A Compressor Unit Operation Control Method Based on Neural Network Algorithm

Abstract

The present invention discloses an operation control method for a compressor unit based on a neural network algorithm. The type and number of air compressors are initially selected according to the compressed air consumption and the pressure demand value of the equipment with the highest air consumption pressure. According to the initially selected type and parameters of the air compressor, its power calculation mathematical model is obtained, a neural network model is established, and the model input variables are reasonably selected. The output quantities are the exhaust volume of the air compressor unit and the shaft power of each air compressor. By training and learning, a non-linear mapping relationship between the input and output quantities is established, which can reduce the pressure fluctuation of the compressed air pipe network and optimize the power consumption of the air compressor unit.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline transportation pressure control, and particularly to an operation control method for a compressor unit based on a neural network algorithm. Background Art

[0002] Compressed air, as an expensive secondary energy source, is widely used in various industrial fields. As the main high-energy-consuming equipment for compressed air, the energy consumption of air compressors accounts for 95% of the total power consumption of the entire air compressor station system. Calculated over a ten-year life cycle, its operating cost accounts for 78% of the total cost, making it a veritable power-consuming mainstay. With the country's increasingly high requirements for energy conservation and emission reduction, on the basis of ensuring the operating capacity of the compressed air system, choosing appropriate technologies to control the air compressor unit to work under the most efficient conditions has become a new energy-saving direction.

[0003] Due to the large number of gas-using equipment and gas-using points, and the fluctuation of gas consumption with the production load, there are sometimes situations where the instantaneous gas consumption is very large, which usually causes a large fluctuation in the pressure of the compressed air system pipeline network. All compressed air systems have a minimum pressure to ensure the normal operation of the system. If the system air supply pressure exceeds the minimum pressure, the system will operate normally. Continuing to increase the system air supply pressure will lead to an increase in the gas consumption and energy consumption of the system. To ensure that the system air supply always meets the normal operation of all production activities, generally, enterprises will increase the air supply pressure of the entire system, making the lowest point of pressure fluctuation higher than the pressure demand value of the highest gas-using pressure equipment. This results in the air supply pressure being higher than the normal gas-using pressure during other periods, causing an increase in system energy consumption. Therefore, how to timely and effectively control the operating conditions of the air compressor unit to maintain the pipeline network pressure of the system and ensure the normal operation of gas-using equipment is the current research direction. In view of the defects of the traditional methods for ensuring the pipeline network pressure of the compressed air system, it is necessary to propose an operation control scheme that can quickly respond to changes in the pipeline network pressure and efficiently optimize the air supply volume and pressure of the compressed air system. Summary of the Invention

[0004] The purpose of this part is to outline some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this part, as well as in the abstract and title of the specification of this application, to avoid obscuring the purpose of this part, the abstract, and the title, but such simplifications or omissions shall not be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the technical problems solved by the present invention are to reduce the pressure fluctuation of the compressed air pipeline network and optimize the power consumption of the air compressor unit.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: including,

[0008] Set the type and quantity of the air compressor according to the actual situation, and establish an operation control model for the air compressor unit;

[0009] According to the type and quantity of the air compressor unit, use a generalized regression neural network to establish a control mathematical model for the air compressor unit. Replace the inherent equation form with a probability density function. The generalized regression neural network is based on nonparametric kernel regression. Take the pressure change rate, air flow, temperature, humidity, network set pressure value, and input air volume of the compressor unit at different positions and times collected during the actual operation of the air supply system as sample data. Collect a large number of multi-group sample data. With the sample data as the posterior condition, after obtaining the joint probability density function between the independent variable and the dependent variable from the sample data, directly calculate the regression value of the dependent variable to the independent variable;

[0010] Each group of the above sample data is divided into training samples and verification samples. The training samples are mainly used to establish the non-linear relationship between the input and output of the generalized regression neural network, and the verification samples are mainly used to evaluate and determine the final value of the smoothing parameter σ.

[0011] As a preferred scheme of the operation control method for the compressor unit based on the neural network algorithm of the present invention, wherein: the selected air compressors include centrifugal air compressors and screw air compressors, and the types of centrifugal air compressors and screw air compressors are A i and B j , and the corresponding numbers of units of type A i and B j are X i and Y j units.

