Method and apparatus for adjusting rotation speed, nonvolatile storage medium, and electronic device

Abstract

This application discloses a method and apparatus for adjusting rotation speed, a non-volatile storage medium, and an electronic device. The method includes: acquiring environmental parameters of a detection area at multiple detection times, wherein the environmental parameters are parameters related to the dust concentration of the detection area; extracting and processing high-order features of the environmental parameters using a dust concentration prediction model to obtain a predicted dust concentration value, wherein the predicted dust concentration value indicates the dust concentration of the detection area at the next detection time, and the high-order features are time-series features of the environmental parameters; adjusting the rotation speed of a target device according to the predicted dust concentration value, wherein different rotation speeds of the target device result in different exhaust intensities for the detection area. This application solves the technical problem of inaccurate adjustment results and increased equipment wear caused by manually adjusting the rotation speed of equipment with air circulation functions in related technologies.

Description

Technical Field

[0001] This application relates to the field of automation control technology, and more specifically, to a method and apparatus for adjusting rotational speed, a non-volatile storage medium, and an electronic device. Background Technology

[0002] Ductless air handling units (DULUs) are crucial equipment in steel production. In steelmaking and coking plant operations, the concentration of metal and coal dust in the air is high. Prolonged operation in this environment poses serious health risks to personnel and pollutes the surrounding environment. DULUs effectively remove dust and exhaust gases through ducts, reducing dust concentration and protecting personnel. They also allow for the processing and recycling of useful materials for reuse. Current technologies primarily rely on personnel to control the speed of DULUs based on ambient dust levels. However, this approach suffers from limitations in timely and accurate speed adjustments, leading to resource waste and increased equipment wear.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides a method and apparatus for adjusting rotational speed, a non-volatile storage medium, and an electronic device, to at least solve the technical problem of increased equipment wear caused by inaccurate adjustment results due to manual adjustment of the rotational speed of equipment with air circulation function in related technologies.

[0005] According to one aspect of the embodiments of this application, a method for adjusting rotation speed is provided, comprising: acquiring environmental parameters of a region to be detected at multiple detection times, wherein the environmental parameters are parameters related to the dust concentration of the region to be detected, and the difference between any two adjacent detection times among the multiple detection times is a preset value; extracting and processing high-order features of the environmental parameters using a dust concentration prediction model to obtain a dust concentration prediction value, wherein the dust concentration prediction value is used to indicate the dust concentration of the region to be detected at the next detection time, the high-order features are time-series features of the environmental parameters related to time, and the dust concentration prediction model is obtained by training a neural network model using historical environmental parameters of the region to be detected as training data; adjusting the rotation speed of a target device according to the dust concentration prediction value, wherein different rotation speeds of the target device result in different exhaust intensities for the region to be detected, and the target device is a device with air circulation function in the region to be detected.

[0006] Optionally, a dust concentration prediction model is used to extract and process high-order features of environmental parameters to obtain predicted dust concentration values. This includes: determining the detection time corresponding to each environmental parameter; arranging multiple different types of environmental parameters corresponding to the same detection time to obtain multiple environmental parameter sequences corresponding to multiple detection times, wherein the types of environmental parameters include: humidity, temperature, air pressure, air velocity, and dust concentration; performing convolution processing on the multiple environmental parameter sequences in different neural network layers of the dust concentration prediction model to obtain high-order features, wherein the first neural network layer is used to extract high-order features, and other neural network layers are used to process high-order features, with different dimensions of convolution kernels used in different neural network layers; and determining the dust concentration prediction value based on the processing results of the high-order features and the multiple environmental parameter sequences.

[0007] Optionally, multiple environmental parameter sequences are convolved in different neural network layers of the dust concentration prediction model to obtain higher-order features. This includes: for each neural network layer, determining the detection time corresponding to each environmental parameter sequence, and arranging multiple environmental parameter sequences into a matrix to be processed according to the detection time from early to late, wherein each row of the matrix to be processed is an environmental parameter sequence; for the Nth row of the matrix to be processed, a new matrix composed of the data from the first N rows of the matrix to be processed is convolved using a preset convolution kernel to obtain a first convolution result, wherein the preset convolution kernel is a matrix with the same dimension as the new matrix; the sum of multiple first convolution results is determined as the convolution sum, and the preset convolution kernel and multiple convolution sums are processed to obtain higher-order features.

[0008] Optionally, the preset convolutional kernel and multiple first convolutional results are processed to obtain higher-order features, including: padding the preset convolutional kernel to obtain multiple other convolutional kernels of different dimensions; using one other convolutional kernel to perform convolution processing on the convolution and obtain a second convolutional result in the target other neural network layer, wherein the target other neural network layer is another neural network layer connected to the first neural network layer; according to the connection order of the remaining other neural network layers, using other convolutional kernels of different dimensions to perform convolution processing on the input data in different other neural network layers in sequence to obtain the target convolutional result output by the last other neural network layer, wherein during the convolution processing, the input of the first other neural network in sequence is the second convolutional result, and the output of the previous other neural network layer is used as the input data of the next other neural network layer; and determining the higher-order features based on the target convolutional result and the environmental parameter sequence.

[0009] Optionally, the higher-order features are processed, including: acquiring hidden features of the dust concentration detection model, wherein the hidden features are used to indicate the relationship between environmental parameters and the detection time, and the hidden features are determined by the dust concentration prediction model based on training data during training; using a recurrent neural memory network included in the dust concentration detection model to perform feature fusion processing on the hidden features and higher-order features to obtain a feature fusion result, wherein the recurrent neural memory network contains multiple sequentially connected recurrent units, and during the feature fusion process, the output of the previous recurrent unit is used as the input of the next recurrent unit; converting the feature fusion result into a target data sequence, wherein the amount of data contained in the target data sequence is the same as the amount of data contained in the environmental parameter sequence, and the dust concentration value contained in the target data sequence is the predicted dust concentration value.

[0010] Optionally, adjusting the rotational speed of the target equipment based on the predicted dust concentration includes: obtaining the historical predicted dust concentration value corresponding to the previous detection time, and determining the predicted value of the control parameter based on the historical predicted dust concentration value and the predicted dust concentration value, wherein the control parameter is a parameter used to control the rotational speed of the target equipment; obtaining multiple historical deviation values ​​corresponding to multiple historical detection times before the current detection time, and determining the predicted deviation value based on the multiple historical deviation values, wherein the historical deviation value is the difference between the historical predicted dust concentration value and the historical measured dust concentration value; determining the target value of the control parameter based on the predicted value and the predicted deviation value, and adjusting the value of the control parameter to the target value.

