Injection molding machine energy-saving control method and system based on multi-source data

Abstract

The application relates to the technical field of energy-saving control, and discloses an injection molding machine energy-saving control method and system based on multi-source data. The method comprises the following steps: collecting voltage and current data and network connection relationship information of multiple injection molding machines, and preprocessing to obtain a device operation state data set; a dynamic correlation graph between devices is constructed according to the data set, and the energy consumption interaction relationship strength is calculated; if the strength exceeds a preset threshold, a high-strength interaction path is extracted, and a potential chain reaction path set is obtained by fusing network information; the voltage and current fluctuation trend is predicted, and a device group with excessively high load in the production peak period is judged; an energy load distribution matrix is obtained based on the device group, and the power parameter range that needs to be adjusted is determined; if the parameter distribution is uneven, balanced power adjustment vectors are obtained by optimizing the load distribution; and the injection molding machine control instruction is updated according to the vectors. The method can solve the problem of uneven load of multiple injection molding machines, and improve the energy utilization efficiency and the device operation stability.

Description

Technical Field

[0001] This invention relates to the field of energy-saving control technology for injection molding machines, and in particular to an energy-saving control method and system for injection molding machines based on multi-source data fusion and dynamic parameter optimization. Background Technology

[0002] As a core production equipment in the manufacturing industry, the energy load control of injection molding machines in multi-machine parallel operation scenarios is of great significance for ensuring production continuity and improving energy utilization efficiency. Achieving dynamic balance and collaborative optimization of the energy load of multiple injection molding machines is the core requirement of current injection molding machine operation management.

[0003] In one existing technology, the voltage and current data of a single injection molding machine are collected by an energy consumption sensor. The collected data is compared with a preset energy consumption threshold. When the data exceeds the threshold, the power parameters of the single machine are adjusted independently to achieve energy consumption control of a single machine.

[0004] However, existing technologies only monitor and adjust individual devices independently, failing to consider the network connections and energy consumption interactions between multiple injection molding machines. This results in an inability to capture the dynamic energy consumption impact between devices, leading to uneven energy load distribution when multiple injection molding machines operate in parallel. Some machines experience instability due to excessive load, while others operate inefficiently due to insufficient load. In summary, existing technologies result in low overall operational stability and energy efficiency for injection molding machine clusters. Summary of the Invention

[0005] This invention provides an energy-saving control method and system for injection molding machines based on multi-source data, in order to solve the problems of low overall operational stability and low energy utilization efficiency of injection molding machine groups caused by existing technologies.

[0006] In a first aspect, the present invention provides an energy-saving control method for injection molding machines based on multi-source data, comprising: The voltage and current data and network connection information of multiple injection molding machines are collected and preprocessed to obtain a dataset of equipment operating status. Based on the equipment operating status dataset, a dynamic correlation diagram between equipment is constructed, and the strength of the energy consumption interaction relationship between each injection molding machine is calculated. If the energy consumption interaction strength exceeds a preset energy consumption interaction strength threshold, then high-intensity interaction paths are extracted, and the network connection relationship information is fused to obtain a set of potential chain reaction paths. Based on the set of potential chain reaction paths, predict future voltage and current fluctuation trends and identify equipment groups with excessive load during peak production periods. Based on the equipment group with excessive load during the peak production period, obtain the current energy load distribution matrix, analyze the load evolution characteristics in the time dimension, and determine the range of equipment power parameters that need to be adjusted; If the range of power parameters of the equipment to be adjusted is unevenly distributed, the energy load allocation scheme is optimized to obtain a balanced power adjustment vector; Based on the balanced power adjustment vector, control commands for multiple injection molding machines are generated and updated.

[0007] Preferably, the step of collecting voltage and current data and network connection information from multiple injection molding machines, and preprocessing them to obtain a data set of equipment operating status data includes: By using sensors on multiple injection molding machines, the voltage and current data of each injection molding machine, as well as the network connection information between the injection molding machines, are collected in real time to generate a raw dataset with time dimension markings. Wavelet transform is performed on the voltage and current data in the original dataset to obtain a preliminary clean dataset; If the voltage fluctuation amplitude in the preliminary clean data exceeds the preset voltage fluctuation amplitude threshold, then the voltage data is subjected to Fourier transform processing to analyze the frequency distribution and determine the characteristics of the small voltage instability signal. By integrating the preliminary cleaning dataset, the characteristics of the minor voltage instability signal, and the network connectivity information, the device operating status dataset is obtained.

[0008] Preferably, the step of constructing a dynamic correlation diagram between devices based on the device operating status dataset and calculating the strength of energy consumption interaction relationships between each injection molding machine includes: Extract energy consumption data and network connection information of each injection molding machine from the equipment operation status dataset, and construct graph nodes corresponding to each injection molding machine; Feature extraction is performed on the graph nodes to obtain a graph node feature set; the graph node features include the average power and energy consumption fluctuation range of the injection molding machine; If the feature difference between any graph node in the graph node feature set and its adjacent graph nodes exceeds a preset feature difference threshold, then the energy consumption-related data of the adjacent graph nodes are weighted and aggregated, and the graph node features are updated to obtain an updated graph node feature set. Based on the updated graph node feature set and the network connection relationship information, the initial connection relationship between each graph node is determined, and a dynamic inter-device association graph is constructed. Based on the dynamic relationship diagram between the devices, the interaction strength score between each pair of injection molding machines is calculated; the interaction strength score is the energy consumption interaction strength between each injection molding machine.

[0009] Preferably, if the energy consumption interaction strength exceeds a preset energy consumption interaction strength threshold, then high-intensity interaction paths are extracted, and the network connection relationship information is fused to obtain a set of potential chain reaction paths, including: When the energy consumption interaction strength exceeds the preset energy consumption interaction strength threshold, the paths whose interaction strength scores meet the energy consumption interaction strength threshold requirements are extracted from the dynamic association graph between devices to obtain a set of high-intensity interaction paths. By integrating the network connection relationship information, the set of high-intensity interaction paths is traversed. If the network connection relationship meets the preset connection stability standard, potential chain reaction paths are selected to obtain a set of potential chain reaction paths.

[0010] Preferably, the step of predicting future voltage and current fluctuation trends and identifying equipment groups with excessive load during peak production periods based on the set of potential chain reaction paths includes: Based on the set of potential chain reaction paths, voltage and current data during peak production periods are extracted from the equipment operation status dataset, and time series analysis is performed to obtain voltage and current fluctuation trend terms. Calculate the probability that the voltage and current fluctuation trend terms exceed a preset fluctuation threshold, and determine the injection molding machine group whose probability exceeds the preset probability threshold as the equipment group with excessive load during peak production period.

[0011] Preferably, the step of obtaining the current energy load allocation matrix based on the equipment group with excessive load during peak production periods, analyzing the load evolution characteristics over time, and determining the range of equipment power parameters that need to be adjusted includes: Based on the equipment group with excessive load during peak production periods, obtain the energy load data of each injection molding unit within a preset time period, and construct the current energy load allocation matrix. Time series analysis is performed on the load data in the current energy load distribution matrix to extract the load change trend of each injection molding unit and obtain a set of load evolution characteristics; If the load change amplitude of any injection molding machine in the load evolution feature set exceeds the preset load change threshold, then the voltage data of the target injection molding machine is subjected to correlation analysis to calculate the voltage instability probability. Based on the voltage instability probability and the current power data of the target injection molding unit, the relationship between power and load change is fitted to determine the range of equipment power parameters that need to be adjusted.

[0012] Preferably, if the range of power parameters of the equipment to be adjusted is unevenly distributed, the energy load allocation scheme is optimized to obtain a balanced power adjustment vector, including: If the range of power parameters of the equipment to be adjusted does not meet the preset equalization standard, then the initial power distribution data is extracted from the current energy load allocation matrix; The initial power distribution data is iteratively optimized, the topology of the dynamic association graph between devices is dynamically updated, and the interaction strength score between each injection molding machine is calculated after each iteration. The power allocation is adjusted according to the interaction intensity score until power distribution data that meets the preset equalization standard is obtained, and the power distribution data is converted into an equalized power adjustment vector.