[0012] As a preferred scheme of the operation control method for the compressor unit based on the neural network algorithm of the present invention, wherein: when fine-tuning the air delivery volume is required, a screw air compressor is used;

[0013] When coarse-tuning the air delivery volume is required, a centrifugal air compressor is used.

[0014] As a preferred scheme of the operation control method for the compressor unit based on the neural network algorithm of the present invention, wherein: establish a mathematical model for the air compressor,

[0015] Select a single variable-frequency screw air compressor or centrifugal air compressor for energy consumption analysis of the compressor. The required driving work of the compressor can be expressed as:

[0016]

[0017] where W t is the required theoretical driving work of a single compressor, J; V iis the volume between the alveoli at the end of inhalation, m 3 ; p t is the pipeline pressure, Pa; V0 is the volume between the alveoli at the end of compression, m 3 ;

[0018] Theoretical driving power:

[0019]

[0020] Actual power consumption:

[0021]

[0022] η T = η i η m

[0023] where z is the number of teeth of the male rotor of the screw air compressor; n s is the rotational speed of the male rotor, r / min; η T is the isothermal efficiency of the air compressor; η i is the indicated efficiency of the air compressor; η m is the mechanical efficiency of the air compressor.

[0024] As a preferred scheme of the operation control method of the compressor unit based on the neural network algorithm described in the present invention, wherein: the change rate of the pressure values at different points of the fixed-length conveying pipeline is used as one of the input variables of the algorithm model, and by changing the input values of different variables, the required compressed air volume and the control scheme of the compressed air unit are output.

[0025] As a preferred scheme of the operation control method of the compressor unit based on the neural network algorithm described in the present invention, wherein: the change rate of the pressure values at different points is measured by setting pressure sensors at different positions of the conveying pipeline, and the required gas production volume of each air compressor under the current state is output by using the neural network algorithm model, and the start-stop and gas transmission volume of the compressor unit are controlled to achieve the required minimum pipeline gas transmission pressure.

[0026] As a preferred scheme of the operation control method of the compressor unit based on the neural network algorithm described in the present invention, wherein: a generalized regression neural network model is established,

[0027] replacing the inherent equation form with a probability density function, and after obtaining the joint probability density function between the independent variable and the dependent variable from the sample data with the sample data as the posterior condition, directly calculating the regression value of the dependent variable to the independent variable.

[0028] That is, the conditional mean is,

[0029]

[0030] For an unknown probability density function f(x,y), it can be obtained by non-parametric estimation from the observed samples of x and y:

[0031]

[0032] where X i , Y i are the sample observed values of the random variables x and y; δ is the smoothing parameter or kernel width;

[0033] Using to replace f(x,y) and substituting it into the above formula and exchanging the order of integration and summation, we can obtain

[0034] Substituting the above two formulas, we have:

[0035]

[0036] For the integral term in the above formula, using the property to simplify, we can obtain

[0037]

[0038] where the estimated value is the weighted average of all sample data values Y i , and the weight factor of each observed value Y i is the exponent of the square of the Euclid distance between the corresponding sample X i and X.

[0039] As a preferred scheme of the compressor unit operation control method based on the neural network algorithm described in the present invention, wherein: determining the value of the smoothing parameter σ in the generalized regression neural network algorithm,

[0040] Let the smoothing parameter increase incrementally by Δσ within a certain range (σ min , σ max );

[0041] In a set of sample data, excluding the verification samples, training the neural network with the remaining samples and testing with the verification samples, and the remaining samples are the training samples;

[0042] Calculating the absolute value of the error of the test samples using the constructed network model, that is, the prediction error;

[0043] Repeating in another set of sample data, excluding the verification samples, training the neural network with the training samples, testing with the verification samples, and calculating the absolute value of the error of the test samples using the constructed network model until all the training samples are used for testing once, obtaining the average value of the prediction errors, and taking it as the objective function for optimization, including,

[0044]

[0045] Lower limit σ of the smoothing parameter when using a Gaussian kernel min including

[0046]

[0047] where D min and ε are respectively the minimum Euclidean distance between training vectors and a number slightly larger than zero;

[0048] Upper limit σ of the smoothing factor max including

[0049]

[0050] where D max is the maximum Euclidean distance between input training vectors.

[0051] As a preferred solution of the compressor unit operation control method based on the neural network algorithm according to the present invention, wherein: initialization is performed, and the initialization is a learning process of training samples. When the training samples are determined, the corresponding network structure and the connection weights between each neuron are also determined accordingly. The training of the network is a process of determining the smoothing factor. The learning algorithm of the generalized regression neural network changes the smoothing factor during the training process and adjusts the transfer functions of each unit to obtain the best regression estimation result.