[0011] Optionally, before extracting and processing the higher-order features of environmental parameters using the dust concentration prediction model, the process includes: classifying environmental parameters of the same type into a dataset; for each dataset, arranging the environmental parameters in order of detection time from early to late as a sequence to be processed; for each sequence to be processed, determining the value range based on the values ​​of the environmental parameters contained in the sequence; and performing numerical correction processing on the environmental parameters if the values ​​of the environmental parameters do not fall within the value range.

[0012] According to another aspect of the embodiments of this application, a device for adjusting rotation speed is also provided, comprising: an acquisition module, configured to acquire environmental parameters of a region to be detected at multiple detection times, wherein the environmental parameters are parameters related to the dust concentration of the region to be detected, and the difference between any two adjacent detection times among the multiple detection times is a preset value; a processing module, configured to extract and process high-order features of the environmental parameters using a dust concentration prediction model to obtain a dust concentration prediction value, wherein the dust concentration prediction value is used to indicate the dust concentration of the region to be detected at the next detection time, the high-order features are time-series features of the environmental parameters, and the dust concentration prediction model is obtained by training a neural network model using historical environmental parameters of the region to be detected as training data; and an adjustment module, configured to adjust the rotation speed of a target device according to the dust concentration prediction value, wherein the target device is a device with air circulation function in the region to be detected.

[0013] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, which stores a computer program, wherein the device where the non-volatile storage medium is located executes the above-described method for adjusting the rotation speed by running the computer program.

[0014] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to execute the above-described method for adjusting the rotation speed through the computer program.

[0015] According to another aspect of the embodiments of this application, a computer program product is also provided, including computer instructions that, when executed by a processor, implement the above-described method for adjusting rotation speed.

[0016] In this embodiment, environmental parameters of the area to be detected are obtained at multiple detection times. These environmental parameters are related to the dust concentration in the area to be detected, and the difference between any two adjacent detection times is a preset value. A dust concentration prediction model is used to extract and process high-order features of the environmental parameters to obtain predicted dust concentration values. These predicted dust concentration values ​​indicate the dust concentration of the area to be detected at the next detection time. The high-order features are time-series features of the environmental parameters. The dust concentration prediction model is trained using historical environmental parameters of the area to be detected as training data on a neural network model. The rotation speed of the target device is adjusted according to the predicted dust concentration values. When the rotation speed of the target device is different, the rotation speed of the area to be detected is adjusted accordingly. By varying exhaust intensities and targeting devices with air circulation functions within the tested area, environmental parameters of the area containing such devices (e.g., ducted air conditioners) are collected and input into a trained neural network model to predict the current dust concentration. Based on this dust concentration, the rotational speed of the air-circulating device (e.g., the ducted air conditioner) is determined. This approach applies the neural network model to predicting dust concentration and adjusting the rotational speed of the ducted air conditioner, thereby improving the accuracy of speed control, reducing equipment wear, and saving energy. It also solves the problem of inaccurate adjustments and increased equipment wear caused by manual adjustment of the rotational speed of air-circulating devices in related technologies. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0018] Figure 1 This is a hardware structure block diagram of a computer terminal for implementing a method for adjusting rotation speed according to an embodiment of this application;

[0019] Figure 2 This is a flowchart illustrating the steps of a method for adjusting rotational speed according to an embodiment of this application;

[0020] Figure 3 This is a schematic diagram of a data processing flow of a Temporal Convolutional Network (TCN) according to an embodiment of this application;

[0021] Figure 4 This is a schematic diagram of an improved data processing flow for a loop unit according to an embodiment of this application;

[0022] Figure 5 This is a structural diagram of a device for adjusting rotational speed according to an embodiment of this application;

[0023] Figure 6 This is a schematic diagram of a system for adjusting rotational speed according to an embodiment of this application;

[0024] Figure 7 This is a comparison of metrics when multiple models are used to process the same test set according to the embodiments of this application;

[0025] Figure 8 The prediction results of dust concentration based on the same validation set by multiple models according to the embodiments of this application are as follows. Detailed Implementation

[0026] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0027] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0028] To better understand the embodiments of this application, the technical terms involved in the embodiments of this application are explained below:

[0029] Proportional-Integral-Differential (PID) control algorithm: An algorithm that controls the deviation by proportional (P), integral (I), and derivative (D).

[0030] In related technologies, when applying neural network models to predict dust concentration, problems arise such as inaccurate feature extraction leading to inaccurate dust concentration predictions, and high energy consumption due to the complex structure of the neural network model. To address these issues, this application provides relevant solutions, which are detailed below.

[0031] According to an embodiment of this application, a method for adjusting rotational speed is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0032] The methods and embodiments provided in this application can be executed on mobile terminals, computer terminals, or similar computing devices. Figure 1 A hardware block diagram of a computer terminal for implementing a method for adjusting rotational speed is shown. Figure 1 As shown, the computer terminal 10 may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0033] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as a form of processor control (e.g., selection of a variable resistor termination path connected to an interface).

[0034] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the speed adjustment method in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the speed adjustment method described above. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0035] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0036] The display may be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10.

[0037] This application provides a method for adjusting rotational speed that can operate under the above-described operating environment. Figure 2 This is a flowchart of the steps of the method for adjusting rotational speed according to the embodiments of this application, as follows: Figure 2 As shown, the method includes the following steps:

[0038] Step S202: Obtain environmental parameters of the area to be detected at multiple detection times. The environmental parameters are parameters related to the dust concentration of the area to be detected. The difference between any two adjacent detection times is a preset value.

[0039] This application provides a control system for predicting rotational speed based on multi-source information fusion. The multi-source information refers to the environmental information of the area where the equipment with air circulation function, controlled by rotational speed, is located (i.e., the area to be detected). Therefore, in step S202, environmental parameters related to dust concentration in the area to be detected are first acquired. These acquired environmental parameters can affect the dust concentration in the area to be detected. Since the dust concentration in the area changes in real time with the environment, in this embodiment, the predicted rotational speed is updated by periodically acquiring the environmental parameters in the area to be detected. For example, according to a pre-set detection cycle (e.g., 6 minutes per cycle), the environmental parameters of the area to be detected are collected at multiple detection moments within the detection cycle. The difference between any two adjacent detection moments is equal, and is the duration of the pre-set detection cycle (i.e., a preset value).

