34 results about "Artificial neural network algorithm" patented technology

The invention discloses a saline-alkali soil improvement control analysis method and system based on digital twinning, and belongs to the technical field of saline-alkali soil improvement control analysis, and the method comprises the steps: firstly obtaining and preprocessing the terrain, soil permeability and precipitation data of saline-alkali soil, predicting a retention water accumulation region through employing a support vector machine algorithm, and estimating the amount of retained water by using a Kriging interpolation method. Then, the dynamic change process of the retained water is simulated through an artificial neural network algorithm, the information is integrated into a digital twinborn model, and visualization of spatial distribution and dynamic change of the retained water is achieved. And finally, on the basis of retained water information in the digital twin model, optimizing the layout and depth of the drainage ditch, selecting proper drainage materials and modes, and forming a complete drainage design scheme. According to the method, accurate prediction, dynamic simulation and visual analysis of the retention water of the saline-alkali soil are realized through a method of combining data driving and an intelligent algorithm, and a scientific basis and decision support are provided for improvement of the saline-alkali soil and agricultural production.

The invention provides a PCELL layout design optimization method and device, a medium and equipment, and is applied to the technical field of chip design. Flexible layout drawing is carried out by acquiring the PCELL containing the process parameters and the graph adjustable parameters, so that the layout structure parameters of the current PCELL are obtained. A three-dimensional device simulation model is constructed by using the parameters, and a simulation compact model is created by using an artificial neural network algorithm to describe the electrical behavior of the device. And verifying the electrical performance trend of the three-dimensional device simulation model by using the PCELL and the simulation compact model. And when the verification is passed, further verifying whether the module test circuit meets a preset PPA index. And if yes, fixing the structural parameters of the current PCELL layout. According to the method, process and design requirements are integrated through a construction and parameterization method, and simulation and verification of closed-loop feedback are utilized, so that the design precision and the process maturity are improved, the development period is shortened, and the reliability and efficiency of the process are enhanced.

The invention, which belongs to the technical field of the power distribution network, discloses an optimization method for a hybrid energy storage system of a power distribution network, comprising the following steps: step 1, constructing a flexible grid-connected model of a photovoltaic-hybrid energy storage system; 2, constructing a double-layer random model for improving the reliability of the power distribution network; 3, setting an optimized operation process of the power distribution network; and 4, setting a control mode. According to the method, the economic advantages of the HESS are developed to the greatest extent through coordinated control of the energy storage units of the hybrid energy storage system, the charge and discharge power of the HESS is controlled through an artificial neural network algorithm, the deviation between the wind power prediction power and the actual power is compensated, the state of charge (SOC) of the HESS is considered, overcharge or deep discharge of the ESS is caused, and the service life is prolonged.

The invention relates to the technical field of high-voltage cable operation inspection, and provides a high-voltage cable operation inspection method and system based on online monitoring and intelligent inspection. An online monitoring system based on an artificial neural network algorithm is used for diagnosing the cable operation fault probability, and when the cable operation fault probability does not exceed a threshold value, the cable operation fault probability is detected; according to the method, image information of a corresponding cable is obtained, image inspection is further carried out on the position of the cable suspected to have a fault, image detection is carried out through an image defect diagnosis algorithm, a predicted value of the operation fault probability of the cable based on the image is obtained, and when the predicted value does not reach a predicted threshold value, the cable is detected. Sending a confirmation operation inspection signal to a cable operation inspection unit corresponding to the position information so as to remind the cable operation inspection unit to inspect and confirm whether a fault occurs in a cable area corresponding to the position information; and when the fault cannot be completely eliminated, the fault detection precision is improved, unnecessary inspection operation is reduced, and manpower and material resources are saved.

The invention discloses a precursor alkali-activated reaction activity prediction method based on physicochemical properties and molecular simulation. The method comprises the following steps: acquiring physicochemical properties of a plurality of different alkali-activated precursor materials; acquiring molecular dynamics parameters of the precursor materials under the corresponding physicochemical properties; obtaining the strength activity index of each precursor material through a traditional unconfined compressive strength test; sequencing the physicochemical properties and the molecular dynamics parameters according to the significant degree of the influence on the strength activity index of the precursor material, rejecting non-significant parameters, and retaining significant influence parameters; and based on the corresponding relationship between the significant influence parameter and the strength activity index of each precursor material, establishing a precursor alkali-activated reaction activity index pre-estimation model taking the significant influence parameter as a variable by combining multiple linear regression with an artificial neural network algorithm. According to the method, the accuracy is ensured, and meanwhile, the alkali excitation reaction activity of the precursor material is rapidly predicted.

