Process industry energy consumption optimization system and method based on neural network

Abstract

The present invention relates to the field of data processing technology, and in particular to a process industry energy consumption optimization system and method based on a neural network, comprising a data acquisition module, a sound fluctuation regularity acquisition module, a sound authenticity acquisition module and a power energy consumption optimization module. The system acquires the sound fluctuation regularity of a production area at each acquisition moment, and acquires the machine sound authenticity of the production area at each acquisition moment according to the sound fluctuation regularity and the difference between sound data at each acquisition moment and other acquisition moments; the machine sound authenticity is input into an LSTM-GRU neural network, and power energy consumption of all production areas in an automated production line is optimized; thereby, more precise control of the operating state of a power system in a process industry is achieved, and the generation of waste electricity can be effectively reduced, and power energy consumption in the process industry can be reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a process industry energy consumption optimization system and method based on a neural network. Background Art

[0002] In the field of process industries, energy consumption accounts for an important part of production costs in the industrial production process, and energy consumption has always been a key factor affecting production costs and environmental sustainability; the industrial energy consumption intelligent optimization system can use advanced information technology and intelligent algorithms to conduct real-time monitoring, predictive analysis and optimal regulation of energy use in the industrial production process.

[0003] In the prior art, in order to avoid the generation of waste electricity, enterprises usually install certain control devices to control the opening or closing of some auxiliary equipment. For example, in large-scale industrial production processes, in order to reduce energy waste and reduce the generation of waste electricity, the lighting system where the industrial equipment is located is usually designed by installing sound sensors. When the machine in the area where the sensor is located starts running, the light turns on normally, and when the machine equipment stops running, the light turns off. However, because there are too many other interfering sounds in the environment in the industrial production environment, these interfering sounds may cause instability of the control system and produce erroneous operations, thereby causing loss of electricity energy. Summary of the invention

[0004] The purpose of the present invention is to provide a process industry energy consumption optimization system and method based on neural network, aiming to more accurately control the operating status of the power system in the process industry and reduce the power energy consumption in the process industry.

[0005] To achieve the above-mentioned purpose, in a first aspect, the present invention provides a process industry energy consumption optimization system based on a neural network, comprising a data acquisition module, a sound fluctuation regularity acquisition module, a sound authenticity acquisition module and an electric energy consumption optimization module, wherein the data acquisition module, the sound fluctuation regularity acquisition module, the sound authenticity acquisition module and the electric energy consumption optimization module are connected in sequence;

[0006] The data acquisition module is used to obtain the sound data sequence of all production areas in the automated production line;

[0007] The sound fluctuation regularity acquisition module is used to compare the differences between the clustering results of the sound data sequences of adjacent production areas, and obtain the sound fluctuation value of the production area at each collection time; obtain a number of data windows of the production area; obtain the sound fluctuation regularity of the production area at each collection time by comparing the differences between the sound fluctuation values ​​at the same position at the collection time in different data windows, and the differences between the sound fluctuation values ​​at different collection times in the data windows to which each collection time belongs;

[0008] The sound authenticity acquisition module acquires the authenticity of the machine sound at each collection moment in the production area based on the regularity of the sound fluctuation and the difference of the sound data between each collection moment and other collection moments;

[0009] The power energy consumption optimization module is used to input the authenticity of machine sounds into the LSTM-GRU neural network to optimize the power energy consumption of all production areas in the automated production line.

[0010] In a second aspect, the present invention further provides a process industry energy consumption optimization method based on a neural network, which is applied to the process industry energy consumption optimization system based on a neural network as described in the first aspect above, and is characterized in that it comprises the following steps:

[0011] Acquire the sound data sequence of all production areas in the automated production line through the data acquisition module;

[0012] The sound fluctuation regularity acquisition module compares the differences between the clustering results of the sound data sequences of adjacent production areas and obtains the sound fluctuation value of the production area at each acquisition time;

[0013] Acquire several data windows of the production area through the sound fluctuation regularity acquisition module;

[0014] The sound fluctuation regularity acquisition module obtains the sound fluctuation regularity of the production area at each collection moment through the differences between the sound fluctuation values ​​at the same location in different data windows at the collection moment, and the differences between the sound fluctuation values ​​at different collection moments in the data window to which each collection moment belongs;

[0015] The sound authenticity acquisition module acquires the authenticity of the machine sound at each collection time in the production area based on the regularity of the sound fluctuation and the difference between the sound data at each collection time and other collection times;

[0016] The power consumption optimization module inputs the authenticity of machine sounds into the LSTM-GRU neural network to optimize the power consumption of all production areas in the automated production line.