[0052] As a preferred solution of the compressor unit operation control method based on the neural network algorithm according to the present invention, wherein: multiple pressure sensors are used to measure the pressure changes at different position points of the gas pipeline. The pressure change rate, air flow rate, temperature, humidity, and network set pressure value at each point are used as the neural network input quantities, and the exhaust gas volume and power of each compressor are used as the output quantities.

[0053] Advantages of the present invention: The compressor unit operation control method based on the neural network algorithm provided by the present invention considers multiple input quantities as neural network input variables, detects the changes of multiple input quantities, predicts the changes of the output quantities, realizes the efficient and accurate prediction of the control model, obtains the optimal gas production strategy of the compressor unit under the determined input quantities through the non-linear relationship of the neural network while ensuring the pressure of the gas pipeline network, so as to reduce the system energy consumption. The output quantities are the gas production volume and power consumption of each compressor. Under the condition that the gas production volume is determined, the requirement of the lowest power consumption of the compressor unit needs to be met, and the control of the compressor unit is reasonably allocated according to the types and capabilities of different compressors.

[0054] Under the same working conditions, the estimated fluctuation range of the internal pressure difference of the gas pipeline network in this control model is about between 0.01 Mpa and 0.03 Mpa. Compared with the pressure difference fluctuation range (ΔP) of the traditional control model, which is about between 0.08 Mpa and 0.1 Mpa, the internal pressure value of the pipeline network will be effectively controlled within an acceptable range.

[0055] The reduction of the pressure difference fluctuation range can effectively reduce the actual supply gas pressure in the pipeline (generally slightly higher than the maximum gas demand pressure in the system gas-using equipment), reduce the power consumption of the air compressor unit by 5-7%, and has certain economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. Among them:

[0057] Figure 1 It is the generalized regression neural network structure diagram of the compressor unit operation control method based on the neural network algorithm.

[0058] Figure 2 It is the prediction flow block diagram of the generalized regression neural network algorithm of the compressor unit operation control method based on the neural network algorithm.

[0059] Figure 3 It is the schematic diagram of the compressor unit operation control of the compressor unit operation control method based on the neural network algorithm.

[0060] Figure 4 It is the comparison diagram of the pressure fluctuation value of the compressor pipeline network of the compressor unit operation control method based on the neural network algorithm.

[0061] Figure 5 It is the comparison diagram of the shaft power of the compressor of the compressor unit operation control method based on the neural network algorithm. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention with reference to the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0063] In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0064] Secondly, as used herein, an "embodiment" or "embodiments" refers to specific features, structures, or characteristics that may be included in at least one implementation of the present invention. The phrase "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that excludes other embodiments.

[0065] The present invention will be described in detail in conjunction with schematic diagrams. When describing the embodiments of the present invention in detail, for the convenience of explanation, the cross-sectional views showing the device structure will be enlarged locally out of the general scale, and the schematic diagrams are only examples and should not limit the scope of protection of the present invention herein. In addition, in actual production, three-dimensional spatial dimensions including length, width, and depth should be included.

[0066] At the same time, in the description of the present invention, it should be noted that the orientation or positional relationships indicated by terms such as "upper, lower, inner, and outer" are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, the terms "first, second, or third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0067] Unless otherwise clearly defined and limited in the present invention, the terms "mounted, connected, and coupled" shall be construed broadly. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may also be a mechanical connection, an electrical connection, or a direct connection, or may be indirectly connected through an intermediate medium, or may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0068] Embodiment 1

[0069] Referring to Figures 1 to 3 , a first embodiment of the present invention provides a method for controlling the operation of a compressor unit based on a neural network algorithm, including:

[0070] The compressed air system mainly includes three parts: production, pressure regulation, and transportation. Among them, the production part is the output source of compressed air and also the main energy-consuming part. The production link is mainly composed of a compressor unit; the transportation part mainly transports the generated compressed air to the gas-using ports, mainly including gas storage tanks, pipe networks, and valves, etc.; the pressure regulation link is mainly achieved by a pressure regulating station.

