster, consisting of multiple air quality monitors located at different points around the purifier, collects air pollutant concentrations, each with its own location information. The purifier multi-point flow rate sensor is composed of multiple flow rate sensors set at different positions of the purifier, which is used to collect the real-time air flow rate at the position point; The neural network turbulence vector correction module connects to the cloud-based purifier turbulence vector pre-trained model, the field air quality monitor cluster, and the purifier's multi-point flow rate sensor. It is used to input the turbulence vector, the air pollutant concentration with location information, and the real-time air flow rate at multiple locations. The neural network turbulence vector correction module corrects the input turbulence vector model based on the input air pollutant concentration and real-time air flow rate to obtain the turbulence vector pre-processing model. A neural network turbulence vector single time slice generation module is generated, connected to a cluster of neighborhood air quality monitors and a multi-point flow velocity sensor of a purifier. The neural network turbulence vector single time slice generation module is used to input air pollutant concentrations with location information and real-time air flow velocities at multiple locations, and based on the input information, generates multiple single-time slice turbulence vector pre-fitting vectors; The discriminant neural network turbulence vector single time slice evaluation module is connected to the turbulence vector preprocessing model and generates the neural network turbulence vector single time slice generation module. It is used to input the turbulence vector preprocessing model and multiple single time slice turbulence vector pre-fitting vectors, and score each turbulence vector pre-fitting vector through the neural network. The highest-scoring turbulence vector is selected as the single time slice turbulence vector and output; Generate a neural network turbulence vector multi-time slice generation module, connect it to the discriminant neural network turbulence vector single time slice evaluation module, input the turbulence vector of the current single time slice, and generate a neural network to generate multiple multi-time slice turbulence vector pre-fitting vectors based on the turbulence vector of the current single time slice and the turbulence vectors of multiple previous time slices; The discriminant neural network turbulence vector multi-time slice evaluation module is used to input a turbulence vector pre-processing model, multiple multi-time slice turbulence vector pre-fitting vectors, and multiple single-time slice turbulence vector scores corresponding to the multi-time slice turbulence vector pre-fitting vectors. Each multi-time slice turbulence vector pre-fitting vector is scored using the trained neural network, and the highest-scoring one is selected as the multi-time slice corrected turbulence vector for single time slice and output. The neural network purifier operation status evaluation module is connected to the discriminant neural network turbulence vector multi-time slice evaluation module, which is used to input the multi-time slice corrected turbulence vector. The neural network corrects the turbulence vector based on the multi-time slice and outputs the purifier operation status evaluation value.

5. The purifier intelligent status assessment system based on adversarial neural network according to claim 4 is characterized in that: The state assessment system also includes A real-time training module for generating a neural network turbulence vector single time slice generation module is connected to the generating neural network turbulence vector single time slice generation module and the discriminative neural network turbulence vector single time slice evaluation module. The module inputs the current network weight of the generating neural network turbulence vector single time slice generation module at every interval, and the evaluation score of the turbulence vector fitting vector of the corresponding time slice given by all the discriminative neural network turbulence vector single time slice evaluation modules within this time interval, so as to train the network weight of the generating neural network turbulence vector single time slice generation module and overwrite the original network weight.

6. The purifier intelligent status assessment system based on adversarial neural network according to claim 4 is characterized in that: The state assessment system also includes Generate a real-time training module for the neural network turbulence vector multi-time slice generation module, connect it to the discriminant neural network turbulence vector multi-time slice evaluation module, input the current network weight of the generated neural network turbulence vector multi-time slice generation module at every interval, and train the network weight of the generated neural network turbulence vector multi-time slice generation module based on the evaluation score of the corresponding multi-time slice turbulence vector fitting vector given by all the discriminant neural network turbulence vector multi-time slice evaluation modules within this time interval, and overwrite the original network weight.

7. The intelligent control system of the purifier based on the adversarial neural network is characterized by: The system comprises the turbulence fitting system according to any one of claims 1 to 3 and the intelligent state assessment system according to any one of claims 4 to 6; The intelligent control system also includes an intelligent control module, which is connected to the neural network purifier operation status evaluation module and the neural network turbulence vector flow direction module, and is used to input the purifier operation status evaluation value and the neighborhood turbulence vector diagram, and adjust the purifier operation mode according to the purifier operation status evaluation value and the neighborhood turbulence vector diagram.

Images