[0013] Preferably, the step of generating and updating control commands for multiple injection molding machines based on the balanced power adjustment vector includes: Based on the balanced power adjustment vector, the target power parameters of each injection molding machine are obtained; Determine if each injection molding unit is overloaded. If an overload is found, adjust the target power parameters of that injection molding unit. Based on the adjusted target power parameters, control commands are generated for multiple injection molding machines, and the control commands are sent to the corresponding injection molding machines to update the operating parameters of the injection molding machines.

[0014] Secondly, the present invention provides an energy-saving control system for injection molding machines based on multi-source data, comprising: The data acquisition and preprocessing module is used to acquire voltage and current data and network connection information of multiple injection molding machines, and preprocess the voltage and current data and network connection information to obtain a data set of equipment operating status. The correlation graph construction and strength calculation module is used to construct a dynamic correlation graph between devices based on the device operating status dataset and calculate the strength of the energy consumption interaction relationship between each injection molding machine. The chain path extraction module is used to extract high-intensity interaction paths and fuse the network connection relationship information to obtain a set of potential chain reaction paths if the energy consumption interaction relationship strength exceeds a preset energy consumption interaction strength threshold. The fluctuation prediction and judgment module is used to predict future voltage and current fluctuation trends based on the set of potential chain reaction paths, and to determine the equipment groups with excessive load during peak production periods. The load matrix processing module is used to obtain the current energy load distribution matrix based on the equipment group with excessive load during the peak production period, analyze the load evolution characteristics in the time dimension, and determine the range of equipment power parameters that need to be adjusted. The power optimization module is used to optimize the energy load allocation scheme to obtain a balanced power adjustment vector if the power parameter range of the equipment to be adjusted is unevenly distributed. The control command update module is used to generate and update control commands for multiple injection molding machines based on the balanced power adjustment vector.

[0015] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the energy-saving control method for injection molding machines based on multi-source data as described above.

[0016] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the energy-saving control method for injection molding machines based on multi-source data described above.

[0017] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention avoids the limitations of a single data dimension by collecting and preprocessing the voltage and current data and network connection information of multiple injection molding machines, and ensures the accuracy of the equipment operation status dataset. Based on this dataset, a dynamic correlation diagram between the equipment is constructed and the energy consumption interaction strength is calculated, which can accurately capture the dynamic energy consumption correlation between the equipment and solve the problem of neglecting the interaction between the equipment in the prior art.

[0018] (2) By extracting high-intensity interaction paths, predicting voltage and current fluctuation trends and judging equipment groups with excessive load, this invention can identify the risk of abnormal energy consumption during peak production periods in advance; and by optimizing the energy load distribution scheme through genetic algorithms, it can achieve balanced adjustment of power parameters, effectively improve the uneven load distribution of multiple injection molding machines, and improve the overall energy utilization efficiency.

[0019] (3) By dynamically updating the equipment association topology and control commands, the present invention can adapt to the changes in the operating status of the injection molding machine in real time, avoid the control lag problem caused by fixed parameters, significantly improve the operating stability of the injection molding machine group, and reduce the probability of chain failures caused by voltage instability. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the energy-saving control method for injection molding machines based on multi-source data provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the energy-saving control system for injection molding machines based on multi-source data provided in the second embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Reference Figure 1 The first embodiment of the present invention provides an energy-saving control method for injection molding machines based on multi-source data, comprising the following steps: S1: Collect voltage and current data and network connection information of multiple injection molding machines, and preprocess them to obtain equipment operating status dataset; S2, Based on the equipment operating status dataset, construct a dynamic correlation diagram between equipment and calculate the energy consumption interaction strength between each injection molding machine; S3, if the energy consumption interaction strength exceeds the preset energy consumption interaction strength threshold, then extract the high-intensity interaction path and fuse the network connection relationship information to obtain a potential chain reaction path set; S4. Based on the set of potential chain reaction paths, predict future voltage and current fluctuation trends and identify equipment groups with excessive load during peak production periods. S5. Based on the equipment group with excessive load during the peak production period, obtain the current energy load distribution matrix, analyze the load evolution characteristics in the time dimension, and determine the range of equipment power parameters that need to be adjusted. S6. If the range of power parameters of the equipment to be adjusted is unevenly distributed, the energy load allocation scheme is optimized to obtain a balanced power adjustment vector. S7. Based on the balanced power adjustment vector, generate and update control commands for multiple injection molding machines.

[0023] In step S1, voltage and current data and network connection information of multiple injection molding machines are collected and preprocessed to obtain a data set of equipment operating status, including: S11 uses sensors on multiple injection molding machines to collect voltage and current data of each injection molding machine in real time, as well as network connection information between the injection molding machines, to generate a raw dataset with time dimension markings. S12, perform wavelet transform processing on the voltage and current data in the original dataset to obtain a preliminary clean dataset; S13, if the voltage fluctuation amplitude in the preliminary cleaning dataset exceeds the preset voltage fluctuation amplitude threshold, then the voltage data is subjected to Fourier transform processing to analyze the frequency distribution and determine the characteristics of the small voltage instability signal. S14, the device operating status dataset is obtained by fusing the preliminary cleaning dataset, the characteristics of the small voltage instability signal, and the network connection relationship information.

[0024] In step S11, the voltage data, current data, and network connection information between each injection molding machine are collected in real time by sensors on multiple injection molding machines to generate an original dataset containing time dimension markers.

[0025] It should be noted that the sensors include a voltage sensor, a current sensor, and a network status sensor; the voltage and current sensors are installed at the power input of the injection molding machine, and the sampling frequency is once every 500ms. The voltage data is in volts (V), and the current data is in amperes (A); the network status sensor is installed at the Ethernet communication interface of the injection molding machine to detect the connection status between injection molding machines in real time, with 1 indicating a stable connection and 0 indicating an interrupted connection; the time dimension is marked by the timestamp of the sampling time, in the format YYYY-MM-DDHH:MM:SS.ms; the raw dataset is stored in the format of timestamp, i.e., [injection molding machine number: voltage value, current value, network connection status].

[0026] For example, a factory deploys 3 injection molding machines (numbered M1, M2, and M3 respectively). The raw data collected at 14:30:00.000 on May 10, 2024 is as follows: 2024-05-10 14:30:00.000: [M1: 220V, 15A, 1; M2: 218V, 14A, 1; M3: 222V, 16A, 1]. The network connection status between M1 and M2, and between M1 and M3 is 1, and the network connection status between M2 and M3 is 1, indicating that the 3 injection molding machines maintain stable connectivity.

[0027] In step S12, wavelet transform is performed on the voltage and current data in the original dataset to obtain a preliminary clean dataset.

[0028] It should be noted that the wavelet transform uses the db4 wavelet basis function, and the decomposition level is set to 3 levels. The first step involves performing 3-level wavelet decomposition on the original voltage and current data respectively, obtaining approximation coefficients (low-frequency part) reflecting the data trend and detail coefficients (high-frequency part) reflecting noise. The second step involves using soft thresholding to denoise the detail coefficients, where the threshold is equal to 0.6745 multiplied by median(|detail coefficient|), where median represents the median and |detail coefficient| represents the absolute value of the detail coefficient. The third step involves reconstructing the denoised detail coefficients and approximation coefficients through inverse wavelet transform to obtain the noise-removed voltage and current data, which together constitute the initial clean dataset.

[0029] For example, the original voltage data of 10 consecutive acquisition points of an M1 injection molding machine is [220,225,218,222,230,215,223,228,219,221]. After db4 wavelet 3-level decomposition and soft threshold denoising, the noise components in the detail coefficients, namely abnormal fluctuation values ​​such as 225, 230, and 215, are filtered out. The reconstructed preliminary clean voltage data is [220,221,219,222,223,218,222,224,219,221]. This data is closer to the voltage stability trend of the actual operation of the equipment.

[0030] In step S13, if the fluctuation amplitude of the voltage data in the preliminary cleaning dataset exceeds the preset voltage fluctuation amplitude threshold, the voltage data is subjected to Fourier transform processing to analyze the frequency distribution and determine the characteristics of the small voltage instability signal.