[0071] The basic working process of the system is as follows: The gas is compressed in the compressor and then output at a higher pressure. After being regulated by the pressure regulating station, it is transported through the pipeline network to the gas-using end at the required pressure. When the gas production is greater than the gas consumption, the gas will be stored in the gas storage tank. On the contrary, when the gas consumption is greater than the gas production, the gas stored in the gas storage tank will be used for supplementation.

[0072] In this embodiment, two types of air compressor units are used, namely the variable frequency screw air compressor and the centrifugal compressor. Taking these two compressors as the objects, a unified mathematical calculation model is established. The specific steps are as follows:

[0073] S1: Determine the actual compression model, derive the calculation equation for the polytropic index of the gas during the compression process from the gas state equation, and calculate the polytropic index of the gas using the inlet and outlet temperatures and the inlet and exhaust pressures:

[0074]

[0075]

[0076] Among them, P i is the suction pressure, Pa; P0 is the final compression pressure, Pa; T i is the suction temperature, K; T0 is the exhaust temperature, K; n is the polytropic index of the gas.

[0077] S2: Derive the theoretical compression work required for air compression, that is, the theoretical compression work required for the air compressor to compress the gas per unit volume includes:

[0078]

[0079] Among them, W t is the theoretical driving work required for a single compressor, J; V i is the volume between the tooth grooves at the end of suction, m 3 ; p t is the pipeline network pressure, Pa; V0 is the volume between the tooth grooves at the end of compression, m 3 ;

[0080] S3: Calculate the theoretical driving power during the loaded operation of the air compressor according to the theoretical driving work of the air compressor, including:

[0081]

[0082] Considering the power consumption of the air compressor in actual situations includes:

[0083]

[0084] η T = η i η m

[0085] Among them, z is the number of teeth of the male rotor of the screw air compressor; n s is the rotational speed of the male rotor, r / min; η T is the isothermal efficiency of the air compressor; η i is the indicated efficiency of the air compressor; η m is the mechanical efficiency of the air compressor.

[0086] S4: Select the appropriate number of air compressor units. According to the actual situation and experience, judge the models and quantities of the screw air compressor units and the centrifugal air compressor units, and require them to be able to meet the minimum to maximum gas consumption requirements in the gas pipeline. And there are at least three or more combination schemes for different gas transmission volume requirements for the control system to compare and select, so as to obtain the best control strategy and meet the ultimate goal of minimum power consumption.

[0087] Furthermore, collect the operation data of the air compressor units during the actual operation process, establish an air compressor operation database, including learning samples and evaluation samples, that is, the network input variable X = [x1…x i , x i+1 …x i+j , x i+j+1 …x 2i+2j , x 2i+2j+1 , x 2i+2j+2 , x 2i+2j+3 , x 2i+2j+4 , x 2i+2j+5 , x 2i+2j+6 , and the output variable Y = [P1, P2, P3, P4…P i+j+3

[0088] Among them: in the input variable, x1 is the pressure drop between the pressure sensor 3 and the position i1…x i is the pressure drop between the pressure sensor 3 and the position ii; x i+1 is the pressure drop between the pressure sensor 3 and the position j1…x i+j is the pressure drop between the pressure sensor 3 and the position jj; x i+j+1 is the pressure change rate of the pressure sensor i1…x 2i+j is the pressure change rate of the pressure sensor ii; x 2i+j+1 is the pressure change rate of the pressure sensor j1…x 2i+2j is the pressure change rate of the pressure sensor jj; x 2i+2j+1 is the pressure drop between the pressure sensors 1 and 2; x 2i+2j+2 is the pressure drop between the pressure sensors 2 and 3; x 2i+2j+3 is the pressure change rate of the pressure sensor 1; x 2i+2j+4 is the pressure change rate of the pressure sensor 2; x 2i+2j+5 is the pressure change rate of the pressure sensor 3; x 2i+2j+3 ​is the flow rate in the gas pipeline; x 2i+2j+4 is the gas temperature in the pipeline; x 2i+2j+5 is the gas humidity in the pipeline; x 2i+2j+6 is the pressure set value in the pipeline.

[0089] Among the output variables, P1, P2, P3, P4…P i+j are the output powers of the screw air compressors 1, 2…i and the centrifugal air compressors 1, 2…j respectively.

[0090] In the actual production process, each set of variables X corresponds to a unique variable Y.