[0017] The sound fluctuation regularity acquisition module compares the differences between the clustering results of the sound data sequences of adjacent production areas and acquires the specific method of the sound fluctuation value of the production area at each collection moment:

[0018] For any production area, the sound data sequence of the production area is input into a two-dimensional coordinate system with the acquisition time as the horizontal coordinate and the sound data as the vertical coordinate to obtain a number of sound data points of the production area;

[0019] All sound data points are clustered using an iterative self-organizing analysis clustering algorithm to obtain several clusters. Based on the number of sound data points in the cluster to which the sound data point at each collection moment belongs, the sound concentration degree of the production area at each collection moment is obtained.

[0020] In an automated production line, production areas adjacent to the production area on the left and right are all used as reference production areas of the production area. For any reference production area of ​​the production area, the product of the sound concentration of the reference production area at the collection time and the sound data at the collection time is recorded as the sound impact factor of the reference production area at the collection time. The average of the sound impact factors of all reference production areas of the production area at the collection time is recorded as the reference area sound impact degree of the production area at the collection time. The absolute value of the difference between the sound impact factor and the reference area sound impact degree is taken as the sound fluctuation value of the production area at the collection time.

[0021] Among them, the specific method of obtaining several data windows of the production area;

[0022] For any production area, all sound data points in the production area are fitted using the least square method to obtain the sound data fluctuation curve of the production area;

[0023] A data point sequence consisting of all data points between any two minimum values ​​in the sound data fluctuation curve is used as a data window of the production area.

[0024] The specific method of obtaining the regularity of sound fluctuations in the production area at each collection moment is as follows:

[0025] For any production area, in the data window at each collection time, the difference between the sound fluctuation values ​​at different collection times is obtained to obtain the fluctuation difference factor of the production area at each collection time;

[0026] Through the differences between the sound fluctuation values ​​at the same location in different data windows at the collection time, the fluctuation regularity factor of the production area at each collection time is obtained;

[0027] The sum of the fluctuation regularity factor of the production area at each collection moment and the fluctuation difference factor of the production area at each collection moment is taken as the sound fluctuation regularity of the production area at each collection moment.

[0028] The process industry energy consumption optimization system based on neural network of the present invention obtains the sound data sequences of all production areas in the automated production line through the data acquisition module; the sound fluctuation regularity acquisition module compares the differences between the clustering results of the sound data sequences of adjacent production areas to obtain the sound fluctuation value of the production area at each acquisition time; obtains a plurality of data windows of the production area; obtains the sound fluctuation regularity of the production area at each acquisition time through the differences between the sound fluctuation values ​​at the same position at the acquisition time in different data windows, and the differences between the sound fluctuation values ​​at different acquisition times in the data windows to which each acquisition time belongs; the sound authenticity acquisition module obtains the sound fluctuation regularity of the production area at each acquisition time based on the sound fluctuation regularity and the differences between the sound data at each acquisition time and other acquisition times The system obtains the regularity of sound fluctuations in the production area at each collection moment, and obtains the authenticity of machine sounds in the production area at each collection moment according to the regularity of sound fluctuations and the difference in sound data between each collection moment and other collection moments; the authenticity of machine sounds is input into the LSTM-GRU neural network, and the power energy consumption of all production areas in the automated production line is optimized; thereby, more precise control of the operating status of the power system in the process industry can be achieved, which can effectively reduce the generation of waste electricity and reduce power consumption in the process industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0030] Figure 1 It is a structural block diagram of the process industry energy consumption optimization system based on neural network of the present invention;

[0031] Figure 2This is a characteristic relationship flow chart of the process industry energy consumption optimization system based on neural network of the present invention.

[0032] Figure 3 The present invention is a flowchart of the process industry energy consumption optimization method based on neural network.

[0033] Figure 4 It is a flowchart of a specific method in which a sound fluctuation regularity acquisition module compares the differences between clustering results of sound data sequences of adjacent production areas and acquires the sound fluctuation value of the production area at each acquisition moment.