[0040] Step S204: The dust concentration prediction model is used to extract and process the high-order features of the environmental parameters to obtain the dust concentration prediction value. The dust concentration prediction value is used to indicate the dust concentration of the area to be detected at the next detection time. The high-order features are the time-series features of the environmental parameters. The dust concentration prediction model is obtained by training the neural network model with the historical environmental parameters of the area to be detected as training data.

[0041] Since the rotational speed is predicted based on environmental parameters, after obtaining the environmental parameters of the area to be detected in step S202, in step S204, the collected environmental parameters are used as input to the dust concentration prediction model. The dust concentration prediction model performs feature extraction processing on the input environmental parameters to obtain high-order features representing the relationship between dust concentration and time, and calculates the predicted dust concentration value based on the high-order features. The predicted dust concentration value is the predicted dust concentration of the area to be detected at the next detection time. As can be seen from the above, since the high-order features represent the relationship between dust concentration and time, they have temporal characteristics. In essence, they are time-related temporal features extracted from the environmental parameters. For example, it is a vector representing the region of change of environmental parameters over time. The reason why the dust concentration prediction model can be used to predict the dust concentration of the area to be detected at the next detection time is that the dust concentration prediction model is a neural network model trained using environmental parameters of the area to be detected collected at multiple historical detection times before the current detection time (or the previous detection time) as training data. During the training process, it learns the relationship between dust concentration, environmental parameters, and time.

[0042] In this embodiment, when using a dust concentration prediction model, the dust concentration prediction model can be loaded into memory. For example, the raw data of the dust concentration prediction model can be loaded from non-volatile memory into volatile memory so that the processor can run the dust concentration prediction model. The raw data of the dust concentration prediction model refers to unprocessed data, which typically includes the parameters and structural data of the dust concentration prediction model. The structural data can be the calculation relationships based on the parameters, such as the forward propagation calculation relationships between intermediate layers and between neurons. Specifically, the structural data can include structure-related code of the dust concentration prediction model, such as code used to perform related calculations between intermediate layers and between neurons.

[0043] In one implementation, a partition can be created in memory for loading the dust concentration prediction model, which may include a structure data storage area and a parameter storage area. The structure data storage area stores structure-related code, and the parameters referenced by it can be accessed via pointers to the addresses of specific parameters in the parameter storage area. During the training of the dust concentration prediction model, frequent parameter updates may be required; in this case, updating the parameter values ​​in the parameter storage area is sufficient.

[0044] According to an optional embodiment of this application, a dust concentration prediction model is used to extract and process high-order features of environmental parameters to obtain predicted dust concentration values. This includes: determining the detection time corresponding to each environmental parameter; arranging multiple different types of environmental parameters corresponding to the same detection time to obtain multiple environmental parameter sequences corresponding to multiple detection times, wherein the types of environmental parameters include: humidity, temperature, air pressure, air velocity, and dust concentration; performing convolution processing on the multiple environmental parameter sequences in different neural network layers of the dust concentration prediction model to obtain high-order features, wherein the first neural network layer is used to extract high-order features, and other neural network layers are used to process the high-order features, with different dimensions of convolution kernels used in different neural network layers; and determining the dust concentration prediction value based on the processing results of the high-order features and the multiple environmental parameter sequences.

[0045] Scenarios requiring dust concentration monitoring (i.e., the area to be monitored) typically include various scenarios such as steel production sites. In these scenarios, parameters related to dust concentration include temperature, humidity, air pressure, air velocity, and dust concentration. These parameters can be acquired through temperature sensors, humidity sensors, air pressure sensors, wind speed sensors, dust concentration detectors, etc., which are pre-set in the area to be monitored and communicate with a device / system running a dust concentration prediction model. These parameters change with the production process; that is, environmental parameters and dust concentrations differ at different times, meaning dust concentration is time-dependent. To predict the dust concentration value at the next monitoring time based on environmental parameters, the input data needs to be preprocessed to make it time-dependent. In this embodiment, the dust concentration prediction model first divides the input environmental parameters into multiple sets according to the monitoring time at which they were collected. Multiple environmental parameters acquired at the same monitoring time are then grouped into an environmental parameter sequence. Therefore, multiple environmental parameter sequences can be obtained, where the time at which each environmental parameter was collected is its corresponding monitoring time. The dust concentration prediction model is a neural network model containing multiple neural network layers. By performing convolution processing on the environmental parameter sequence generated by the aforementioned arrangement within different neural network layers, the model extracts time-related temporal features (i.e., higher-order features) from the environmental parameters. Different convolution kernels are used in different neural network layers. In this embodiment, the dust concentration prediction model uses the first neural network layer to extract higher-order features, and the remaining neural network layers process the extracted higher-order features. The dust concentration prediction model further predicts the dust concentration value of the area to be detected at the next detection time based on the processing results of the higher-order features and the input environmental parameter sequence, and outputs the predicted dust concentration value.

[0046] Optionally, multiple environmental parameter sequences are convolved in different neural network layers of the dust concentration prediction model to obtain higher-order features. This includes: for each neural network layer, determining the detection time corresponding to each environmental parameter sequence, and arranging multiple environmental parameter sequences into a matrix to be processed according to the detection time from early to late, wherein each row of the matrix to be processed is an environmental parameter sequence; for the Nth row of the matrix to be processed, a new matrix composed of the data from the first N rows of the matrix to be processed is convolved using a preset convolution kernel to obtain a first convolution result, wherein the preset convolution kernel is a matrix with the same dimension as the new matrix; the sum of multiple first convolution results is determined as the convolution sum, and the preset convolution kernel and multiple convolution sums are processed to obtain higher-order features.