The invention discloses an ankylosing spondylitis prediction model construction method based on a neural network, and belongs to the technical field of biomedicine, and the method specifically comprises the following steps: sample collection: collecting a blood sample; preparing a sample; performing mass spectrum identification: performing mass spectrum analysis by adopting a Linear mode of a mass spectrometer; performing data analysis; performing MS file processing; performing mass spectrum detection on peak profile data, and setting a threshold value as Plt; 0.01 and log2 (FC) gt; 1, screening a differential mass spectrum peak; carrying out LASSO dimension reduction on the differential mass spectrum peaks to screen mass spectrum characteristic peaks; analyzing the identification efficiency of the mass spectrum characteristic peak on the ankylosing spondylitis by an ROC method; carrying out artificial neural network algorithm modeling; and analyzing and predicting the data of the test set by using the artificial neural network model. More detailed and comprehensive biomarker analysis can be provided, and the overall metabolic state and the pathological process of a patient and good medication curative effect evaluation can be reflected.

The invention discloses a multispectral remote sensing-based mountain original red soil nutrient content inversion method and system. Relates to the technical field of soil nutrient monitoring. The method comprises the following steps: step 1, collecting multispectral remote sensing image data of a mountain original red soil area; 2, preprocessing the multispectral remote sensing image data; 3, extracting soil nutrient content data from the preprocessed multispectral remote sensing image; 4, establishing an inversion model of the mountain original red soil nutrient content by adopting an artificial neural network algorithm; and 5, processing the soil nutrient content data by using the inversion model to obtain a soil nutrient content distribution diagram, and evaluating and analyzing the soil nutrient content distribution diagram. According to the inversion model, through combination of unmanned aerial vehicle multispectral remote sensing and an artificial neural network algorithm, the main nutrient content of the mountain original red soil can be efficiently and accurately inverted, a rapid new means is provided for soil nutrient monitoring, and precision agriculture development is assisted.

The application relates to the technical field of marine hydrological monitoring, and discloses a seawater suspended sediment concentration low-altitude observation method, which uses a low-altitude observation platform to carry digital photography equipment, obtains seawater suspended sediment concentration images, collects water samples in the photographed sea area, positions the water samples through a global satellite positioning system, uses a filtering method to measure the suspended sediment content, pre-processes the obtained seawater suspended sediment concentration images, realizes automatic splicing of the images and automatic imaging of orthographic images, trains the processed image data and the measured suspended sediment concentration data based on an optimized artificial neural network algorithm, establishes a seawater suspended sediment concentration inversion model, and evaluates the model precision. The method uses the low-altitude observation platform to obtain high-resolution images, uses the optimized artificial neural network model to invert the suspended sediment concentration, has the advantages of high precision, low cost, flexibility, timeliness and the like, and can effectively make up for the deficiency of satellite remote sensing.

The present invention discloses a data stream architecture neuromorphic processor, which includes a control unit, a 4D computing unit, a 1D computing unit, an activation unit, a data type conversion unit, a parameter on-chip memory, a 4D data on-chip memory, and a 1D data on-chip memory; the data stream architecture neuromorphic processor is connected to an off-chip memory through a bus, the control unit obtains the configuration information of the neural network algorithm from the off-chip memory, and configures each module according to the configuration information, and configures the computing function of the data stream architecture neuromorphic processor as a certain operator function; through multiple configuration switches by the control unit, the calculations of different operators in the neural network algorithm are sequentially completed, so as to realize the complete calculation of a certain neural network algorithm, and write the final output data into the off-chip memory. Based on the reconfigurable data stream architecture technology, this processor can simultaneously meet the calculations of artificial neural network algorithms and spiking neural network algorithms.

The invention provides a method for predicting nasopharyngeal carcinoma stem cell depletion characteristics by using a blood marker, which comprises the following steps: acquiring cancer tissue samples of a plurality of nasopharyngeal carcinoma patients, and constructing a transcriptome-based nasopharyngeal carcinoma stem cell depletion score for each patient by using single sample gene enrichment analysis ssGSEA; collecting blood marker data of the same batch of patients, including 21 blood routine data, and performing data preprocessing; creating a stem cell depletion scoring model based on the blood marker, and training 21 blood routine data of a nasopharyngeal carcinoma patient by using an artificial neural network ANN algorithm to obtain a trained model; joint visualization is carried out through SHAP feature importance, and the effectiveness of senescence scoring and blood scoring is evaluated by combining correlation analysis and survival analysis; and collecting blood marker data of another batch of patients, performing data preprocessing, inputting the data into the trained model to obtain a score corresponding to each patient, and performing survival analysis to verify the efficiency of the model. According to the invention, effective and low-cost individual aging prediction can be realized.