[0031] It should be noted that the preset voltage fluctuation amplitude threshold is determined based on the rated voltage of the injection molding machine, and is calculated by multiplying the sum of 1 and ±5% by the rated voltage. For example, for an injection molding machine with a rated voltage of 220V, the preset voltage fluctuation amplitude threshold range is 209V-231V. The voltage fluctuation amplitude is the difference between the maximum and minimum values ​​of the initial cleaning voltage data within a certain continuous acquisition period. The Fourier transform uses the Fast Fourier Transform (FFT) algorithm to convert the voltage data in the time domain into the frequency domain data. The frequency resolution is set to 1Hz. By analyzing the frequency components with amplitudes exceeding 0.5V in the frequency domain, the characteristics of the small voltage instability signal are determined. The characteristic representation format is fluctuation frequency: XHz, amplitude: YV.

[0032] For example, the maximum value of the initial cleaning voltage data of the M2 injection molding machine within 1 minute (120 sampling points) is 232V, the minimum value is 208V, and the fluctuation range is 14V, which exceeds the preset threshold of 220V±5% (11V). After performing FFT transformation on this voltage data, the amplitude at 15Hz in the frequency domain is 1.2V, the amplitude at 20Hz is 0.8V, and the amplitude at other frequencies is less than 0.5V. Therefore, the characteristics of the small voltage instability signal are determined as follows: fluctuation frequency: 15Hz, amplitude: 1.2V; fluctuation frequency: 20Hz, amplitude: 0.8V.

[0033] In step S14, the device operating status dataset is obtained by fusing the features of the preliminary cleaning dataset, the micro voltage instability signal, and the network connection relationship information.

[0034] It should be noted that the fusion operation is performed according to the timestamp alignment principle: First, the preliminary clean dataset (voltage data, current data), the characteristics of minor voltage instability signals (fluctuation frequency, amplitude), and network connectivity information (0 / 1 states) are associated according to the same timestamp. Second, the dimensional structure of the device operating status dataset is constructed as [time series length × (voltage dimension + current dimension + fluctuation frequency dimension + fluctuation amplitude dimension + network status dimension)]. The time series length is the total number of collected data, and each dimension stores the corresponding data type. Third, the associated data is standardized to ensure that the data for each dimension corresponding to each timestamp is complete and without missing data, ultimately forming the device operating status dataset.

[0035] For example, the fused data at 14:30:01.000 on 2024-05-10 is 221V (voltage), 15A (current), 15Hz / 20Hz (fluctuation frequency), 1.2V / 0.8V (fluctuation amplitude), and 1 (network status). If the time series length is 120 (corresponding to 2 minutes of data collection), then the device operation status dataset is a matrix of 120 rows × 5 columns, with each row corresponding to a complete data set of a timestamp.

[0036] In step S2, based on the equipment operating status dataset, a dynamic correlation diagram between equipment is constructed, and the strength of the energy consumption interaction relationship between each injection molding machine is calculated, including: S21, extract the energy consumption-related data of each injection molding machine and the network connection relationship information from the equipment operation status dataset, and construct the graph nodes corresponding to each injection molding machine; S22, extract features from the graph nodes to obtain a graph node feature set; the graph node features include the average power and energy consumption fluctuation range of the injection molding machine. S23, if the feature difference between any graph node in the graph node feature set and its adjacent graph nodes exceeds a preset feature difference threshold, then the energy consumption related data of the adjacent graph nodes are weighted and aggregated, and the graph node features are updated to obtain an updated graph node feature set. S24, Based on the updated graph node feature set and the network connection relationship information, determine the initial connection relationship between each graph node and construct a dynamic association graph between devices; S25, based on the dynamic relationship diagram between the devices, calculate the interaction strength score between each pair of injection molding machines; the interaction strength score is the energy consumption interaction strength between each injection molding machine.

[0037] In step S21, energy consumption-related data of each injection molding machine and network connection information are extracted from the equipment operation status dataset to construct graph nodes corresponding to each injection molding machine.

[0038] It should be noted that the energy consumption related data is calculated from the voltage and current data in the equipment operation status dataset, including average power (obtained by multiplying the average voltage data over a time period by the average current data over a time period, in kilowatts (kW)) and energy consumption fluctuation range (obtained by subtracting the minimum power value from the maximum power value over a time period, in kW); network connection relationship information is directly extracted from the network status dimension in the equipment operation status dataset; graph nodes are uniquely identified by the injection molding machine number, and the initial attributes of each graph node are injection molding machine number: [average power, energy consumption fluctuation range].

[0039] For example, the average voltage of injection molding machine M1 over 10 minutes is extracted from the equipment operation status dataset as 220V and the average current as 15A, and the average power is calculated as 220V×15A=3.3kW; the maximum power during this period is 3.5kW and the minimum power is 3.0kW, and the energy consumption fluctuation range is 3.5kW-3.0kW=0.5kW; the average voltage of injection molding machine M2 during the same period is 218V and the average current is 14A, and the average power is 218V×14A=3052W≈3.05kW, and the energy consumption fluctuation range is 0.4kW; the network connection relationship is M1 connected to M2, M1 connected to M3, and M2 connected to M3; then the constructed graph nodes are M1: [3.3kW, 0.5kW]; M2: [3.05kW, 0.4kW]; M3: [3.55kW, 0.6kW].

[0040] In step S22, feature extraction is performed on the graph nodes to obtain a graph node feature set; the graph node features include the average power and energy consumption fluctuation range of the injection molding machine.

[0041] It should be noted that the feature extraction is performed using a Graph Convolutional Layer (GCN). The specific steps are as follows: First, the initial attributes (average power, energy consumption fluctuation) of the graph nodes are normalized. The normalized value is obtained by calculating the difference between the original value and the minimum attribute value, and the difference between the maximum attribute value and the minimum attribute value, and then dividing the former by the latter. The minimum and maximum attribute values ​​are the minimum and maximum values ​​of the corresponding attributes for all graph nodes, respectively. Second, the normalized attributes are input into a preset graph convolutional layer. The weight matrix of the graph convolutional layer is generated through random initialization (optimized during subsequent training), mapping the 2D initial attributes to a 128-dimensional high-dimensional feature vector. Third, the 128-dimensional feature vectors of all graph nodes are collected to form a graph node feature set.

[0042] For example, the minimum average power of all graph nodes is 3.05kW and the maximum is 3.55kW. The average power of M1, 3.3kW, is normalized to (3.3-3.05) / (3.55-3.05)=0.5. The minimum energy consumption fluctuation range is 0.4kW and the maximum is 0.6kW. The energy consumption fluctuation range of M1, 0.5kW, is normalized to (0.5-0.4) / (0.6-0.4)=0.5. Inputting [0.5, 0.5] into the graph convolutional layer, after mapping by the weight matrix, a 128-dimensional feature vector [0.21, 0.15, 0.33, ..., 0.28] (a total of 128 elements) is obtained. This vector is the graph node feature of M1. Similarly, the graph node features of M2 and M3 are obtained, which together constitute the graph node feature set.

[0043] In step S23, if the feature difference between any graph node in the graph node feature set and its adjacent graph nodes exceeds a preset feature difference threshold, then the energy consumption related data of the adjacent graph nodes are weighted and aggregated, and the graph node features are updated to obtain an updated graph node feature set.

[0044] It should be noted that the feature differences are calculated using Euclidean distance, and the calculation formula is as follows: ,in - These are elements of the 128-dimensional feature vector of the current graph node. - d represents the elements of the 128-dimensional feature vectors of adjacent graph nodes, where d is the feature difference value.

[0045] For example, the feature vector of M1 is [0.21, 0.15, ..., 0.28], and the feature vector of the neighboring node M2 ​​is [0.18, 0.12, ..., 0.25]. The feature difference is calculated. .

[0046] It is worth noting that the preset feature difference threshold is determined by statistically analyzing the feature difference values ​​between all graph nodes. It is calculated by multiplying the average feature difference values ​​of all graph nodes by 1.2 (1.2 is a weighting coefficient, i.e., reserving 20% ​​redundancy space to balance screening sensitivity and computational stability). For example, if the average feature difference value between all graph nodes is 0.6, then the preset feature difference threshold is 0.6 × 1.2 = 0.72. The judgment logic is that if the feature difference d is greater than the preset feature difference threshold, it is determined to exceed it; otherwise, it is determined not to exceed it. For example, the feature difference d between M1 and M2 is 0.8 > 0.72, which is determined to exceed the preset feature difference threshold; the feature difference d between M2 and M3 is 0.65 < 0.72, which is determined not to exceed it.