[0034] Figure 5 It is a flowchart of the specific method of obtaining several data windows of the production area.

[0035] Figure 6 It is a flowchart of a specific method for obtaining the regularity of sound fluctuations in the production area at each collection moment by determining the differences between the sound fluctuation values ​​at the same location in different data windows at the collection moment, and the differences between the sound fluctuation values ​​at different collection moments in the data window to which each collection moment belongs.

[0036] In the figure: 1-data acquisition module, 2-sound fluctuation regularity acquisition module, 3-sound authenticity acquisition module, 4-power energy consumption optimization module. DETAILED DESCRIPTION

[0037] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0038] See also Figure 1 In a first aspect, the present invention provides a process industry energy consumption optimization system based on a neural network, comprising a data acquisition module 1, a sound fluctuation regularity acquisition module 2, a sound authenticity acquisition module 3 and an electric energy consumption optimization module 4, wherein the data acquisition module 1, the sound fluctuation regularity acquisition module 2, the sound authenticity acquisition module 3 and the electric energy consumption optimization module 4 are connected in sequence;

[0039] The data acquisition module 1 is used to obtain the sound data sequence of all production areas in the automated production line;

[0040] The sound fluctuation regularity acquisition module 2 is used to compare the differences between the clustering results of the sound data sequences of adjacent production areas, and obtain the sound fluctuation value of the production area at each collection time; obtain a number of data windows of the production area; obtain the sound fluctuation regularity of the production area at each collection time by comparing the differences between the sound fluctuation values ​​at the same position at the collection time in different data windows, and the differences between the sound fluctuation values ​​at different collection times in the data windows to which each collection time belongs;

[0041] The sound authenticity acquisition module 3 acquires the authenticity of the machine sound at each collection time in the production area based on the regularity of the sound fluctuation and the difference of the sound data between each collection time and other collection times;

[0042] The power consumption optimization module 4 is used to input the authenticity of machine sounds into the LSTM-GRU neural network to optimize the power consumption of all production areas in the automated production line.

[0043] In an embodiment of the present invention, the data acquisition module 1 is used to acquire the sound data sequences of all production areas in the automated production line; the sound fluctuation regularity acquisition module 2 compares the differences between the clustering results of the sound data sequences of adjacent production areas to acquire the sound fluctuation values ​​of the production areas at each acquisition moment; a plurality of data windows of the production areas are acquired; the sound fluctuation regularity of the production areas at each acquisition moment is acquired by the differences between the sound fluctuation values ​​at the same position at the acquisition moment in different data windows, and the differences between the sound fluctuation values ​​at different acquisition moments in the data windows to which each acquisition moment belongs; the sound authenticity acquisition module 3 acquires the sound authenticity of the production areas based on the sound fluctuation regularity and the differences between the sound data at each acquisition moment and other acquisition moments. The authenticity of the machine sounds in the production area at each collection moment; the power energy consumption optimization module 4 inputs the authenticity of the machine sounds into the LSTM-GRU neural network, and optimizes the power energy consumption of all production areas in the automated production line. The system obtains the regularity of sound fluctuations in the production area at each collection moment, and obtains the authenticity of the machine sounds in the production area at each collection moment according to the regularity of sound fluctuations and the difference in sound data between each collection moment and other collection moments; the authenticity of the machine sounds is input into the LSTM-GRU neural network, and the power energy consumption of all production areas in the automated production line is optimized; thereby achieving more precise control of the operating status of the power system in the process industry, which can effectively reduce the generation of waste electricity and reduce power consumption in the process industry.

[0044] See also Figure 2-Figure 6In a second aspect, the present invention further provides a process industry energy consumption optimization method based on a neural network, which is applied to the process industry energy consumption optimization system based on a neural network as described in the first aspect above, and is characterized in that it comprises the following steps:

[0045] S1 obtains the sound data sequence of all production areas in the automated production line through the data acquisition module 1;

[0046] In an embodiment of the present invention, the data acquisition module 1 uses sound sensors to collect sound data from machines and equipment in all production areas in an automated production line. For any production area, one collection moment is every 1 second, and the sound data of the production area is collected each time, for a total of 2 hours; the sound data at each collection moment is used as the sound data sequence of the production area.