The invention discloses an optimal dip angle optimization design method based on a plane roof photovoltaic system, which comprises the following steps: taking local past load demand data as reference, and combining a long-short-term memory artificial neural network algorithm to predict the local electricity load demand in the next year; the optimal inclination angle is optimized according to the predicted load demand, the obtained optimal inclination angle can improve the utilization rate of the electric energy load in one year, and the problem of excessive power shortage in individual months is avoided. In addition, the capacity of the photovoltaic device per unit occupied area under the actual installation cost is considered, and the photovoltaic dip angle when the capacity of the photovoltaic device per unit occupied area is maximum is calculated. And finally, optimizing the two schemes by using a game theory method, so that the finally obtained optimal inclination angle not only can improve the utilization rate of an electric energy load in one year, but also can obtain the capacity of a photovoltaic device per unit occupied area.

The invention discloses a solution for edge computing operation and maintenance alarm, and relates to the technical field of edge computing. Comprising the following steps: step 1, collecting all equipment information in an edge computing environment, including running states, performance indexes and fault records of equipment; 2, constructing a state model of the equipment according to the running state and the performance index of the equipment; 3, according to the state model of the equipment, constructing an alarm model of the equipment; 4, when the equipment gives an alarm, determining an alarm reason according to the alarm model of the equipment; step 5, taking corresponding measures for processing according to the reason of the alarm; and 6, performing deep learning and pattern recognition on the operation data of the equipment by applying an artificial neural network algorithm. By applying the artificial neural network and the reinforcement learning technology, the complex relation between the equipment state and the alarm can be automatically identified, and automatic detection and classification of the alarm are realized; the need of manual intervention is greatly reduced, and the operation and maintenance efficiency is improved.

The invention relates to the technical field of data processing, and particularly discloses an AI-based air quality prediction method and system, and the method comprises the steps: S1, dividing a preset region into independent grids of a preset size, and setting the information of each grid; s2, extracting influence factors of grid air quality, wherein the influence factors comprise influence factors of spatial change and influence factors of time change; s3, constructing a space classifier based on an artificial neural network algorithm, and capturing the influence of space factors on the air quality; constructing a time classifier based on a convolution LSTM algorithm, and learning time periodic features in the time sequence data; and S4, through a collaborative algorithm, according to results of the space classifier and the time classifier, obtaining gridding air quality data with a time sequence. By adopting the technical scheme of the invention, the air quality prediction accuracy of mountainous regions and river valley cities can be effectively improved.

The invention discloses a shield tunnel segment assembly error quantification and segment position prediction method, and relates to the technical field of shield subway tunnel construction. The position deviation of a point B is calculated through coordinate data of a central position A of a segment ring where a laser target is located, the distance of a central point B of a segment ring on the outermost side and the position of the laser target actually measured on site in combination with the horizontal and vertical coordinate orientation of the laser target, and then the measured data is compared with the data of the actual position of the segment. And central position data and error rules of the segment assembly rings under different tunneling stratums are obtained through calculation. Error correction is carried out by combining data comparison of a trial tunneling section, a segment position prediction model is established through an artificial neural network algorithm, and the relative position of a next segment is determined. According to the method, the relative position relation between the duct piece position and the tunnel line type can be judged in the tunneling process, the relative position of the duct piece of the next ring can be predicted, errors in the duct piece splicing process can be remarkably reduced, and the splicing quality and the forming quality of the shield tunnel duct pieces are guaranteed.

The application discloses a kind of edge computing operation and maintenance alarm solutions, it is related to edge computing technical field;Including: step 1: collect all equipment information in edge computing environment, including the running state of equipment, performance index, fault record;Step 2: according to the running state of equipment and performance index, the state model of equipment is constructed;Step 3: according to the state model of equipment, the alarm model of equipment is constructed;Step 4: when equipment occurs alarm, according to the alarm model of equipment, the reason of alarm is determined;Step 5: according to the reason of alarm, corresponding measures are taken to handle;Step 6: the running data of equipment is carried out deep learning and pattern recognition by applying artificial neural network algorithm.The application can automatically identify the complex relationship between equipment state and alarm by applying artificial neural network and reinforcement learning technology, realize the automatic detection and classification of alarm;Greatly reduce the need of manual intervention, improve operation and maintenance efficiency.