[0047] Figure 3 This is a diagram illustrating the data processing flow of TCN, such as... Figure 3As shown, in this embodiment, when the dust concentration prediction model extracts high-order features, it first arranges multiple environmental parameter sequences in order from earliest to latest according to their corresponding detection times, obtaining a matrix to be processed. Each environmental parameter sequence serves as a row in the matrix, and each column contains multiple environmental parameters acquired at different detection times. For example, the environmental parameter sequences include: X0, X1, X2, X3, and X4, where X0 corresponds to the earliest detection time, X1 corresponds to a later detection time than X0, X2 corresponds to a later detection time than X1, X3 corresponds to a later detection time than X2, and X4 corresponds to the latest detection time. The matrix to be processed (X) can then be represented as: X = {X0, X1, X2, X3, X4}, where each environmental parameter sequence X... k(k＝0、1、2、3、4) ={x M}, where M represents M environmental parameters; each column X of the matrix to be processed l(k＝0,….M) ={x N}, where N represents N detection times; x N This refers to the same environmental parameter collected at the Nth time point. The dust concentration prediction model uses causal convolution and dilated convolution to extract high-order features. The first neural network layer is a Temporal Convolutional Network (TCN) layer, containing multiple TCNs. These multiple TCNs extract high-order features from the input data in parallel, enabling rapid modeling of dust concentration changes. Specifically, when using TCN for causal convolution, only environmental parameters from multiple times before the prediction time are considered, without considering future parameters. Since the data at a given time point is one row of the matrix to be processed, extracting high-order features in each temporal convolutional network layer requires the combined effect of environmental parameter sequences from multiple times before that time point to extract the temporal features between the environmental parameter and time at each time point. Specifically, when processing the matrix, after extracting each row... When extracting temporal features, the first N rows of a given row are needed to work together. In this embodiment, the first N rows of the matrix to be processed are considered as a new matrix. When extracting temporal features from each row, a pre-defined convolution kernel with the same dimension as the new matrix is ​​used to convolve with the new matrix, resulting in a (first) convolution result. For example, when predicting the temporal features of the time point corresponding to the 3rd row of the matrix to be processed, the environmental parameter sequence of the first 3 rows of the matrix to be processed is used. Through the above method, convolution processing of a matrix to be processed containing N rows can yield N (first) convolution results. The sum of these N (first) convolution results (i.e., the convolution sum) and the pre-defined convolution kernel are then used to obtain higher-order features. The above causal convolution process can be expressed as the formula: Where X is the matrix to be processed, W is a parameter matrix containing multiple preset convolutional kernels, and each preset convolutional kernel is also a matrix. i Represents the environmental parameters at the i-th detection time; k represents the k-th element, for example, w k x represents the k-th element in the parameter matrix. k This represents the k-th element in the matrix to be processed, and N represents the dimension of the convolution kernel, which is also the dimension of the matrix to be processed.

[0048] Optionally, the preset convolutional kernel and multiple first convolutional results are processed to obtain higher-order features, including: padding the preset convolutional kernel to obtain multiple other convolutional kernels of different dimensions; using one other convolutional kernel in the target other neural network layer to perform convolution processing on the convolution and obtain a second convolutional result, wherein the target other neural network layer is another neural network layer connected to the first neural network layer; according to the connection order of the remaining other neural network layers, using other convolutional kernels of different dimensions to perform convolution processing on the input data in different other neural network layers in sequence to obtain the target convolutional result output by the last other neural network layer, wherein during the convolution processing, the input of the first other neural network in sequence is the second convolutional result, and the output of the previous other neural network layer is used as the input data of the next other neural network layer; determining the higher-order features based on the target convolutional result and the environmental parameter sequence.

[0049] Still Figure 3As shown, in this embodiment, multiple (first) convolution results from the first layer output of the TCN and a preset convolution kernel are processed together in other neural network layers of the dust concentration prediction model to obtain higher-order features. Specifically, the dimension of the preset convolution kernel is changed by padding it to obtain multiple new convolution kernels (i.e., other convolution kernels) with different dimensions. In this embodiment, to avoid affecting the accuracy of the prediction results, the preset convolution kernel can be padded with the number "0". After obtaining multiple new convolution kernels (i.e., other convolution kernels) with different dimensions, convolution processing is performed on the new convolution kernels (i.e., other convolution kernels) with different dimensions and the input data of the layer in other (target) neural network layers connected to the neural network layer (i.e., the first neural network layer) used to extract higher-order features in the dust concentration prediction model. The input data of each target neural network layer is the output of the previous target neural network layer connected to it. In this embodiment, the first layer of the TCN outputs a sum of multiple (first) convolution results (i.e., a convolution sum). The neural network layer connected to the first layer of the TCN (i.e., other target neural network layers) then performs convolution processing on the convolution sum using a new convolution kernel with the same dimension as the convolution sum. The resulting (second) convolution result is used as the input to the next neural network layer connected to the other target neural network layers. For example, pre-defined convolution kernels with dilation coefficients of 2 and 4 can be used in the second and third layers of the TCN for dilated convolution. In this embodiment, by using the above dilated convolution method, the temporal coverage of the input data is increased by increasing the convolution kernel size, which can better extract long-term dependencies in the time series and solve the problem of limited receptive field in the original network structure. The dilated convolution process in this embodiment can be expressed as the formula: Where X is the convolution sum, which is in matrix form, and W is a matrix composed of multiple new convolution kernels. k is the k-th element in W, i.e., the k-th new convolution kernel, representing the number of new convolution kernels, s represents the stride of the new convolution kernel, i.e., the interval at which the new convolution kernel slides on the input data, and d represents the inflation factor (inflation coefficient). By convolving the convolution kernel at time k with the environmental parameters (sequence) at time [k-(Ns)d], the long-term dependencies in the time series can be better extracted. After the dust concentration prediction model uses the above method to perform convolution and multi-layer processing, the last layer of the TCN will output a final convolution result (i.e., the target convolution result). In this embodiment, the target convolution result and the input environmental parameter sequence are subjected to residual connection processing to obtain higher-order features. For example, if the TCN has 3 layers, the result output by the 3rd layer of the TCN (target convolution result) is subjected to residual operation with the input data (environmental parameter sequence) at the same time. Taking time t... 31 For example, the environmental parameter sequence x at this moment 31After a 1×1 convolution branch, the output x of the third layer, which has undergone ReLU activation and Dropout regularization, is... 31 Then, perform element-wise addition to obtain the final output of the TCN network (i.e., higher-order features); this step alleviates the problem of gradient vanishing or gradient explosion caused by the increase in model depth.

[0050] According to another optional embodiment of this application, processing the higher-order features includes: acquiring hidden features of the dust concentration detection model, wherein the hidden features are used to indicate the relationship between environmental parameters and the detection time, and the hidden features are determined by the dust concentration prediction model based on training data during the training process; performing feature fusion processing on the hidden features and higher-order features using a recurrent neural memory network included in the dust concentration detection model to obtain a feature fusion result, wherein the recurrent neural memory network contains multiple sequentially connected recurrent units, and during the feature fusion process, the output of the previous recurrent unit is used as the input of the next recurrent unit; converting the feature fusion result into a target data sequence, wherein the amount of data contained in the target data sequence is the same as the amount of data contained in the environmental parameter sequence, and the dust concentration value contained in the target data sequence is the predicted dust concentration value.