[0047] It should be noted that the weight value of the weighted aggregation is the reciprocal of the feature difference d. The smaller the feature difference, the larger the weight. The energy consumption data after aggregation is calculated as follows: First, the energy consumption data of the current graph node (including the average power and energy consumption fluctuation amplitude corresponding to the current graph node) is multiplied by 0.6 to obtain the first part of the result; then, the energy consumption data of all adjacent graph nodes (including the average power and energy consumption fluctuation amplitude corresponding to each adjacent graph node) is multiplied by the corresponding weight, and these products are added together to obtain the sum. At the same time, the sum of the weights corresponding to all adjacent graph nodes is calculated. The sum of the former is divided by the sum of the latter, and the result is multiplied by 0.4 to obtain the second part of the result; finally, the first part of the result and the second part of the result are added together to obtain the final aggregated energy consumption data.

[0048] For example, M1 has an average power of 3.3kW and an energy consumption fluctuation of 0.5kW, while M2 has an average power of 3.05kW and an energy consumption fluctuation of 0.4kW. The characteristic difference d is 0.8, and the weight is 1 / 0.8=1.25. The average power after aggregation is (3.3×0.6)+(3.05×1.25 / 1.25)×0.4=3.2kW. The energy consumption fluctuation after aggregation is (0.5×0.6)+(0.4×1.25 / 1.25)×0.4=0.46kW.

[0049] It should be noted that the graph node features are generated by re-inputting the aggregated energy consumption data (average power, energy consumption fluctuation range) into the graph convolutional layer, generating a new 128-dimensional feature vector according to the feature extraction method in step S22, and replacing the original graph node features; all updated graph node feature vectors are collected to form an updated graph node feature set.

[0050] For example, after normalizing the average power of M1 (3.2kW) and energy consumption fluctuation (0.46kW), the new feature vector [0.198, 0.138, ..., 0.26] is input into the graph convolutional layer and replaced with the original feature vector of M1. The feature vectors of M2 and M3 are not updated, and the updated graph node feature set is finally obtained.

[0051] In step S24, the initial connection relationship between each graph node is determined based on the updated graph node feature set and the network connection relationship information, and a dynamic association graph between devices is constructed.

[0052] It should be noted that the initial connection relationship is determined based on the network connection relationship information: if the network connection status of two injection molding machines is 1 (connected), then there is an edge between the corresponding two graph nodes; if the network connection status is 0 (disconnected), then there is no edge between the corresponding two graph nodes; the structure of the dynamic association graph between devices includes a graph node set, an edge set, and edge attributes, wherein the graph node set is the graph nodes corresponding to all injection molding machines, the edge set is the graph node pairs with connection relationships, and the initial weight of the edge attributes is set to 1.

[0053] For example, if the network connection status of M1 and M2, M1 and M3, and M2 and M3 are all 1, then the edge set is (M1-M2), (M1-M3), and (M2-M3), and the initial weight of each edge is 1. The final constructed dynamic association graph is an undirected graph containing 3 nodes and 3 edges.

[0054] In step S25, based on the dynamic correlation diagram between the devices, the interaction strength score between each pair of injection molding machines is calculated; the interaction strength score is the energy consumption interaction strength between each injection molding machine.

[0055] It should be noted that the calculation of the interaction strength score includes: performing a vector dot product operation on the feature vector of the current graph node and the feature vector of the adjacent graph node. Specifically, the vector dot product refers to multiplying the corresponding elements of the two 128-dimensional feature vectors respectively, and then adding all the results of the multiplication; next, multiplying the result of the above vector dot product operation with the edge attribute weight of the corresponding edge between the current graph node and the adjacent graph node in the dynamic association graph, where the initial value of the edge attribute weight is set to 1; finally, the calculated result is rounded to two decimal places. This rounded result is the energy consumption interaction strength between the injection molding machine corresponding to the current graph node and the injection molding machine corresponding to the adjacent graph node.

[0056] For example, the dot product of the updated feature vector of M1 and the feature vector of M2 is 0.7, and the edge attribute weight is 1, so the interaction strength score is 0.7×1=0.70; the dot product of the feature vectors of M1 and M3 is 0.82, so the interaction strength score is 0.82×1=0.82; the dot product of the feature vectors of M2 and M3 is 0.65, so the interaction strength score is 0.65×1=0.65.

[0057] In step S3, if the energy consumption interaction strength exceeds a preset energy consumption interaction strength threshold, then high-intensity interaction paths are extracted, and the network connection relationship information is fused to obtain a set of potential chain reaction paths, including: S31, when the energy consumption interaction relationship strength exceeds the preset energy consumption interaction strength threshold, extract the path whose interaction strength score meets the energy consumption interaction strength threshold requirement from the dynamic association graph between devices to obtain a set of high-intensity interaction paths. S32, integrate the network connection relationship information, traverse the set of high-intensity interaction paths, and if the network connection relationship meets the preset connection stability standard, then filter out potential chain reaction paths to obtain a set of potential chain reaction paths.

[0058] In step S31, when the energy consumption interaction strength exceeds the preset energy consumption interaction strength threshold, the paths whose interaction strength scores meet the energy consumption interaction strength threshold requirement are extracted from the dynamic association graph between devices to obtain a set of high-intensity interaction paths.

[0059] It should be noted that the preset energy consumption interaction intensity threshold is determined by statistically analyzing the energy consumption interaction strength between all injection molding machines over the past 30 days, i.e., the 75th percentile of the historical interaction intensity score. For example, if the 75th percentile of the historical interaction intensity score is 0.6, then the preset threshold is set to 0.6. The interaction intensity score meets the threshold requirement, meaning that the interaction intensity score is greater than or equal to the preset energy consumption interaction intensity threshold. The high-intensity interaction path is the path formed by the edges between adjacent graph nodes in the dynamic association graph where the interaction intensity score is greater than or equal to the threshold. The path is represented in the format of injection molding machine number 1, injection molding machine number 2, etc. All paths that meet the requirements together constitute the high-intensity interaction path set.

[0060] For example, if the interaction intensity score of M1-M2 is 0.70≥0.6, the score of M1-M3 is 0.82≥0.6, and the score of M2-M3 is 0.65≥0.6, then the set of high-intensity interaction paths is [M1-M2, M1-M3, M2-M3]; if the score of M2-M3 is 0.58<0.6, then the path is removed, and the set becomes [M1-M2, M1-M3].

[0061] In step S32, the network connection relationship information is integrated, and the set of high-intensity interaction paths is traversed. If the network connection relationship meets the preset connection stability standard, potential chain reaction paths are selected to obtain a set of potential chain reaction paths.

[0062] It should be noted that the fusion operation involves associating each injection molding machine pair in the high-intensity interaction path with the network connection relationship information in the equipment operation status dataset, and obtaining the network connection status time series of the injection molding machine pair in the most recent 5 minutes (e.g., [1,1,1,0,1], representing the connection status within 5 1-minute time intervals); the traversal method is to read each path in the high-intensity interaction path set one by one in sequence, and perform the association operation on all injection molding machine pairs contained in the path.

[0063] For example, when traversing the high-intensity interaction path M1-M2, the network connection status time series of M1 and M2 in the last 5 minutes is [1,1,1,1,1]; when traversing the path M1-M3, the time series is [1,1,0,1,1].

[0064] It should be noted that the process of traversing the set of high-intensity interaction paths, and filtering out potential chain reaction paths if the network connection relationship meets the preset connection stability standard, to obtain a set of potential chain reaction paths, involves directly adding paths that meet the standard to a temporary set; then removing paths that do not meet the standard from the traversal list and not entering the temporary set; the judgment result of each path needs to be recorded during the operation; then all paths in the temporary set are sorted from high to low according to the interaction intensity score to form an ordered set of potential chain reaction paths; the set format is [path1 (score), path2 (score), ...].

[0065] For example, M1-M2 meets the criteria and is included in the temporary set; M1-M3 does not meet the criteria and is removed; if the stable connectivity rate of M2-M3 is 95% ≥ 90%, it is included in the temporary set; the paths in the temporary set are M1-M2 (0.70) and M2-M3 (0.65). After sorting by score, the potential chain reaction path set is [M1-M2 (0.70), M2-M3 (0.65)].