The invention relates to an intelligent depression assessment method and system based on an ERP and eye movement data fusion model. The method comprises the steps that a data acquisition module obtains psychological scale data, multiple ERP indexes and multiple eye movement indexes; the data analysis module is used for screening ERP (Enterprise Resource Planning) and eye movement indexes significantly related to depressive symptoms; the feature fusion module is used for fusing the screened ERP data and eye movement data and splicing the data into a fusion feature vector, and the model training module is used for inputting the fusion feature vector into a depression dichotomy recognition model constructed by combining Transform and an artificial neural network (ANN) algorithm and outputting a depression risk assessment result; and the prediction output module is used for predicting the data acquired by the new user by adopting the trained model and outputting a depression risk assessment result through a visual interface. According to the system, the technical bottleneck problems of subjective defect of traditional psychological scale evaluation, insufficient sensitivity and specificity of single physiological index detection and multi-modal data fusion are solved.

This invention discloses an optimal tilt angle design method for a planar rooftop photovoltaic (PV) system. Using historical local load demand data as a reference, it combines a long short-term memory (LSTM) artificial neural network algorithm to predict the local electricity load demand for the next year. Based on this predicted load demand, the optimal tilt angle is optimized. The obtained optimal tilt angle improves the annual electricity load utilization rate and avoids excessive power shortages in certain months. Furthermore, this invention considers the PV device capacity per unit area under actual installation costs and calculates the PV tilt angle that maximizes the PV device capacity per unit area. Finally, game theory is used to optimize the two schemes, ensuring that the final optimal tilt angle improves both the annual electricity load utilization rate and the PV device capacity per unit area.

The invention discloses a complex part design method and system based on a digital-real fusion test, and belongs to the field of electronic engineering and computer science, and the method comprises the steps: designing a model construction module of a complex part, then, physical constraints in the entity environment and additive manufacturing constraints in the manufacturing process are introduced, iteration and verification of the model are completed based on the additive manufacturing technology and user requirements, and a complex part model with data and real fusion is obtained; a complex part model topological optimization module is designed, the module carries out finite element division on an original model, optimization parameters are set based on multiple topological optimization algorithms and an artificial neural network algorithm, and a group of topological optimization result models are obtained; and a detection and correction module of the complex part model is designed, the module constructs an evaluation system based on user requirements and model constraints, the optimization model is evaluated through digital testing, and an optimal design scheme or correction suggestion is given.

The application discloses a duckweed carbon sink evaluation method, system and equipment suitable for a rice-duckweed symbiotic system, relates to the technical field of agricultural production and ecological environment, and discloses the following technical scheme: after rice is harvested, the information of duckweed in the rice-duckweed symbiotic system is collected by using spectral scanning and laser point cloud technology, and the volume of the duckweed is determined based on the duckweed information; the duckweed biomass in the rice-duckweed symbiotic system is determined according to the duckweed volume; the duckweed carbon sink is estimated by using an artificial neural network algorithm and the duckweed biomass, or the duckweed carbon sink is estimated by using a duckweed carbon sink estimation formula and the duckweed biomass. The application can flexibly evaluate the duckweed carbon sink according to information such as rice agronomic management, soil and climate, and provides good technical support for the management and application of duckweed in a rice field.

The application discloses a kind of molecular potential energy field analysis methods based on multivariate feature information transmission, it is related to molecular dynamics technical field, including steps: obtaining the element characteristic vector and structure characteristic vector of target molecular formula;By undercomplete self-encoder constitutes coding network, and the multivariate characteristic vector formed by element characteristic vector and structure characteristic vector is handled to obtain atom representation;According to atom representation, obtain interatomic distance and aggregate adjacent atoms, and update adjacent aggregated atom representation by continuous convolution filter;According to the updated atom representation, atom potential is predicted by artificial neural network algorithm, and the total potential energy of system is obtained according to the atom potential of each atom in target molecular formula.The application is by the use of general local environment characteristics, and makes atom representation contain more feature information, guarantees the feature description of different material systems, and then improves the generality of model.