[0047] S2 Sound fluctuation regularity acquisition module 2 compares the differences between the clustering results of the sound data sequences of adjacent production areas, and obtains the sound fluctuation value of the production area at each acquisition time;

[0048] In an embodiment of the present invention, because when the machinery and equipment in the production area where the sensor is located is running, the machinery and equipment in the adjacent production areas in the automated production line are also running and producing, so the sound received by the sensor in the production area will inevitably be affected by the interference of the sound generated by the operation of the equipment in other adjacent production areas. That is, when the sensor collects sound data, its sound contains the sound influence of the machinery and equipment in other production areas, so as to calculate the sound fluctuation value of the production area at each collection moment.

[0049] Specific method:

[0050] S21, for any production area, using the acquisition time as the horizontal coordinate and the sound data as the vertical coordinate, inputting the sound data sequence of the production area into a two-dimensional coordinate system to obtain a plurality of sound data points of the production area;

[0051] S22 clusters all the sound data points using an iterative self-organizing analysis clustering algorithm to obtain a number of clusters, and obtains the sound concentration of the production area at each collection moment based on the number of sound data points in the cluster to which the sound data point at each collection moment belongs;

[0052] In the embodiment of the present invention, the method for obtaining the sound concentration of the production area at each collection time is: The ratio between the number of sound data points in the cluster to which the sound data point belongs and the number of all sound data points in the production area is used as the production area in the The sound concentration at the time of collection.

[0053] The specific formula is:

[0054]

[0055] In the formula, Indicates that the production area is The sound concentration at each collection moment; Indicates the production area The number of sound data points in the cluster to which the sound data point belongs; Represents the number of all sound data points in the production area.

[0056] S23 In the automated production line, the production areas adjacent to the production area on the left and right are taken as reference production areas of the production area. For any reference production area of ​​the production area, the product of the sound concentration of the reference production area at the collection time and the sound data at the collection time is recorded as the sound impact factor of the reference production area at the collection time. The average of the sound impact factors of all the reference production areas of the production area at the collection time is recorded as the reference area sound impact degree of the production area at the collection time. The absolute value of the difference between the sound impact factor and the reference area sound impact degree is taken as the sound fluctuation value of the production area at the collection time.

[0057] In the embodiment of the present invention, the specific formula is:

[0058]

[0059] In the formula, Indicates that the production area is The sound fluctuation value at the time of collection; Indicates that the production area is The sound concentration at each collection moment; The first in the sound data sequence representing the production area The sound data at the time of collection; The number of all reference production areas representing the production area; Indicates the production area The reference production area is in The sound concentration at each collection moment; Indicates the production area The first The sound data at the time of collection; Indicates taking the absolute value.

[0060] in, Indicates that the production area is The sound impact factor at each collection moment; It indicates the degree of influence of the machines and equipment in the adjacent production areas on the sound of the production area; the iterative self-organizing analysis clustering algorithm is the existing technology.

[0061] S3 obtains several data windows of the production area through the sound fluctuation regularity acquisition module 2;

[0062] Specific method:

[0063] S31, for any production area, fitting all sound data points of the production area using the least square method to obtain a sound data fluctuation curve of the production area;

[0064] S32 uses a data point sequence consisting of all data points between any two minimum values ​​in the sound data fluctuation curve as a data window for the production area.

[0065] S4 Sound fluctuation regularity acquisition module 2 obtains the sound fluctuation regularity of the production area at each collection moment through the difference between the sound fluctuation values ​​at the same location in different data windows at the collection moment, and the difference between the sound fluctuation values ​​at different collection moments in the data window to which each collection moment belongs;

[0066] Specific method:

[0067] S41 For any production area, in the data window belonging to each collection time, the difference between the sound fluctuation values ​​at different collection times is obtained to obtain the fluctuation difference factor of the production area at each collection time;

[0068] In the embodiment of the present invention, in the sound data sequence of the production area, All the acquisition moments in the data window to which the acquisition moment belongs are taken as the The reference collection time of each collection time;

[0069] For The first collection time The reference collection time is the production area at the The sound fluctuation value at the collection time is related to the production area at the The absolute value of the difference between the sound fluctuation values ​​at the reference collection time is recorded as The fluctuation difference value of the reference collection time; The inverse proportional value of the cumulative sum of the fluctuation difference values ​​of all reference collection times at the collection time is used as the production area at the The volatility difference factor at each collection moment.