[0066] It should be noted that the preset connection stability standard is that the network connection status of each adjacent injection molding machine in the path has a stable connectivity rate of ≥90% within the most recent 5 minutes. The stable connectivity rate is calculated as follows: first, count the total number of time periods marked as 1 for network connection status; then, divide this number by the total number of time periods within the statistical range; finally, multiply the result by 100%. The resulting value is the stable connectivity rate. The total number of time periods is set to 5 (corresponding to 5 minutes, with 1 time period per minute). The judgment logic is: if the stable connectivity rate of all adjacent injection molding machines in the path is greater than or equal to 90%, the standard is met; otherwise, it is not met.

[0067] For example, the stable connectivity rate of path M1-M2 is (5 / 5)×100%=100%, which is greater than 90%; the stable connectivity rate of path M1-M3 is (4 / 5)×100%=80%<90%; therefore, M1-M2 meets the standard, while M1-M3 does not meet the standard.

[0068] In step S4, based on the set of potential chain reaction paths, the future voltage and current fluctuation trends are predicted, and the equipment groups with excessive load during peak production periods are identified, including: S41, Based on the set of potential chain reaction paths, extract voltage and current data during peak production periods from the equipment operation status dataset, and perform time series analysis to obtain voltage and current fluctuation trend items; S42, calculate the probability that the voltage and current fluctuation trend item exceeds the preset fluctuation threshold, and determine the injection molding machine group whose probability exceeds the preset probability threshold as the equipment group with excessive load during the peak production period.

[0069] In step S41, voltage and current data during peak production periods are extracted from the equipment operation status dataset based on the set of potential chain reaction paths, and time series analysis is performed to obtain voltage and current fluctuation trend terms.

[0070] It should be noted that the peak production period is determined according to the factory's production plan, for example, 14:00-16:00 daily. The extraction operation is to filter out the voltage and current data of injection molding machines that belong to the potential chain reaction path set during the peak production period from the equipment operation status dataset. The time series analysis uses the ARIMA model, and the model parameters (p, d, q) are determined by the AIC criterion (p is the autoregression order, d is the difference order, and q is the moving average order). The analysis steps are as follows: First, perform a stationarity test (ADF test) on the extracted voltage / current data. If it is not stationary, perform d-order differencing (usually d=1). Second, determine the q value (ACF truncation order) through the ACF plot (autocorrelation function plot) and the p value (PACF truncation order) through the PACF plot (partial autocorrelation function plot). Third, fit the ARIMA(p, d, q) model, perform seasonal decomposition on the data, and separate the trend term, seasonal term, and residual term, where the trend term is the voltage and current fluctuation trend term.

[0071] For example, the injection molding machines in the potential chain reaction path set are M1 and M2. Voltage data from 14:00 to 16:00 are extracted. The ADF test p value is 0.03 < 0.05 (stationary). The ACF is truncated at order 2 and the PACF is truncated at order 1. The ARIMA(2,0,1) is determined. After fitting the model, the voltage fluctuation trend term is found to be an increase of 2V every 10 minutes, and the current fluctuation trend term is an increase of 0.5A every 10 minutes.

[0072] In step S42, the probability that the voltage and current fluctuation trend term exceeds the preset fluctuation threshold is calculated, and the injection molding machine group whose probability exceeds the preset probability threshold is determined to be the equipment group with excessive load during the peak production period.

[0073] It should be noted that the preset fluctuation threshold includes a preset voltage fluctuation threshold and a preset current fluctuation threshold. The preset voltage fluctuation threshold is the rated voltage multiplied by ±5%; the preset current fluctuation threshold is determined according to the rated current of the injection molding machine, that is, the rated current multiplied by ±10%. For example, for an injection molding machine with a rated current of 20A, the threshold range is 18A-22A; the preset probability threshold is set to 80%, that is, a probability greater than or equal to 80% is judged as an overload risk.

[0074] For example, if M1 has a rated voltage of 220V and a rated current of 20A, then the preset voltage fluctuation threshold is 209V-231V, the preset current fluctuation threshold is 18A-22A, and the preset probability threshold is 80%.

[0075] It should be noted that the prediction operation is based on the ARIMA model fitted in step S41, with a prediction duration of one future production peak period (e.g., 14:00-16:00 the next day), a prediction interval of 1 minute, and a total of 120 prediction points. The prediction results are stored in the format of time point: [predicted voltage value, predicted current value]. For example, based on the ARIMA(2,0,1) model, the predicted voltage value of M1 at 14:00 the next day is 222V and the current value is 18.5A; the predicted voltage value at 14:10 is 224V and the current value is 19A; and the predicted voltage value at 15:00 is 232V and the current value is 22.5A, for a total of 120 prediction point data.

[0076] It should be noted that the calculation steps are as follows: First, count the number of times the predicted voltage value exceeds the preset voltage fluctuation threshold upper limit (e.g., 231V) or falls below the lower limit (e.g., 209V), and record it as Nv; Second, count the number of times the predicted current value exceeds the preset current fluctuation threshold upper limit (e.g., 22A) or falls below the lower limit (e.g., 18A), and record it as Ni; Third, calculate the total number of times the value exceeds the threshold. The total number of predictions is 120; the probability calculation formula is (N / 120) × 100%. For example, the number of times the predicted voltage value of M1 exceeds 231V is 50 (Nv=50), the number of times the predicted current value exceeds 22A is 30 (Ni=30), and the total number of times exceeding the standard is N=80; the probability = (80 / 120) × 100% ≈ 66.7%; if Nv=60, Ni=36, N=96, the probability = (96 / 120) × 100% = 80%.

[0077] The determination logic is as follows: if the probability of a certain injection molding machine is greater than or equal to a preset probability threshold (e.g., 80%), then the injection molding machine is classified into the overloaded equipment group; if multiple injection molding machines meet the conditions, then a set of equipment groups is formed with the injection molding machine number as the element. For example, if the probability of M1 is 80%, the probability of M2 is 85%, and the probability of M3 is 75%, then the overloaded equipment group during peak production is [M1, M2].

[0078] In step S5, based on the equipment group with excessive load during peak production periods, the current energy load allocation matrix is ​​obtained, the load evolution characteristics over time are analyzed, and the range of equipment power parameters that need to be adjusted is determined, including: S51, Based on the equipment group with excessive load during peak production period, obtain the energy load data of each injection molding unit within a preset time period, and construct the current energy load allocation matrix; S52, perform time series analysis on the load data in the current energy load distribution matrix, extract the load change trend of each injection molding unit, and obtain a load evolution feature set; S53, if the load change amplitude of any injection molding unit in the load evolution feature set exceeds the preset load change threshold, then perform correlation analysis on the voltage data of the target injection molding unit and calculate the voltage instability probability. S54. Based on the voltage instability probability and the current power data of the target injection molding unit, fit the relationship between power and load change, and determine the range of equipment power parameters that need to be adjusted.

[0079] In step S51, based on the equipment group with excessive load during peak production period, the energy load data of each injection molding unit within a preset time period is obtained, and the current energy load allocation matrix is ​​constructed.

[0080] It should be noted that the preset time period is set to 30 minutes; the energy load data is the real-time power of the injection molding machine, which is calculated from the voltage and current data in the equipment operation status dataset (the real-time power is calculated by multiplying the instantaneous voltage value by the instantaneous current value, unit: kW); the rows of the current energy load allocation matrix represent the injection molding machine group number (the injection molding machine in the overloaded equipment group), the columns represent the time points (1 time point every 1 minute, 30 in total), and the matrix elements are the real-time power values ​​of the corresponding time points.

[0081] For example, the equipment group with excessive load during peak production periods is [M1, M2], with a preset time period of 14:00-14:30, totaling 30 time points; the real-time power of M1 is 3.3kW at 14:00, 3.4kW at 14:01, and 3.8kW at 14:29; the power of M2 is 3.05kW at 14:00, 3.1kW at 14:01, and 3.5kW at 14:29; then the current energy load allocation matrix is: Time points 14:00, 14:01, ..., 14:29 M1 3.3kW 3.4kW...3.8kW M2 3.05kW 3.1kW...3.5kW In step S52, time series analysis is performed on the load data in the current energy load allocation matrix to extract the load change trend of each injection molding unit and obtain a load evolution feature set.