The application provides an air compressor optimization method and system based on artificial intelligence and cloud edge collaboration. The method comprises the following steps: using a preset full connection neural network algorithm to model an air compressor performance model, obtaining a power-yield neural network model and a power-frequency neural network model, and storing them in the cloud; the edge end adjusts the frequency of the air compressor according to the model; using a preset long short-term memory artificial neural network algorithm to construct a demand prediction model and storing it in the cloud; the edge end predicts the current demand of the air compressor unit according to the demand prediction model, obtains the expected demand, calculates the expected total power of the air compressor unit when the air compressor unit reaches the expected demand, obtains the full load flow rate of the air compressor, compares the expected demand with the full load flow rate, selects the best scheduling optimization scheme of the air compressor unit according to the comparison result, and realizes that the air compressor unit is always in the optimal efficiency state under the working condition, and achieves the purpose of energy saving and cost reduction.

The invention relates to the technical field of biological medicine, and discloses a method for constructing a pregnant woman health index evaluation system, which comprises the following steps: collecting original data of a pregnant woman in a perinatal period, cleaning the original data, removing abnormal values, filling missing values, and carrying out standardization processing to obtain preprocessed data; a random forest feature screening technology is adopted, variables with feature importance higher than an average value are selected, a composite perinatal period bad outcome prediction model is established based on an artificial neural network algorithm, and a prediction value output by the prediction model is converted into a pregnant woman health index through linear conversion; according to the method, features screened in the construction process of a composite perinatal period bad outcome prediction model are used as independent variables, a Logistic regression model is used for prediction, model fitting is carried out for different bad outcomes, respective Logistic regression equations are obtained, and a single bad outcome prediction model is constructed; according to the invention, the early warning capability is improved, the clinical management efficiency is enhanced, and the occurrence rate of bad outcomes in the perinatal period is reduced.

The invention discloses a building waterproof layer damage trend prediction method, which is characterized in that a plurality of groups of sensor clusters are dispersedly arranged on a building waterproof layer, whether the building waterproof layer is damaged or not is reflected by monitoring physical property change data of the building waterproof layer through the sensor clusters, and the physical property change data comprise strain, humidity, temperature, optical wavelength and audio frequency; monitoring physical property change data of the building waterproof layer in real time by each group of sensor clusters in a wireless transmission mode, and transmitting the data to analysis equipment; on an analysis device, layout position data of a sensor cluster on a building waterproof layer, acquired physical performance change data and a time sequence are used as input data of an artificial neural network algorithm; and generating damage probability and damage trend data of the building waterproof layer at each layout node position through an artificial neural network algorithm to predict the damage trend of the building waterproof layer. According to the invention, the problem of monitoring hysteresis due to the damage of the waterproof layer of the video monitoring building can be avoided.

An electronic device is provided. The electronic device may include: a microphone; a communication module including a communication circuit configured to be connected to one or more external electronic devices; and a processor. The processor may be configured to: identify a user based on at least voice data including phoneme information acquired using a microphone; identify first usage pattern information corresponding to the first user based on a specified artificial neural network (ANN) algorithm when the user is a first user, and control a first specified electronic device belonging to one or more external electronic devices and corresponding to the phoneme information via the communication module based on the identified first usage pattern information to perform a first specified operation; and identify second usage pattern information corresponding to the second user based on a specified ANN algorithm when the user is a second user, and control a second specified electronic device belonging to one or more external electronic devices and corresponding to the phoneme information via the communication module based on the identified second usage pattern information to perform a second specified operation. Various embodiments of the present disclosure may use an ANN algorithm to easily control at least one electronic device desired by a user based on user pattern information corresponding to the user.

The invention relates to an intelligent recognition method and system for website sensitive words and a storage medium. The intelligent recognition method for the website sensitive words comprises the following steps: S1, website text data acquisition: in a website data acquisition stage, acquiring original data of a website, and fully acquiring source code data of an entered website in real time; s2, website text data preprocessing: after website data acquisition is completed, preprocessing the data; and S3, intelligent recognition of the sensitive words of the website text: an artificial neural network algorithm is realized by constructing a neuron model, and sensitive word information in the text is automatically and quickly recognized. Iterative self-organizing feature extraction is carried out by means of an artificial neural network algorithm, sensitive information is automatically recognized, high-precision recognition capability is achieved, and the detection rate and the accuracy rate of the sensitive words of the website are increased; high throughput and low delay are maintained, the detection efficiency is improved, and the timeliness is ensured.