[0070] The specific formula is:

[0071]

[0072] In the formula, Indicates that the production area is The volatility difference factor at each collection moment; Indicates The number of all reference collection moments for a collection moment; Indicates that the production area is The sound fluctuation value at the time of collection; Indicates that the production area is The first collection time The sound fluctuation value at the reference collection moment; Indicates taking the absolute value; represents an exponential function with a natural constant as the base, and the embodiment adopts Model to present inverse proportional relationship and normalization, As the input of the model, the implementer can choose the inverse proportional function and the normalization function according to the actual situation.

[0073] S42 obtains the fluctuation regularity factor of the production area at each collection time by comparing the differences between the sound fluctuation values ​​at the same position in different data windows at the collection time;

[0074] In the embodiment of the present invention, The sequence number of each acquisition moment in the corresponding data window is recorded as ; In each data window of the production area, The collection moments are all taken as The target collection time for each collection time;

[0075] For The first collection time At the target collection time, the production area is The sound fluctuation value at the collection time is related to the production area at the The absolute value of the difference in sound fluctuation values ​​at the target collection time is recorded as The floating difference value of the target collection time; The inverse proportional value of the cumulative sum of the floating difference values ​​of all target collection times at the collection time is used as the production area at the The fluctuation regularity factor at each collection moment.

[0076] The specific formula is:

[0077]

[0078] In the formula, Indicates that the production area is The fluctuation regularity factor at each collection moment; Indicates The number of all target collection moments for a collection moment; Indicates that the production area is The sound fluctuation value at the time of collection; Indicates that the production area is The first collection time The sound fluctuation value at the reference collection moment; Indicates taking the absolute value; represents an exponential function with a natural constant as the base, and the embodiment adopts Model to present inverse proportional relationship and normalization, As the input of the model, the implementer can choose the inverse proportional function and the normalization function according to the actual situation.

[0079] S43 takes the sum of the fluctuation regularity factor of the production area at each collection moment and the fluctuation difference factor of the production area at each collection moment as the sound fluctuation regularity of the production area at each collection moment.

[0080] In the embodiment of the present invention, the specific formula is:

[0081]

[0082] In the formula, Indicates that the production area is The regularity of sound fluctuations at each acquisition moment; Indicates that the production area is The volatility difference factor at each collection moment; Indicates that the production area is The fluctuation regularity factor at each collection moment.

[0083] S5 Sound authenticity acquisition module 3 acquires the authenticity of machine sound at each collection time in the production area based on the regularity of sound fluctuations and the difference of sound data between each collection time and other collection times;

[0084] In the embodiment of the present invention, because the distance between the sensor in the production area and the machine equipment in the production area is relatively close and the sensor is fixed, and the distance between the sensor and the machine equipment is fixed, for the sensor, the volume of the sound of the running machine equipment collected by the sensor is basically unchanged or changes little; while other mechanical equipment in the area where the sensor is not located, because the distance from the production area is relatively far, the sound generated by the vibration of the machine equipment during operation will be attenuated after transmission, and the intensity of the sound will become smaller. Therefore, calculating the difference between the sound data at each sampling moment and the sound data at the historical sampling moments can indicate the possibility that the sound at the sampling moment is generated by the machine equipment in the area where the sensor is located.

[0085] According to the difference of sound data between each collection moment and other collection moments, the similarity of sound intensity in the production area at each collection moment is obtained;

[0086] Specifically, in the sound data sequence of the production area, All collection times other than the first collection time are recorded as the The comparison collection time of the collection time; for the The first collection time At the comparison collection time, the production area is The sound data at the collection time and the production area at the The absolute value of the difference between the sound data at the time of comparison and collection is recorded as The sound difference value of the comparison acquisition time; The inverse proportional value of the mean of the sound difference values ​​of all the comparison collection moments at the collection moment is used as the production area at the The similarity of sound intensity at each acquisition moment.

[0087] For any production area, place the production area in The regularity of sound fluctuations at the time of collection is related to the production area at the The normalized value of the product of the sound intensity similarity at the collection time is used as the production area at the The authenticity of the machine sound at the time of collection.

[0088] The specific formula is:

[0089]

[0090] In the formula, Indicates that the production area is The authenticity of the machine sound at the time of collection; Indicates that the production area is The regularity of sound fluctuations at each acquisition moment; Indicates that the production area is Similarity of sound intensity at each acquisition moment; represents the linear normalization function.