[0082] It should be noted that the aforementioned time series analysis involves extracting the load change patterns of each injection molding machine from the current energy load distribution matrix to form quantifiable load evolution characteristics.

[0083] For example, the current energy load allocation matrix is ​​for injection molding machines with excessive load during peak production periods, listed as time points within a preset time period (usually one time point every minute). The matrix elements are the load data (i.e., real-time power, unit: kW) of the corresponding injection molding machine at the corresponding time point. It is necessary to confirm that there are no missing data in the matrix and that the format is consistent. For example, time series analysis is performed separately on the load data (one column of data) of each injection molding machine in the matrix. A linear regression model is used to model the data, with time points as independent variables (e.g., the first time point is denoted as 0, the second as 1, and so on), and the real-time power at the corresponding time point as the dependent variable. The regression equation is obtained by fitting the data using the least squares method: p = k × t + b; where p is the real-time power; t is the time point; k is the regression coefficient (i.e., the load change amplitude, k > 0 indicates that the load increases with time, k < 0 indicates that the load decreases with time, and the larger the absolute value of k, the more obvious the change); and b is the intercept (reflecting the load baseline value at the initial time point). For example, the regression coefficient k and intercept b of each injection molding machine are used as the load evolution characteristics of the unit (i.e., the quantitative indicators of load change trends). The characteristics of all units with excessive load are organized according to the injection molding machine number, and finally a set of load evolution characteristics is formed.

[0084] For example, during peak production periods, the equipment group with excessive load consists of two injection molding machines (numbered M1 and M2); the preset time cycle is 30 minutes (corresponding to 30 time points, denoted as 0-29, corresponding to 14:00-14:29, one time point per minute); in the current energy load distribution matrix, the load data (real-time power) of M1 are as follows: 3.3kW (14:00, time point 0), 3.4kW (14:01, time point 1), 3.5kW (14:02, time point 2), ..., 3.8kW (14:29, time point 29); the load data (real-time power) of M2 are as follows: 3.05kW (14:00, time point 0), 3.1kW (14:01, time point 1), 3.15kW (14:02, time point 2), ..., 3.5kW (14:29, time point 29).

[0085] For example, a linear regression analysis was performed on 30 load data points of M1, with time point (0-29) as the independent variable and real-time power as the dependent variable. The regression equation was obtained as p=0.02×t+3.3. Then, the load evolution characteristics of M1 are the regression coefficient k=0.02 (indicating that the load increases by 0.02kW every minute) and intercept b=3.3 (indicating that the initial load at 14:00 is 3.3kW). For example, a linear regression analysis was performed on 30 load data points of M2, and the regression equation p = 0.015 × t + 3.05 was obtained. Then, the load evolution characteristics of M2 are k = 0.015 (meaning that the load increases by 0.015 kW every minute) and b = 3.05 (meaning that the initial load at 14:00 is 3.05 kW). The load evolution characteristics of M1 and M2 were sorted out, and the load evolution characteristic set was obtained as [M1: {regression coefficient k = 0.02, intercept b = 3.3}, M2: {regression coefficient k = 0.015, intercept b = 3.05}].

[0086] In step S53, if the load change amplitude of any injection molding unit in the load evolution feature set exceeds the preset load change threshold, then the voltage data of the target injection molding unit is subjected to correlation analysis to calculate the voltage instability probability.

[0087] It should be noted that the preset load change threshold is the statistical value of the k value of all injection molding machine units, that is, the average value of the k value of all injection molding machine units multiplied by 1.5. For example, the average value of the k value of M1 and M2 is (0.02+0.015) / 2=0.0175, and the preset threshold is 0.0175×1.5≈0.026; the load change amplitude is the k value. If the k value is greater than the preset threshold, then the injection molding machine is the target injection molding machine unit; the voltage correlation analysis uses the Pearson correlation coefficient to calculate the correlation coefficient r between the real-time power and real-time voltage of the target unit (the value of r is in the range of [-1,1], and the closer r is to 1, the stronger the positive correlation); the voltage instability probability is (1-r)×100%.

[0088] For example, if the k value of M1 is 0.02 < 0.026, it is not listed as a target unit; if the k value of M2 is 0.03 > 0.026, it is listed as a target unit; if the correlation coefficient r between the real-time power and voltage of M2 is calculated to be 0.3, then the probability of voltage instability is (1-0.3) × 100% = 70%.

[0089] In step S54, based on the voltage instability probability and the current power data of the target injection molding unit, the relationship between power and load change is fitted to determine the range of equipment power parameters that need to be adjusted.

[0090] It should be noted that the fitting operation adopts a linear regression model, using the current power data of the target unit (real-time power extracted from the energy load distribution matrix) as the independent variable and the load change amplitude k as the dependent variable to obtain a regression equation. The range of equipment power parameters to be adjusted is the power interval that makes k less than or equal to the preset load change threshold. Here, the current power p is less than or equal to the difference between the preset load change threshold and c, divided by a. Here, a and c are regression coefficients in the regression equation, and a is usually a positive value (the higher the power, the larger the k value). The parameter range is expressed as [lower power limit, upper power limit], and the lower power limit is set as the minimum operating power of the injection molding machine (e.g., 1.5kW).

[0091] For example, the fitting equation of M2 is k=0.001×p+0.005, and the preset load change threshold is 0.026; then the upper limit of power is (0.026-0.005) / 0.001=21kW; the minimum operating power of M2 is 1.5kW, then the range of equipment power parameters that need to be adjusted is [1.5kW, 21kW].

[0092] In step S6, if the range of power parameters of the equipment to be adjusted is unevenly distributed, the energy load allocation scheme is optimized to obtain a balanced power adjustment vector, including: S61, if the range of power parameters of the equipment to be adjusted does not meet the preset equalization standard, then the initial power distribution data is extracted from the current energy load allocation matrix; S62, iteratively optimize the initial power distribution data, dynamically update the topology of the dynamic association graph between devices, and calculate the interaction strength score between each injection molding machine after each iteration; S63, adjust the power allocation according to the interaction intensity score until power distribution data that meets the preset equalization standard is obtained, and convert the power distribution data into an equalized power adjustment vector.

[0093] In step S61, if the range of power parameters of the equipment to be adjusted does not meet the preset equalization standard, the initial power distribution data is extracted from the current energy load allocation matrix.

[0094] It should be noted that the preset balancing standard is that the power standard deviation of each injection molding machine unit is less than or equal to 5% of the rated power of the injection molding machine, where the power standard deviation is the standard deviation of the average power of each injection molding machine; for example, if the rated power of the injection molding machine is 30kW, 5% is 1.5kW. If the power standard deviation is >1.5kW, then the balancing standard is not met; the initial power distribution data is the average power of each injection molding machine in the current energy load distribution matrix.

[0095] For example, the average power of M1 is The average power of M2 is ( Power standard deviation is If the rated power is 5kW, 5% is 0.25kW, and 0.15≤0.25, which meets the balance standard; if the average power of M1 is 4kW and the average power of M2 is 2kW, the power standard deviation is about 1.41, which is less than 0.25, which does not meet the standard. The initial power distribution data is extracted as [4kW, 2kW].

[0096] In step S62, the initial power distribution data is iteratively optimized, the topology of the dynamic association graph between devices is dynamically updated, and the interaction strength score between each injection molding machine is calculated after each iteration.

[0097] It should be noted that the initial power distribution data is the average power of each injection molding machine within a preset time period. A genetic algorithm is used for optimization, with fixed parameters: population size 20, maximum number of iterations 50, crossover probability 0.8, and mutation probability 0.1. The optimization objective is to minimize the standard deviation of the average power of each injection molding machine. Iteration is achieved through a loop of generating a power scheme population, calculating fitness (fitness is the reciprocal of the power standard deviation), crossover and mutation to generate new schemes, and selecting the optimal scheme. The dynamic relationship graph topology between devices is dynamically updated, using the new power data after each iteration as the new attribute of the graph nodes, and combining this with the energy consumption fluctuation amplitude to re-update the graph. The graph convolutional layer extracts 128-dimensional graph node features, and then adjusts the edge attribute weights based on the interaction strength score calculated from the new features. The adjustment formula is that the new edge weight is equal to the old edge weight multiplied by the new interaction strength score divided by the old interaction strength score, and the initial value of the old edge weight is set to 1. The interaction strength score is calculated after each iteration. Based on the updated graph node features and the new edge weights, the interaction strength score is equal to the dot product of the current graph node feature vector and the product of the adjacent graph node feature vectors, multiplied by the new edge weights. The vector dot product is a 128-dimensional feature vector. The corresponding elements are multiplied and summed. The calculation result is rounded to two decimal places, which is the strength of the energy consumption interaction relationship between the corresponding injection molding machines.