[0091] Based on the above, a running control network model of the air compression system's compressor unit is established using a generalized regression neural network. The running data of the compressor unit during the actual operation is used as the training sample database and verification sample database of the generalized regression neural network. The running control model of the compressor unit based on the generalized regression neural network algorithm is defined and set up to realize the basic establishment of the network model.

[0092] The theoretical basis of the generalized regression neural network is non - linear kernel regression analysis. The regression analysis of the non - independent variable y with respect to the independent variable x is actually to calculate the y with the maximum probability value.

[0093] Let the joint probability density function of the random variables x and y be f(x, y). Given the observed value of x as X, the regression of y with respect to X, that is, the conditional mean includes:

[0094]

[0095] For the unknown probability density function f(x, y), it can be obtained by non - parametric estimation from the observed samples of x and y:

[0096]

[0097] where, X i , Y i are the sample observed values of the random variables x and y; δ is the smoothing parameter (or kernel width);

[0098] Using to replace f(x, y) and substituting it into the above formula and exchanging the order of integration and summation, we can get

[0099] Substituting the above two formulas, we have:

[0100]

[0101] For the integral term in the above formula, using the property to simplify, we can get

[0102]

[0103] In the above formula, the estimated value is the weighted average of all sample observations Y i , and the weight factor for each observation Y i is the exponent of the square of the Euclid distance between the corresponding sample X i and X.

[0104] Using an artificial neural network to establish an operation control model for a compressor unit regards the entire compressor unit as a "black box", and uses the self-organization, self-adaptation, and self-learning capabilities of the artificial neural network to learn its input-output characteristics in order to obtain model network parameters (weight value threshold matrix, smoothing parameter, etc.) that can more realistically represent the load characteristics. Through the learning sample database composed of the input variable X and the output variable Y collected, the generalized regression neural network model is trained and learned. The purpose of network training is to generate a suitable weight value matrix and threshold vector.

[0105] Preferably, determine the smoothing parameter σ value in the generalized regression neural network algorithm, so that the smoothing parameter σ increases by Δσ in (σ min , σ max ). During the learning process, the actually collected parameters are divided into training samples and verification samples. The generalized neural network model is constructed with the training samples and compared with the evaluation samples to obtain the error value between the predicted value and the actual value. As many samples as possible are involved in the training calculation, and the effective value of the obtained error sequence is used to evaluate the network performance. After repeated training, the smoothing parameter corresponding to the minimum error is used as the final value and used for subsequent calculations.

[0106] Calculate the lower limit σ min of the smoothing parameter, including:

[0107]

[0108] where D min and ε are respectively the minimum Euclidean distance between training vectors and a number slightly larger than zero.

[0109] Calculate the upper limit σ max of the smoothing parameter, including:

[0110]

[0111] where D max is the maximum Euclidean distance between input training vectors.

[0112] Determine the smoothing parameter σ value:

[0113] (1) Let the smoothing parameter increase by an increment Δσ within a certain range (σ min , σ max)Incremental change;

[0114] (2) Train a neural network in the learning samples and test it with the validation samples;

[0115] (3) Calculate the absolute value of the error of the test samples using the constructed network model, and call it the prediction error;

[0116] (4) Repeat steps (2) and (3) until all training samples have been used for testing once, calculate the average value of the prediction errors, and use it as the objective function for optimization, including:

[0117]

[0118] Establish a control model for the compressor unit based on the generalized neural network using the above method.

[0119] This embodiment comprehensively considers different types of compressor units, uses the pressure change rate and pressure fluctuation conditions at different positions in the gas pipeline as the characteristic dependent variables of the compressed air demand in the gas pipeline, and takes them as the main input variables of the generalized regression neural network model. By establishing an operation control model for the compressor unit based on the generalized regression neural network algorithm, calculate the compressed air demand in the gas pipeline, and finally output the gas production volume of the air compressor unit and the optimal output power of each compressor to obtain the minimum power consumption, and obtain a control method with the minimum output power under the condition of meeting the gas transmission volume.