[0051] Figure 4 This is a schematic diagram of the data processing flow of the improved recurrent unit in the dust concentration prediction model. The method provided in this application further adds feature fusion operations to the Long Short Term Memory (LSTM) unit in the dust concentration prediction model. Specifically, before the recurrent unit gating structure operation of the LSTM unit, a multiplication operation between the hidden state and the input state is added. In this embodiment, when processing higher-order features in the LSTM unit, the hidden features and higher-order features are fused in the recurrent neural memory network. The hidden features are features extracted from the training data during the dust concentration prediction model's training process to indicate the correlation between environmental parameters and the detection time (for example, the hidden feature can be a feature vector indicating the changing trend of environmental parameters with the detection time). Figure 4 As shown, the feature fusion process can be represented by the following formula: Among them, X t H represents the sequence of environmental parameters input at time t. t-1 The hidden state (hidden feature) at time (t-1) is represented by σ, which represents the activation function; the feature fusion result is... Used to update the input state, feature fusion result To update the hidden state, after performing feature fusion in the above manner, the updated hidden state can be obtained. and the updated input status Furthermore, as illustrated in the above formula, a recurrent unit structure is adopted during feature fusion. The output of the previous recurrent unit is used as the input of the next recurrent unit connected to it. This feature fusion enhances the semantic relevance between the current input data and the hidden state output by the previous unit, preventing the recurrent unit from forgetting important features and improving the accuracy of the prediction results. In this embodiment, during feature fusion, after updating the hidden state and the input state, the activation function σ is used to update the hidden state. Updated input status A linear memory cell unit (C) in LSTM t-1 ), multiple different convolution kernels (w f w i w g w o The fusion process is as follows:

[0052]

[0053] Among them, f t For the forget gate of the loop unit, i t As the input gate of the recurrent unit, g t For the input modulation gate of the cyclic unit, o t b is the output gate of the loop unit. f b i b c b o These are pre-set biases of different values. tanh is an activation function. Through the above steps, during feature fusion, f is made... t Unimportant features in the forgetting memory cell unit make i t and g t The learning unit learns key features from the current input state, updates the memory cell unit based on the sum of the above two results, and then normalizes it using the tanh activation function before combining it with o. t The operation yields the output result. Among them, generating The process is shown in the following formula: C t =f t ⊙C t-1 +i t ⊙g t , Among them, C t This is the updated memory cell unit. Since the LSTM in this embodiment uses a self-attention mechanism, the output of the gated structure operation is... Then, three different convolution kernels (w) will be used. qw k w v Under the influence of ) , we obtain three vectors for the self-attention mechanism: Query(q) t Key(k) t ) and Value(V t When using a self-attention mechanism for computation, q t With k t The similarity score is obtained by performing multiplication, and then normalized by scaling and a softmax function to obtain the weight coefficient matrix, which is then compared with V. t Perform multiplication to obtain the final output H of the loop unit. t (i.e., feature fusion results). This step, through weighting, enhances the dust concentration prediction model's ability to extract global features from long sequences, thereby improving the accuracy of dust concentration prediction. This step can be expressed as the following formula: During feature fusion, the above steps are repeated, stacking two layers of recurrent units to increase network depth and improve the model's ability to extract high-order features from the input sequence. After obtaining the feature fusion result, in this embodiment, the result is converted into a sequence form through dimensionality reduction to obtain the target data sequence. The target data sequence contains the same amount of data as the environmental parameter sequence; that is, the target data sequence and the environmental parameter sequence have the same dimension. In fact, the target data sequence contains predicted values ​​of different types of environmental parameters. The dust concentration prediction value is then obtained by querying the set within the target data sequence. In the above embodiment, converting the feature fusion result into the target data sequence can be implemented in the fully connected layer (Dense layer) of the dust concentration prediction model. In the fully connected layer, the dimension of the feature fusion result is adjusted to 1 to obtain the final prediction sequence. The variable corresponding to the dust concentration in the prediction sequence is the value predicted by the dust concentration prediction model for the dust concentration in the field environment at the next detection time.

[0054] Step S206: Adjust the rotation speed of the target equipment according to the predicted dust concentration value. When the rotation speed of the target equipment is different, the exhaust intensity of the area to be tested is different. The target equipment is the equipment with air circulation function in the area to be tested.

[0055] After predicting the dust concentration value of the area to be detected at the next detection time in step S204, in step S206, the rotation speed determined based on the predicted dust concentration value is the rotation speed of the device with air circulation function (i.e., the target device) in the area to be detected at the next detection time. The target device reduces the dust concentration of the area to be detected by exhausting air into the area to be detected. The exhaust intensity of the target device into the area to be detected varies when it runs at different rotation speeds. Therefore, by adjusting the rotation speed of the target device, the technical effect of adjusting the dust concentration of the area to be detected can be achieved.

[0056] Optionally, adjusting the rotational speed of the target equipment based on the predicted dust concentration includes: obtaining the historical predicted dust concentration value corresponding to the previous detection time, and determining the predicted value of the control parameter based on the historical predicted dust concentration value and the predicted dust concentration value, wherein the control parameter is a parameter used to control the rotational speed of the target equipment; obtaining multiple historical deviation values ​​corresponding to multiple historical detection times before the current detection time, and determining the predicted deviation value based on the multiple historical deviation values, wherein the historical deviation value is the difference between the historical predicted dust concentration value and the historical measured dust concentration value; determining the target value of the control parameter based on the predicted value and the predicted deviation value, and adjusting the value of the control parameter to the target value.

[0057] After predicting the dust concentration of the area to be detected at the next detection time through the above embodiments, in this embodiment, the rotational speed of the target equipment is determined based on the predicted dust concentration. Specifically, the latest predicted dust concentration value is read, a PID control algorithm is used to calculate a result, and the rotational speed of the target equipment is adjusted using this result. In this embodiment, when adjusting the rotational speed of the target equipment based on the predicted dust concentration value, firstly, the predicted value u(t) of the control parameter of the rotational speed corresponding to the predicted dust concentration value c(t) needs to be determined according to the following formula, where the target equipment is currently rotating at the rotational speed corresponding to u(t).