[0098] For example, the overloaded equipment group includes two injection molding machines, M1 and M2. The initial power distribution data is M1=4kW, M2=2kW, and the power standard deviation is approximately 1.41kW (exceeding the preset equalization standard by 0.25kW). In the initial dynamic relationship graph between the equipment, the edge weight of M1 and M2 is 1, and the initial interaction strength score is 0.7. During the first iteration of optimization, the genetic algorithm generates 20 power allocation schemes, such as [4kW, 2kW], [3.9kW, 2.1kW], etc. Through crossover operations (such as [3.9kW, 2.1kW] and [3.8kW, 2.2kW] crossover to generate [3.85kW, 2.15kW]) and mutation operations (such as [3.8kW, 2.2kW] crossover to generate [3.85kW, 2.15kW]), the power allocation schemes are further optimized. The power [W] mutated to [3.72kW, 2.28kW]), and the optimal solution was finally selected as M1=3.8kW and M2=2.2kW, with a power standard deviation of approximately 0.89kW. Subsequently, the topology was updated, and M1=3.8kW and M2=2.2kW were used as new node attributes. Combined with the energy consumption fluctuation range of the two (M1=0.5kW and M2=0.4kW), a 128-dimensional feature vector was re-extracted, and the new interaction strength score was calculated to be 0.75. The edge weight was adjusted according to the formula 1×(0.75 / 0.7)≈1.07. Finally, the interaction strength score was calculated to be 0.75×1.07≈0.80, that is, the energy consumption interaction strength between M1 and M2 after this iteration is 0.80.

[0099] In step S63, the power allocation is adjusted according to the interaction intensity score until power distribution data that meets the preset equalization standard is obtained, and the power distribution data is converted into an equalized power adjustment vector.

[0100] It should be noted that the power allocation adjustment rule is that the higher the interaction strength score of the injection molding machine, the larger the corresponding power adjustment range. Specifically, the power adjustment method is to subtract the product of the interaction strength score and the adjustment step size from the current power of the injection molding machine. The adjustment step size is calculated by subtracting the initial minimum power from the maximum power in the initial power distribution data, and then dividing the difference by the maximum number of iterations. For example, if the initial maximum power is 4kW, the initial minimum power is 2kW, and the maximum number of iterations is 50, then the adjustment step size is the difference between 4kW and 2kW divided by 50, resulting in 0.04kW. If the interaction strength score of injection molding machine M1 is 0.86 and its current power is 3.8kW, then the new power of M1 is 3.8kW minus (0.86 multiplied by 0.04kW), resulting in approximately 3.766kW. This power adjustment is repeated until the standard deviation of the average power of each injection molding machine is less than or equal to the preset equilibrium standard.

[0101] It is worth noting that the aforementioned balanced power adjustment vector refers to the vector formed by subtracting the initial power from the final power of each injection molding machine, and then organizing all the resulting differences according to the injection molding machine number.

[0102] For example, the final power distribution data that meets the standard is [3.1kW, 2.9kW], and the initial power distribution data is [4kW, 2kW]. Then the power adjustment vector is [3.1-4=-0.9kW, 2.9-2=+0.9kW].

[0103] In step S7, control commands for multiple injection molding machines are generated and updated based on the balanced power adjustment vector, including: S71, Based on the balanced power adjustment vector, obtain the target power parameters of each injection molding machine; S72, determine whether each injection molding unit is overloaded. If an overload is found, adjust the target power parameter of the injection molding unit. S73, based on the adjusted target power parameters, generate control commands for multiple injection molding machines, send the control commands to the corresponding injection molding machines, and update the operating parameters of the injection molding machines.

[0104] In step S71, the target power parameters of each injection molding machine are obtained based on the balanced power adjustment vector.

[0105] It should be noted that the target power parameter is calculated as the current power of the injection molding machine plus the corresponding value in the balanced power adjustment vector; the current power is the average power in the initial power distribution data; the unit of the target power parameter is kW, and it is rounded to two decimal places.

[0106] For example, M1 has a current power of 4kW, an adjustment vector value of -0.9kW, and a target power parameter of 3.10kW; M2 has a current power of 2kW, an adjustment vector value of +0.9kW, and a target power parameter of 2.90kW.

[0107] In step S72, it is determined whether each injection molding unit is overloaded. If an overload is found, the target power parameter of the injection molding unit is adjusted.

[0108] It should be noted that the overload judgment standard is that the target power parameter is greater than 110% of the rated power of the injection molding machine, where the rated power is the rated operating power marked at the factory of the injection molding machine (e.g., M1 rated power = 3.3kW, 110% = 3.63kW); the judgment reference is 110% of the rated power of the injection molding machine, with a 10% redundancy margin reserved for the rated power of the injection molding machine to avoid false triggering of overload judgment due to short-term normal fluctuations; if the target power parameter exceeds this value, it is judged as overload.

[0109] For example, M1 has a rated power of 3.3kW, and the overload judgment criterion is that the target power parameter is greater than 3.63kW; M2 has a rated power of 3.0kW, and the overload judgment criterion is that the target power parameter is greater than 3.3kW.

[0110] It should be noted that the adjusted target power parameter is 110% of the rated power of the injection molding machine, that is, the maximum value of the overload judgment standard is taken; for injection molding machines that are not overloaded, the target power parameter remains unchanged as calculated in step S71.

[0111] For example, if M2 is overloaded, the adjusted target power parameter is 3.0kW × 110% = 3.30kW; if M1 is not overloaded, it remains unchanged at 3.10kW.

[0112] In step S73, control commands for multiple injection molding machines are generated based on the adjusted target power parameters, and the control commands are sent to the corresponding injection molding machines to update the operating parameters of the injection molding machines.

[0113] It should be noted that the control command is a digital signal, in the format of injection molding machine number: [adjusted target power parameter, execution time], where the execution time is the current time + 10 seconds (to avoid real-time adjustment from impacting the equipment); the control command is sent to the injection molding machine's PLC controller via the Modbus protocol; after receiving the command, the injection molding machine adjusts its operating power to the target power parameter within the execution time, completing the operating parameter update.

[0114] For example, the current time is 14:30:00, the target power parameter of M1 after adjustment is 3.10kW, and that of M2 after adjustment is 3.30kW; the generated control instructions are M1: [3.10kW, 14:30:10]; M2: [3.30kW, 14:30:10]; the PLC controller receives the instructions at 14:30:10 and controls the power of M1 and M2 to be adjusted to 3.10kW and 3.30kW respectively, completing the update.

[0115] Reference Figure 2 The second embodiment of the present invention provides an energy-saving control system for injection molding machines based on multi-source data, comprising: The data acquisition and preprocessing module is used to acquire voltage and current data and network connection information of multiple injection molding machines, and preprocess the voltage and current data and network connection information to obtain a data set of equipment operating status. The correlation graph construction and strength calculation module is used to construct a dynamic correlation graph between devices based on the device operating status dataset and calculate the strength of the energy consumption interaction relationship between each injection molding machine. The chain path extraction module is used to extract high-intensity interaction paths and fuse the network connection relationship information to obtain a set of potential chain reaction paths if the energy consumption interaction relationship strength exceeds a preset energy consumption interaction strength threshold. The fluctuation prediction and judgment module is used to predict future voltage and current fluctuation trends based on the set of potential chain reaction paths, and to determine the equipment groups with excessive load during peak production periods. The load matrix processing module is used to obtain the current energy load distribution matrix based on the equipment group with excessive load during the peak production period, analyze the load evolution characteristics in the time dimension, and determine the range of equipment power parameters that need to be adjusted. The power optimization module is used to optimize the energy load allocation scheme to obtain a balanced power adjustment vector if the power parameter range of the equipment to be adjusted is unevenly distributed. The control command update module is used to generate and update control commands for multiple injection molding machines based on the balanced power adjustment vector.