[0120] Embodiment 2

[0121] This embodiment provides an operation control method for a compressor unit based on a neural network algorithm. The pressure drop between different positions in the gas pipeline, the pressure change rate, the flow rate in the gas pipeline, the output control quantity, the temperature, the humidity, and the network pressure set value are used as the input values of the compressor unit operation control model. Multiple input quantities are used as neural network input variables. By detecting the changes of multiple input quantities, predict the changes of the output quantity, realize the efficient and accurate prediction of the control model, and obtain the optimal gas production strategy of the compressor unit under the determined input quantity through the non-linear relationship of the neural network to reduce the system energy consumption. Control the compressor unit according to the types and capabilities of different compressors to meet the gas production demand. Its advantages are:

[0122] ① Under the model of this embodiment, by using two air compressors with different quantities to supplement the gas inside the system, it can meet the gas demand under different working conditions. Based on the neural network model, it has good prediction and rapid replenishment capabilities for pressure changes and gas demand;

[0123] ② Refer to Figure 4, the pressure difference fluctuation range (△P) generated by the traditional gas supply network with the fluctuation of production load is about 0.08Mpa-0.1Mpa. Under the same working conditions, the pressure difference fluctuation range (△P) of the gas network in this control model is estimated to be about 0.01Mpa-0.03Mpa, and the pressure value inside the network will be effectively controlled within an acceptable range;

[0124] ③Reference Figure 5 The reduction of the pressure difference fluctuation range can effectively reduce the actual gas supply pressure in the pipeline (generally slightly higher than the maximum gas demand pressure in the system gas equipment), reduce the power consumption of the air compressor unit by 5-7%, and have certain economic benefits;

[0125] ④ Under the condition of meeting the pressure demand, the pressure difference fluctuates little, which can effectively reduce the repeated impact of gas pressure fluctuations on the equipment. The service life of each part of the system equipment will be increased, and it also has high economic efficiency.

[0126] It should be appreciated that embodiments of the present invention may be implemented or enforced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program may be implemented in an assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed ASIC for this purpose.

[0127] Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions that may be executed by one or more processors.

[0128] Further, the method can be implemented in any type of computing platform operably connected to a suitable one, including but not limited to personal computers, minicomputers, mainframes, workstations, network or distributed computing environments, separate or integrated computer platforms, or communicating with charged particle tools or other imaging devices, etc. Aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into the computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it can be read by a programmable computer and, when read by the storage medium or device, can be used to configure and operate the computer to perform the processes described herein. In addition, the machine-readable code, or portions thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the above-described steps in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention also includes the computer itself. A computer program can be applied to input data to perform the functions described herein, thereby transforming the input data to generate output data stored in non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the transformed data represents physical and tangible objects, including a specific visual depiction of the physical and tangible objects generated on a display.

[0129] As used in this application, the terms "component", "module", "system", etc. are intended to refer to a computer-related entity, which can be hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to: a process running on a processor, a processor, an object, an executable file, a thread in execution, a program, and / or a computer. As an example, an application running on a computing device and the computing device can both be components. One or more components can exist in a process and / or thread in execution, and the components can be located in one computer and / or distributed between two or more computers. In addition, these components can execute from various computer-readable media having various data structures thereon. These components can communicate in a local and / or remote procedure manner through signals such as in accordance with one or more data packets (e.g., data from one component that interacts with another component in a local system, a distributed system, or communicates with other systems via a network such as the Internet in a signal manner).

[0130] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for controlling the operation of a compressor unit based on a neural network algorithm, characterized in that: including, setting the type and quantity of air compressors according to the actual situation and establishing an operating control model for the air compressor unit; establishing a control mathematical model for the air compressor unit using a generalized regression neural network according to the type and quantity of the air compressor unit, replacing the inherent equation form with a probability density function. The generalized regression neural network is based on non-parametric kernel regression. Using the pressure change rate, air flow, temperature, humidity, network set pressure value, and air input volume of the compressor unit at different positions and times collected during the actual operation of the air supply system as sample data, collecting a large number of groups of sample data. After obtaining the joint probability density function between the independent variable and the dependent variable from the sample data with the sample data as the posterior condition, directly calculating the regression value of the dependent variable to the independent variable; each group of the above sample data is divided into training samples and verification samples. The training samples are used to establish the input-output non-linear relationship of the generalized regression neural network, and the verification samples are used to evaluate and determine the final value of the smoothing parameter σ; using multiple pressure sensors to measure the pressure changes at different positions of the gas transmission pipeline, taking the pressure change rate, air flow, temperature, humidity, and network set pressure value at each point as the input of the neural network, and taking the exhaust volume and power of each compressor as the output; 2. The compressor unit operation control method based on the neural network algorithm according to claim 1, characterized in that: The selected air compressors include centrifugal air compressors and screw air compressors, and the types of centrifugal air compressors and screw air compressors are A i and B j respectively, and the corresponding numbers of units of type A i and B j are a i and b j units respectively.