[0058]

[0059] Where c(t-1) is the predicted dust concentration of the area to be detected at the previous detection time (i.e., the historical predicted dust concentration), T is the sampling time (i.e., the detection time), and K... p T is the proportionality coefficient. i T is the integral coefficient. dHere, is the differential coefficient, i represents the i-th detection time, and t is the total number of detection times. Further, to improve the accuracy of speed adjustment, the deviation value of the currently output parameter value (i.e., the predicted deviation value Δu(t)) is determined based on the deviation value between the predicted dust concentration at the predicted time (i.e., the current time) and the preset dust concentration value, as well as the deviation values ​​of the dust concentration at multiple historical detection times close to the predicted time (current detection time) (i.e., historical deviation values). The speed of the target device is then corrected using the predicted deviation value Δu(t). In this embodiment, the deviation value of the previously output parameter value (i.e., the predicted deviation value Δu(t)) is determined according to the following formula:

[0060]

[0061] Where e(t) is the deviation of the predicted dust concentration at time t (i.e., the prediction time / current time), e(t-1) is the deviation of the predicted dust concentration at the previous detection time (t-1) before the current time, and e(t-2) is the deviation of the predicted dust concentration at the two previous detection times (t-2) before the current time. Both e(t-1) and e(t-2) are historical deviation values. Finally, the sum of u(t) and Δu(t) is the final adjusted value of the control parameter of the target device (i.e., the target value). The control parameter can be a current percentage, and Δu(t) can be positive or negative.

[0062] According to some optional embodiments of this application, before extracting and processing high-order features of environmental parameters using a dust concentration prediction model, the method includes: classifying environmental parameters of the same type into a dataset; for each dataset, arranging the environmental parameters in order of detection time from early to late as a sequence to be processed; for each sequence to be processed, determining the value range based on the values ​​of the environmental parameters contained in the sequence; and performing numerical correction processing on the environmental parameters if the values ​​of the environmental parameters do not fall within the value range.

[0063] In the above embodiments, the collected environmental parameters are directly input into the dust concentration prediction model to predict the dust concentration at the next detection time. This can also achieve the technical effect of adjusting the rotation speed of the target equipment. However, if the environmental parameters are preprocessed before being input into the dust concentration prediction model, the accuracy of the prediction results can be improved. Therefore, in some embodiments, the environmental parameters can be preprocessed as follows: a preset number of environmental parameters are selected from the acquired environmental parameters, and the selected environmental parameters are processed using a sliding window method. For example, the latest 32 values ​​of five environmental parameters collected by sensors at the steel production site—temperature, humidity, air pressure, air velocity, and dust mass concentration—can be read from the database and processed using a sliding window. Before performing sliding window processing, environmental parameters belonging to the same category among the selected environmental parameters are grouped into a dataset. For each dataset, the environmental parameters are arranged in ascending order of their corresponding detection times to form a sequence (i.e., the sequence to be processed). The sliding window is then moved across each sequence in a preset order. Each time the window is moved, multiple environmental parameters included in the preset window are selected, and a value range is determined based on these multiple environmental parameters. Environmental parameters whose values ​​do not fall within the determined value range are corrected. Here, the preset window is the preset number of environmental parameters selected each time, and the value range can be based on the average value μ of these multiple environmental parameters. k and standard deviation σ k Commonly determined, for example, the range of values ​​can be (μ) k -3σ k μ k +3σ k The correction process can be to replace an environmental parameter whose value does not belong to the range with the average of multiple environmental parameters contained in a preset window. Taking a certain category of environmental parameter x as an example, for time t... n Collected value x n The sequence X is formed by selecting the historical observation values ​​from the previous n-1 consecutive time points. k ={x k1 ,x k2 ,…,x kn}, calculate its mean μ k and standard deviation σ k : If x n ∈(μ k -3σ k μ k +3σ k If x is determined, then x is determined. n This is a normal value. If x nIf the value falls outside this interval, it is considered an outlier and undergoes denoising and interpolation. The mean of the observed values ​​from the previous 5 time points is used as the current corrected observed value. All environmental parameters are preprocessed according to the above procedure, and then the data is normalized to the [0,1] interval, thus completing the preprocessing.

[0064] Through the above steps, the dust concentration prediction model can extract the temporal dimension and high-order features of the input environmental parameters using causal convolution and dilated convolution operations, solving the problems of low feature extraction efficiency and limited receptive field in existing network models. By adding a multiplication operation between the hidden state and the input state of the unit in the prediction process of the dust concentration prediction model, the problem of forgetting or failing to extract important features of dust concentration changes caused by the weak semantic relevance of the two in the temporal context during the gating operation of the recurrent unit is avoided, thus improving the accuracy of the prediction results. By adding a step of assigning weights to the operation results after the recurrent gating structure operation, the dust concentration prediction model's ability to extract global features of dust concentration changes is improved, further enhancing the accuracy of the prediction results.

[0065] Figure 5 This is a structural diagram of a device for adjusting rotational speed according to an embodiment of this application, as shown below. Figure 5 As shown, the device includes: an acquisition module 50, used to acquire environmental parameters of the area to be detected at multiple detection times, wherein the environmental parameters are parameters related to the dust concentration of the area to be detected, and the difference between any two adjacent detection times is a preset value; a processing module 52, used to extract and process the high-order features of the environmental parameters using a dust concentration prediction model to obtain a dust concentration prediction value, wherein the dust concentration prediction value is used to indicate the dust concentration of the area to be detected at the next detection time, the high-order features are time-series features of the environmental parameters, and the dust concentration prediction model is obtained by training a neural network model using historical environmental parameters of the area to be detected as training data; and an adjustment module 54, used to adjust the rotation speed of the target device according to the dust concentration prediction value, wherein the target device is a device with air circulation function in the area to be detected.

[0066] It should be noted that, Figure 5 Preferred embodiments of the shown examples can be found in [reference needed]. Figure 2 The relevant descriptions of the embodiments shown will not be repeated here.