[0091] The specific method for obtaining the similarity of the sound intensity of the production area at each acquisition time is as follows: In the sound data sequence of the production area, All collection times other than the first collection time are recorded as the The comparison collection time of the collection time;

[0092] For The first collection time At the comparison collection time, the production area is The sound data at the collection time and the production area at the The absolute value of the difference between the sound data at the time of comparison and collection is recorded as The sound difference value of the comparison acquisition time; The inverse proportional value of the mean of the sound difference values ​​of all the comparison collection moments at the collection moment is used as the production area at the The similarity of sound intensity at each acquisition moment.

[0093] The specific formula is:

[0094]

[0095] In the formula, Indicates that the production area is Similarity of sound intensity at each acquisition moment; Indicates The number of all comparison collection moments for a collection moment; Indicates that the production area is The sound data at the time of collection; Indicates that the production area is The first collection time The sound data at the time of comparison collection; Indicates taking the absolute value; represents an exponential function with a natural constant as the base, and the embodiment adopts Model to present inverse proportional relationship and normalization, As the input of the model, the implementer can choose the inverse proportional function and the normalization function according to the actual situation.

[0096] The specific method for obtaining the similarity of the sound intensity of the production area at each acquisition time is as follows: In the sound data sequence of the production area, All collection times other than the first collection time are recorded as the The comparison collection time of the collection time;

[0097] For The first collection time At the comparison collection time, the production area is The sound data at the collection time and the production area at the The absolute value of the difference between the sound data at the time of comparison and collection is recorded as The sound difference value of the comparison acquisition time; The inverse proportional value of the mean of the sound difference values ​​of all the comparison collection moments at the collection moment is used as the production area at the The similarity of sound intensity at each acquisition moment.

[0098] The specific formula is:

[0099]

[0100] In the formula, Indicates that the production area is Similarity of sound intensity at each acquisition moment; Indicates The number of all comparison collection moments for a collection moment; Indicates that the production area is The sound data at the time of collection; Indicates that the production area is The first collection time The sound data at the time of comparison collection; Indicates taking the absolute value; represents an exponential function with a natural constant as the base, and the embodiment adopts Model to present inverse proportional relationship and normalization, As the input of the model, the implementer can choose the inverse proportional function and the normalization function according to the actual situation.

[0101] The S6 power consumption optimization module 4 inputs the authenticity of machine sounds into the LSTM-GRU neural network to optimize the power consumption of all production areas in the automated production line.

[0102] In the embodiment of the present invention, a threshold parameter is preset , wherein this embodiment is based on This example is described as an example, and this embodiment is not specifically limited. Depends on the specific implementation situation;

[0103] For any production area, if the mean value of the authenticity of the machine sound in the production area at all acquisition times is greater than or equal to the threshold parameter , then the type of the production area is recorded as a working production area; if the mean of the authenticity of the machine sound in the production area at all acquisition times is less than the threshold parameter , then the type of the production area is recorded as a non-working production area;

[0104] For any production area in the automated production line, the type of the production area is input into the trained neural network to obtain power energy consumption optimization measures for the production area.

[0105] The neural network used in this embodiment is LSTM-GRU, and the method for obtaining the data set for training the neural network is:

[0106] The specific method of artificially setting the power energy consumption optimization measures is as follows: if the type of the production area is a working production area, the power energy consumption optimization measures of the production area are set to: keep the power system of the production area in an open state; if the type of the production area is a non-working production area, the power energy consumption optimization measures of the production area are set to: keep the power system of the power production area in a disconnected state; this setting result is recorded as a label of the power energy consumption optimization measures; the power energy consumption optimization measures and their corresponding labels constitute a data set; the neural network is trained using the data set, and the loss function used in the training process is a cross loss function; wherein the specific training process is a well-known content of the neural network.

[0107] What is disclosed above is only a preferred embodiment of the process industry energy consumption optimization system and method based on neural network of the present invention. Of course, it cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiments are implemented, and equivalent changes made according to the claims of the present invention are still within the scope of the invention.