3. The compressor unit operation control method based on the neural network algorithm according to claim 2, characterized in that: if fine adjustment of the air delivery volume is required, a screw air compressor is used; if coarse adjustment of the air delivery volume is required, a centrifugal air compressor is used; 4. The compressor unit operation control method based on neural network algorithm according to claim 1 or 2, characterized in that: establishing a mathematical model of the air compressor; selecting a single variable-frequency screw air compressor or a centrifugal air compressor for energy consumption analysis of the compressor. The driving work required by the compressor can be expressed as: Among them, W t is the theoretical driving work required for a single compressor, J; V i is the volume between the tooth grooves at the end of suction, m 3 ; p t is the pipeline network pressure, Pa; V0 is the volume between the tooth grooves at the end of compression, m 3 ; P i is the suction pressure, Pa; P0 is the final compression pressure, Pa, and n is the polytropic index of the gas; Theoretical driving power: Actual power consumption: η T = η i η m Among them, z is the number of teeth of the male rotor of the screw air compressor; n s is the rotational speed of the male rotor, r / min; η T is the isothermal efficiency of the air compressor; η i is the indicated efficiency of the air compressor; η m is the mechanical efficiency of the air compressor.

5. The operation control method of the compressor unit based on the neural network algorithm according to claim 1, characterized in that: taking the pressure value change rate at different points of the fixed-length transmission pipeline as one of the input variables of the algorithm model, and outputting the required compressed air volume and the control scheme of the compressed air unit through the change of the input values of different variables; 6. The operating control method for a compressor unit based on a neural network algorithm according to claim 5, characterized in that: measuring the pressure value change rate at different points by setting pressure sensors at different positions of the transmission pipeline, using the neural network algorithm model to output the required gas production volume of each air compressor under the current state, controlling the start and stop of the compressor unit and the air delivery volume to achieve the required minimum pipeline gas transmission pressure; 7. The operation control method of the compressor unit based on the neural network algorithm according to claim 1 or 5, characterized in that: establishing a generalized regression neural network model; replacing the inherent equation form with a probability density function, and directly calculating the regression value of the dependent variable to the independent variable after obtaining the joint probability density function between the independent variable and the dependent variable from the sample data with the sample data as the posterior condition; that is, the conditional mean is for the unknown probability density function f(x, y), it can be obtained by non-parametric estimation from the observed samples of x and y: where X i , Y i are the sample observations of the random variables x and y; σ is the smoothing parameter or kernel width; Use to replace f(x,y), substitute it into the above formula and exchange the order of integration and summation, we can get substituting the above two formulas, we get: For the integral term in the above equation, using the property to simplify, we can obtain Among them, the estimated value is the weighted average of all sample values Y i , and the weight factor of each sample value Y i is the exponent of the square of the Euclid distance between the corresponding sample X i and X.

8. The compressor unit operation control method based on the neural network algorithm according to claim 7, characterized in that: determining the value of the smoothing parameter σ in the generalized regression neural network algorithm; Let the smoothing parameter increase incrementally by Δσ within a certain range (σ min , σ max ); in a group of sample data, excluding the verification samples, training the neural network with the remaining samples, and testing with the verification samples. The remaining samples are the training samples; calculating the absolute value of the error of the test samples using the constructed network model, that is, the prediction error; Repeat. In another set of sample data, excluding the validation samples, use the training samples to train the neural network, use the validation samples for testing, and use the constructed network model to calculate the absolute value of the error of the test samples until all the training samples have been used for testing once. Then, find the average value of the prediction errors and use it as the objective function for optimization, including Lower limit σ of the smoothing parameter when using a Gaussian kernel min including where D min and ε are respectively the minimum Euclidean distance between training vectors and a number slightly larger than zero; Upper limit σ of the smoothing parameter max including where D max is the maximum Euclidean distance between the input training vectors.

9. The compressor unit operation control method based on neural network algorithm according to claim 8, characterized in that: Perform initialization. The initialization is the learning process of the training samples. Once the training samples are determined, the corresponding network structure and the connection weights between neurons are also determined. The training of the network is the process of determining the smoothing parameter. The learning algorithm of the generalized regression neural network changes the smoothing parameter during the training process and adjusts the transfer functions of each unit to obtain the best regression estimation result.

Images