[0067] Figure 6 This is a schematic diagram of a system for adjusting rotational speed. The method provided in this application embodiment can also be used for... Figure 6 The system shown, such as Figure 6As shown, the system includes the following hardware architecture capable of communication: a data acquisition sensor group for collecting environmental parameters, a database server and a high-performance computing server for data processing, and a field automation control cabinet for controlling equipment with air circulation functions. The data acquisition sensor group communicates with the database server via a 5G gateway. The sensor group includes a temperature sensor, a humidity sensor, a pressure sensor, a wind speed sensor, and a dust concentration detector. Each sensor is deployed in the dust-removing area of ​​the detection zone to collect temperature, humidity, pressure, wind speed, and dust concentration data. Each sensor is connected to the 5G gateway, which has corresponding ports to read sensor information and send it to the message broker server (MQTT Broker) at 6-minute intervals. The database server deploys a data subscription program and database service, subscribing to the data collected by the sensors and sending it to the MQTT Broker, parsing it, and storing it in the database on the server. The high-performance computing server deploys a dust concentration prediction model and a rotation speed mapping program. The high-performance computing server communicates with the automated control cabinet via a 5G gateway. It adjusts the speed of the ducted air conditioner (i.e., a device with air circulation function) through a programmable logic controller (PLC). In this embodiment, a PID algorithm is used to determine the values ​​of speed-related control variables and output current percentages based on the output increment. The PID control algorithm requires multiple parameters such as derivative coefficients, integral coefficients, and proportional coefficients. In this embodiment, the sampling time T in the PID algorithm is set to 180, the preset dust concentration value is set to 25, and the proportional coefficient (K...) is... p The integral coefficient (T) is set to 2.0. i The differential coefficient (T) is set to 0.8. d The value is set to 0.2. In this embodiment, data from actual steel plant production over three consecutive years, collected at a frequency of 6 minutes, is used to train the dust concentration prediction model. The data is divided into training, validation, and test sets at a ratio of 60%, 20%, and 20%, respectively. During the training phase, the input data is processed to obtain predicted values. The difference in the loss function is calculated by comparing the predicted values ​​with the actual values, thereby continuously updating the network parameters. The validation set is used to verify whether the model is overfitting. During the testing phase, the predictive performance of the trained model is tested based on the test set. In this embodiment, Mean Square Error (MAE), Root Mean Square Error (RMSE), and Coefficient of Determination (R-squared) are used to evaluate the accuracy of the model in predicting dust concentration. The calculation formulas for the three indicators are as follows: Where m is the total number of samples, i is the i-th sample, and x i Let be the actual dust concentration value of the i-th sample. For the i-th sample, predict the dust concentration value using the model. This represents the average dust concentration of the sample. Lower MAE and RMSE scores, and higher R-Squared scores, indicate better predictive performance of the model.

[0068] Figure 7 This is a comparison of metrics when multiple models are used to process the same test set. Figure 8 This involves comparing the dust concentration prediction results of multiple models based on the same validation set. To verify the dust concentration prediction model provided in this application's embodiments, the dust concentration prediction model from this application's embodiments, a Long Short-Term Memory network model combining convolutional neural networks (CNN-LSTM), a Long-Term and Short-Term Time Series Network model (LSTNet), a Recurrent Neural Network model combining residual connections and skip connections (RSC-GRU), and a deep learning-based prediction model (DeepAR) were processed on the same test set for comparative experiments. The scores of each model on the test set are shown below. Figure 7 As shown, according to Figure 7 It is understood that the dust concentration prediction model provided in the embodiments of this application can predict dust concentration more accurately, and therefore, adjusting the speed of the duct fan based on the dust concentration prediction model provided in the embodiments of this application is also more precise. Figure 8 As shown, the dust concentration prediction values ​​at multiple detection times output by multiple models are compared with the actual dust concentration values ​​(i.e., Ground Truth) at each detection time included in the validation set. The dust concentration prediction value predicted by the dust concentration prediction model (i.e., Proposed Model) provided in this application embodiment is closer to the actual dust concentration value.

[0069] This application also provides a non-volatile storage medium storing a computer program, wherein the device containing the non-volatile storage medium executes the above-mentioned method for adjusting the rotation speed by running the computer program.

[0070] The aforementioned non-volatile storage medium is used to store a program that performs the following functions: acquiring environmental parameters of the area to be detected at multiple detection times, wherein the environmental parameters are parameters related to the dust concentration of the area to be detected, and the difference between any two adjacent detection times is a preset value; extracting and processing high-order features of the environmental parameters using a dust concentration prediction model to obtain a dust concentration prediction value, wherein the dust concentration prediction value is used to indicate the dust concentration of the area to be detected at the next detection time, the high-order features are time-series features of the environmental parameters, and the dust concentration prediction model is obtained by training a neural network model using historical environmental parameters of the area to be detected as training data; adjusting the rotation speed of the target device according to the dust concentration prediction value, wherein different rotation speeds of the target device result in different exhaust intensities for the area to be detected, and the target device is a device with air circulation function in the area to be detected.

[0071] This application also provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor is configured to execute the above-described method for adjusting the rotation speed through the computer program.

[0072] The processor in the aforementioned electronic device is used to run a program that performs the following functions: acquiring environmental parameters of the area to be detected at multiple detection times, wherein the environmental parameters are parameters related to the dust concentration of the area to be detected, and the difference between any two adjacent detection times is a preset value; extracting and processing high-order features of the environmental parameters using a dust concentration prediction model to obtain a dust concentration prediction value, wherein the dust concentration prediction value is used to indicate the dust concentration of the area to be detected at the next detection time, the high-order features are time-series features of the environmental parameters, and the dust concentration prediction model is obtained by training a neural network model using historical environmental parameters of the area to be detected as training data; adjusting the rotation speed of the target device according to the dust concentration prediction value, wherein different rotation speeds of the target device result in different exhaust air intensities for the area to be detected, and the target device is a device with air circulation function in the area to be detected.

[0073] This application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the above-mentioned method for adjusting rotation speed.

[0074] It should be noted that the modules in the above-mentioned speed adjustment device can be program modules (e.g., a set of program instructions to implement a certain function) or hardware modules. For the latter, they can be in the following forms, but are not limited to these: each of the above modules is in the form of a processor, or the functions of each of the above modules are implemented by a processor.