The invention relates to the technical field of marine hydrology monitoring, and discloses a low-altitude observation method for seawater suspended sediment concentration, which comprises the following steps of: carrying digital photographic equipment by using a low-altitude observation platform, acquiring a seawater suspended sediment concentration image, simultaneously carrying out water sample collection on a photographic sea area, positioning by using a global satellite positioning system, and measuring the suspended sediment content by using a filtering method. The obtained seawater suspended sediment concentration image is preprocessed, and automatic splicing of the image and automatic imaging of the orthoimage are achieved. And based on an optimized artificial neural network algorithm, training by using the processed image data and actually measured suspended sediment concentration data, establishing a seawater suspended sediment concentration inversion model, and evaluating the model precision. According to the method, a high-resolution image is obtained by using a low-altitude observation platform, suspended sediment concentration inversion is carried out by optimizing an artificial neural network model, the method has the advantages of high precision, low cost, flexibility, timeliness and the like, and the defects of satellite remote sensing can be effectively overcome.

The invention discloses a face mask binding method based on real-time face 3D key point detection, and the method comprises the following steps: S1, constructing a deep learning model which is a software or hardware computing system based on an artificial neural network algorithm, and comprises one or more trained or untrained artificial neural networks; s2, acquiring a face image through a camera; s3, extracting feature information of the key points of the face image, wherein the feature information refers to related information which is included in the key points of the face image and is required for three-dimensional coordinate measurement of the key points of the face image; s4, inputting the feature information of the key points of the face image into the deep learning model; s5, the deep learning model calculates three-dimensional coordinates of key points of the face image; and S6, constructing a face grid model according to the three-dimensional coordinates, and generating a finished face mask for real-time face 3D key point detection. According to the invention, the problem that the three-dimensional key point coordinates of the human face are difficult to accurately obtain in real time can be solved.

The present invention discloses a method for rapid analysis of fertilizer application based on grass yield, and specifically relates to the technical field of increasing plant yield. The processing method comprises the following steps: monitoring and collecting information related to soil properties in a target area, and using an artificial neural network algorithm to construct a model using crop yield data and historical fertilizer application data. The present invention is based on a method for rapid analysis of fertilizer application based on grass yield, evaluates soil properties in a target area, and constructs a model for soil suitable for plants required by users through an artificial neural network algorithm based on the evaluation results and using crop yield data and historical fertilizer application data as factors, outputs fertilizer application data with the highest grass yield, and optimizes and adjusts the obtained model to make the acquired data more accurate.

The application relates to a shield tunnel segment assembling error quantification and segment position prediction method, and relates to the technical field of shield subway tunnel construction. The position deviation of a point B is calculated by combining the horizontal and vertical coordinate positions of the position deviation of the point B with the coordinate data of a center position A of a segment ring at a position of a laser target and the distance of the center position B of the outermost segment ring and the actually measured position of the laser target. Then, the measured data and the actual position data of the segment are compared, and the center position data of a segment assembling ring in different tunneling strata and error rules are calculated. The error is corrected by comparing the data of a trial tunneling section, a segment position prediction model is established by using an artificial neural network algorithm, and the relative position of a next segment ring is determined. The application can judge the relative position relationship between the segment position and the tunnel line type during tunneling, and can predict the relative position of the next segment ring. The application can significantly reduce the error in the segment assembling process, and can guarantee the assembling quality and forming quality of the shield tunnel segment.

The invention provides an air compressor adjusting and optimizing method and system based on artificial intelligence and cloud edge collaboration. The method comprises the steps that a preset full-connection neural network algorithm is applied to conduct air compressor performance model modeling, a power-yield neural network model and a power-frequency neural network model are obtained and stored in a cloud end, and an edge end conducts frequency adjustment on an air compressor according to the models. A demand prediction model is constructed by applying a preset long-short-term memory artificial neural network algorithm and is stored in the cloud, the edge end predicts the current demand of the air compressor unit according to the demand prediction model to obtain the expected demand, the edge end calculates the expected total power of the air compressor unit when the air compressor unit reaches the expected demand, and the expected total power of the air compressor unit reaches the expected demand. And the edge end obtains the full-load flow velocity of the air compressor, compares the expected demand with the full-load flow velocity, and selects the optimal scheduling optimization scheme of the air compressor unit according to the comparison result, so that the air compressor unit can be always in the optimal efficiency state under the working condition, and the purposes of saving energy and reducing cost are achieved.