Claims

1. A process industry energy consumption optimization system based on neural network, characterized in that: It includes a data acquisition module, a sound fluctuation regularity acquisition module, a sound authenticity acquisition module and a power energy consumption optimization module, wherein the data acquisition module, the sound fluctuation regularity acquisition module, the sound authenticity acquisition module and the power energy consumption optimization module are connected in sequence; The data acquisition module is used to obtain the sound data sequence of all production areas in the automated production line; The sound fluctuation regularity acquisition module is used to compare the differences between the clustering results of the sound data sequences of adjacent production areas, and obtain the sound fluctuation value of the production area at each collection time; obtain a number of data windows of the production area; obtain the sound fluctuation regularity of the production area at each collection time by comparing the differences between the sound fluctuation values ​​at the same position at the collection time in different data windows, and the differences between the sound fluctuation values ​​at different collection times in the data windows to which each collection time belongs; The sound authenticity acquisition module acquires the authenticity of the machine sound at each collection moment in the production area based on the regularity of the sound fluctuation and the difference of the sound data between each collection moment and other collection moments; The power consumption optimization module is used to input the authenticity of machine sounds into the LSTM-GRU neural network to optimize the power consumption of all production areas in the automated production line; The sound fluctuation regularity acquisition module compares the differences between the clustering results of the sound data sequences of adjacent production areas, and obtains the specific method of the sound fluctuation value of the production area at each collection moment: For any production area, the sound data sequence of the production area is input into a two-dimensional coordinate system with the acquisition time as the horizontal coordinate and the sound data as the vertical coordinate to obtain a number of sound data points of the production area; All sound data points are clustered using an iterative self-organizing analysis clustering algorithm to obtain several clusters. Based on the number of sound data points in the cluster to which the sound data point at each collection moment belongs, the sound concentration degree of the production area at each collection moment is obtained. In an automated production line, production areas adjacent to the production area on the left and right are all used as reference production areas of the production area. For any reference production area of ​​the production area, the product of the sound concentration of the reference production area at the collection time and the sound data at the collection time is recorded as the sound impact factor of the reference production area at the collection time. The average of the sound impact factors of all reference production areas of the production area at the collection time is recorded as the reference area sound impact degree of the production area at the collection time. The absolute value of the difference between the sound impact factor and the reference area sound impact degree is taken as the sound fluctuation value of the production area at the collection time.

2. A method for optimizing energy consumption in process industries based on neural networks, applied to a system for optimizing energy consumption in process industries based on neural networks as claimed in claim 1, characterized in that: The following steps are involved: Acquire the sound data sequence of all production areas in the automated production line through the data acquisition module; The sound fluctuation regularity acquisition module compares the differences between the clustering results of the sound data sequences of adjacent production areas and obtains the sound fluctuation value of the production area at each acquisition time; Acquire several data windows of the production area through the sound fluctuation regularity acquisition module; The sound fluctuation regularity acquisition module obtains the sound fluctuation regularity of the production area at each collection moment through the differences between the sound fluctuation values ​​at the same location in different data windows at the collection moment, and the differences between the sound fluctuation values ​​at different collection moments in the data window to which each collection moment belongs; The sound authenticity acquisition module acquires the authenticity of the machine sound at each collection time in the production area based on the regularity of the sound fluctuation and the difference between the sound data at each collection time and other collection times; The power consumption optimization module inputs the authenticity of machine sounds into the LSTM-GRU neural network to optimize the power consumption of all production areas in the automated production line.

3. The process industry energy consumption optimization method based on neural network according to claim 2, characterized in that: The specific method of obtaining several data windows of the production area; For any production area, all sound data points in the production area are fitted using the least square method to obtain the sound data fluctuation curve of the production area; A data point sequence consisting of all data points between any two minimum values ​​in the sound data fluctuation curve is used as a data window of the production area.

4. The process industry energy consumption optimization method based on neural network according to claim 2, characterized in that: The specific method of obtaining the regularity of sound fluctuations in the production area at each collection moment is as follows: For any production area, in the data window at each collection time, the difference between the sound fluctuation values ​​at different collection times is obtained to obtain the fluctuation difference factor of the production area at each collection time; Through the differences between the sound fluctuation values ​​at the same location in different data windows at the collection time, the fluctuation regularity factor of the production area at each collection time is obtained; The sum of the fluctuation regularity factor of the production area at each collection moment and the fluctuation difference factor of the production area at each collection moment is taken as the sound fluctuation regularity of the production area at each collection moment.

Images