[0075] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0076] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0077] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0078] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0079] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0080] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to related technologies, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0081] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for adjusting rotational speed, characterized in that, include: The environmental parameters of the area to be detected are obtained at multiple detection times, wherein the environmental parameters are related to the dust concentration of the area to be detected, and the difference between any two adjacent detection times is a preset value. A dust concentration prediction model is used to extract and process the high-order features of the environmental parameters to obtain the dust concentration prediction value. The dust concentration prediction value is used to indicate the dust concentration of the area to be detected at the next detection time. The high-order features are the time-related temporal features of the environmental parameters. The dust concentration prediction model is obtained by training a neural network model with the historical environmental parameters of the area to be detected as training data. The rotational speed of the target device is adjusted according to the predicted dust concentration value. This involves: acquiring the historical predicted dust concentration value corresponding to the previous detection time; determining the predicted value of a control parameter based on the historical predicted dust concentration value and the predicted dust concentration value; the control parameter being a parameter used to control the rotational speed of the target device; acquiring multiple historical deviation values ​​corresponding to multiple historical detection times prior to the current detection time; determining a predicted deviation value based on the multiple historical deviation values; the historical deviation value being the difference between the historical predicted dust concentration value and the historical measured dust concentration value; determining the target value of the control parameter based on the predicted value and the predicted deviation value; and adjusting the value of the control parameter to the target value. The control parameter includes a current percentage. Different rotational speeds of the target device result in different exhaust intensities for the area to be detected. The target device is a device with air circulation functionality in the area to be detected. Processing the higher-order features includes: Hidden features of the dust concentration detection model are obtained, wherein the hidden features are used to indicate the relationship between the environmental parameters and the detection time, and the hidden features are determined by the dust concentration prediction model based on the training data during the training process; the hidden features and the higher-order features are fused using a recurrent neural memory network included in the dust concentration detection model to obtain a feature fusion result, wherein the recurrent neural memory network contains multiple sequentially connected recurrent units, and the output of the previous recurrent unit is used as the input of the next recurrent unit during the feature fusion process; the feature fusion result is converted into a target data sequence, wherein the amount of data contained in the target data sequence is the same as the amount of data contained in the environmental parameter sequence, and the dust concentration value contained in the target data sequence is the predicted dust concentration value.

2. The method according to claim 1, characterized in that, The high-order features of the environmental parameters are extracted and processed using a dust concentration prediction model to obtain predicted dust concentration values, including: Determine the detection time corresponding to each of the environmental parameters; Arrange multiple environmental parameters of different types corresponding to the same detection time to obtain multiple environmental parameter sequences corresponding to multiple detection times, wherein the types of environmental parameters include: humidity, temperature, air pressure, air velocity, and dust concentration; The dust concentration prediction model performs convolution processing on multiple environmental parameter sequences in different neural network layers to obtain the higher-order features. The first neural network layer is used to extract the higher-order features, and the other neural network layers are used to process the higher-order features. Different neural network layers use convolution kernels of different dimensions. The dust concentration prediction value is determined based on the processing results of the higher-order features and multiple environmental parameter sequences.

3. The method according to claim 2, characterized in that, The dust concentration prediction model performs convolution processing on multiple environmental parameter sequences at different neural network layers to obtain the higher-order features, including: For each neural network layer, the detection time corresponding to each environmental parameter sequence is determined, and multiple environmental parameter sequences are arranged into a processing matrix in order from early to late according to the detection time, wherein each row of the processing matrix is ​​an environmental parameter sequence; For the Nth row of the matrix to be processed, a new matrix composed of the data from the first N rows of the matrix to be processed is convolved using a preset convolution kernel to obtain a first convolution result, wherein the preset convolution kernel is a matrix with the same dimension as the new matrix; The sum of multiple first convolution results is determined as the convolution sum, and the preset convolution kernel and multiple convolution sums are processed to obtain the higher-order features.

4. The method according to claim 3, characterized in that, The higher-order features are obtained by processing the preset convolution kernel and multiple first convolution results, including: The preset convolutional kernel is filled to obtain multiple other convolutional kernels with different dimensions; In the target other neural network layer, one of the other convolution kernels is used to perform convolution processing on the convolution and to obtain a second convolution result, wherein the target other neural network layer is the other neural network layer connected to the first neural network layer; According to the connection order of the remaining other neural network layers, the input data is convolved sequentially in different other neural network layers using other convolution kernels of different dimensions to obtain the target convolution result output by the last other neural network layer. In the convolution process, the input of the first other neural network is the second convolution result, and the output of the previous other neural network layer is used as the input data of the next other neural network layer. The higher-order features are determined based on the target convolution result and the environmental parameter sequence.

5. The method according to claim 1, characterized in that, Before extracting and processing the higher-order features of the environmental parameters using a dust concentration prediction model, the process includes: The environmental parameters of the same type are grouped into a single dataset; For each of the data sets, the environmental parameters are arranged in order from earliest to latest according to the detection time to form a sequence to be processed; For each of the sequences to be processed, the range of values ​​is determined based on the values ​​of the environmental parameters contained in the sequence to be processed; If the value of the environmental parameter is not within the specified range, the environmental parameter is corrected.

6. A device for adjusting rotational speed, characterized in that, include: The acquisition module is used to acquire environmental parameters of the area to be detected at multiple detection times, wherein the environmental parameters are parameters related to the dust concentration of the area to be detected, and the difference between any two adjacent detection times among the multiple detection times is a preset value. The processing module is used to extract and process high-order features of the environmental parameters using a dust concentration prediction model to obtain predicted dust concentration values. Specifically, it acquires hidden features of the dust concentration detection model, whereby the hidden features indicate the relationship between the environmental parameters and the detection time, and these hidden features are determined by the dust concentration prediction model based on training data during training. The module then uses a recurrent neural memory network included in the dust concentration detection model to perform feature fusion processing on the hidden features and the high-order features to obtain a feature fusion result. The recurrent neural memory network contains multiple sequentially connected recurrent units. During the fusion process, the output of the previous loop unit is used as the input of the next loop unit; the feature fusion result is converted into a target data sequence, wherein the amount of data contained in the target data sequence is the same as the amount of data contained in the environmental parameter sequence, and the dust concentration value contained in the target data sequence is the dust concentration prediction value; the dust concentration prediction value is used to indicate the dust concentration of the area to be detected at the next detection time, the higher-order feature is the time-related temporal feature of the environmental parameter, and the dust concentration prediction model is obtained by training the neural network model with the historical environmental parameters of the area to be detected as training data; An adjustment module is used to adjust the rotational speed of the target device according to the predicted dust concentration value. This includes: acquiring a historical predicted dust concentration value corresponding to the previous detection time; determining a predicted value for a control parameter based on the historical predicted dust concentration value and the predicted dust concentration value; acquiring multiple historical deviation values ​​corresponding to multiple historical detection times prior to the current detection time; determining a predicted deviation value based on the multiple historical deviation values; and determining a target value for the control parameter based on the predicted value and the predicted deviation value, and adjusting the value of the control parameter to the target value. The control parameter includes a current percentage, and the target device is a device with air circulation function in the area to be detected.

7. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores a computer program, wherein the device containing the non-volatile storage medium executes the method for adjusting the rotational speed as described in any one of claims 1 to 5 by running the computer program.

8. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the method for adjusting the rotational speed according to any one of claims 1 to 5 through the computer program.

9. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the method for adjusting the rotational speed as described in any one of claims 1 to 5.

Images