[0116] It should be noted that the energy-saving control system for injection molding machines based on multi-source data provided in this embodiment of the invention is used to execute all the process steps of the energy-saving control method for injection molding machines based on multi-source data in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0117] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as an energy-saving control program for injection molding machines based on multi-source data. When the processor executes the computer program, it implements the steps in the various embodiments of the energy-saving control method for injection molding machines based on multi-source data described above, for example... Figure 1 Step S1 is shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the data acquisition and preprocessing module.

[0118] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0119] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0120] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting various parts of the electronic device through various interfaces and lines.

[0121] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0122] If the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0123] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0124] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An energy-saving control method for injection molding machines based on multi-source data, characterized in that, include: The voltage and current data and network connection information of multiple injection molding machines are collected and preprocessed to obtain a dataset of equipment operating status. Based on the equipment operating status dataset, a dynamic correlation diagram between equipment is constructed, and the strength of the energy consumption interaction relationship between each injection molding machine is calculated. If the energy consumption interaction strength exceeds a preset energy consumption interaction strength threshold, then high-intensity interaction paths are extracted, and the network connection relationship information is fused to obtain a set of potential chain reaction paths. Based on the set of potential chain reaction paths, predict future voltage and current fluctuation trends and identify equipment groups with excessive load during peak production periods. Based on the equipment group with excessive load during the peak production period, obtain the current energy load distribution matrix, analyze the load evolution characteristics in the time dimension, and determine the range of equipment power parameters that need to be adjusted; If the range of power parameters of the equipment to be adjusted is unevenly distributed, the energy load allocation scheme is optimized to obtain a balanced power adjustment vector; Based on the balanced power adjustment vector, control commands for multiple injection molding machines are generated and updated; The step of constructing a dynamic association graph between devices based on the device operation status dataset and calculating the energy consumption interaction strength between each injection molding machine includes: extracting energy consumption-related data of each injection molding machine and network connection information from the device operation status dataset to construct graph nodes corresponding to each injection molding machine; extracting features from the graph nodes to obtain a graph node feature set; the graph node features include the average power and energy consumption fluctuation range of the injection molding machine; if the feature difference between any graph node in the graph node feature set and its adjacent graph nodes exceeds a preset feature difference threshold, then the energy consumption-related data of the adjacent graph nodes are weighted and aggregated, and the graph node features are updated to obtain an updated graph node feature set; determining the initial connection relationship between each graph node based on the updated graph node feature set and the network connection information to construct a dynamic association graph between devices; and calculating the interaction strength score between each pair of injection molding machines based on the dynamic association graph between devices; the interaction strength score is the energy consumption interaction strength between each injection molding machine. Specifically, if the energy consumption interaction strength exceeds a preset energy consumption interaction strength threshold, then extracting high-intensity interaction paths and fusing the network connection relationship information to obtain a potential chain reaction path set includes: when the energy consumption interaction strength exceeds the preset energy consumption interaction strength threshold, extracting paths whose interaction strength scores meet the energy consumption interaction strength threshold requirement from the device dynamic association graph to obtain a high-intensity interaction path set; fusing the network connection relationship information, traversing the high-intensity interaction path set, and if the network connection relationship meets a preset connection stability standard, then filtering out potential chain reaction paths to obtain a potential chain reaction path set.

2. The energy-saving control method for injection molding machines based on multi-source data according to claim 1, characterized in that, The process involves collecting voltage and current data from multiple injection molding machines and their network connectivity information, followed by preprocessing to obtain a dataset of equipment operating status, including: By using sensors on multiple injection molding machines, the voltage and current data of each injection molding machine, as well as the network connection information between the injection molding machines, are collected in real time to generate a raw dataset with time dimension markings. Wavelet transform is performed on the voltage and current data in the original dataset to obtain a preliminary clean dataset; If the voltage fluctuation amplitude in the preliminary clean data exceeds the preset voltage fluctuation amplitude threshold, then the voltage data is subjected to Fourier transform processing to analyze the frequency distribution and determine the characteristics of the small voltage instability signal. By integrating the preliminary cleaning dataset, the characteristics of the minor voltage instability signal, and the network connectivity information, the device operating status dataset is obtained.

3. The energy-saving control method for injection molding machines based on multi-source data according to claim 1, characterized in that, The step of predicting future voltage and current fluctuation trends and identifying equipment groups with excessive load during peak production periods based on the set of potential chain reaction paths includes: Based on the set of potential chain reaction paths, voltage and current data during peak production periods are extracted from the equipment operation status dataset, and time series analysis is performed to obtain voltage and current fluctuation trend terms. Calculate the probability that the voltage and current fluctuation trend terms exceed a preset fluctuation threshold, and determine the injection molding machines whose probability exceeds the preset probability threshold as equipment groups with excessive load during peak production periods.

4. The energy-saving control method for injection molding machines based on multi-source data according to claim 1, characterized in that, The process involves obtaining the current energy load allocation matrix based on the equipment group with excessive load during peak production periods, analyzing the load evolution characteristics over time, and determining the range of equipment power parameters that need to be adjusted, including: Based on the equipment group with excessive load during peak production periods, obtain the energy load data of each injection molding machine within a preset time period, and construct the current energy load allocation matrix. Time series analysis is performed on the load data in the current energy load distribution matrix to extract the load change trend of each injection molding machine and obtain a set of load evolution characteristics; If the load change amplitude of any injection molding machine in the load evolution feature set exceeds the preset load change threshold, then the voltage data of the target injection molding machine is subjected to correlation analysis to calculate the voltage instability probability. Based on the voltage instability probability and the current power data of the target injection molding machine, the relationship between power and load change is fitted to determine the range of equipment power parameters that need to be adjusted.

5. The energy-saving control method for injection molding machines based on multi-source data according to claim 1, characterized in that, If the power parameter range of the equipment to be adjusted is unevenly distributed, the energy load allocation scheme is optimized to obtain a balanced power adjustment vector, including: If the range of power parameters of the equipment to be adjusted does not meet the preset equalization standard, then the initial power distribution data is extracted from the current energy load allocation matrix; The initial power distribution data is iteratively optimized, the topology of the dynamic association graph between devices is dynamically updated, and the interaction strength score between each injection molding machine is calculated after each iteration. The power allocation is adjusted according to the interaction intensity score until power distribution data that meets the preset equalization standard is obtained, and the power distribution data is converted into an equalized power adjustment vector.

6. The energy-saving control method for injection molding machines based on multi-source data according to claim 1, characterized in that, The step of generating and updating control commands for multiple injection molding machines based on the balanced power adjustment vector includes: Based on the balanced power adjustment vector, the target power parameters of each injection molding machine are obtained; Determine if each injection molding machine is overloaded. If an overload is found, adjust the target power parameter of that injection molding machine. Based on the adjusted target power parameters, control commands are generated for multiple injection molding machines, and the control commands are sent to the corresponding injection molding machines to update the operating parameters of the injection molding machines.

7. An energy-saving control system for injection molding machines based on multi-source data, characterized in that, The method for implementing the energy-saving control of injection molding machines based on multi-source data as described in any one of claims 1 to 6 includes: The data acquisition and preprocessing module is used to acquire voltage and current data and network connection information of multiple injection molding machines, and preprocess the voltage and current data and network connection information to obtain a data set of equipment operating status. The correlation graph construction and strength calculation module is used to construct a dynamic correlation graph between devices based on the device operating status dataset and calculate the strength of the energy consumption interaction relationship between each injection molding machine. The chain path extraction module is used to extract high-intensity interaction paths and fuse the network connection relationship information to obtain a set of potential chain reaction paths if the energy consumption interaction relationship strength exceeds a preset energy consumption interaction strength threshold. The fluctuation prediction and judgment module is used to predict future voltage and current fluctuation trends based on the set of potential chain reaction paths, and to determine the equipment groups with excessive load during peak production periods. The load matrix processing module is used to obtain the current energy load distribution matrix based on the equipment group with excessive load during the peak production period, analyze the load evolution characteristics in the time dimension, and determine the range of equipment power parameters that need to be adjusted. The power optimization module is used to optimize the energy load allocation scheme to obtain a balanced power adjustment vector if the power parameter range of the equipment to be adjusted is unevenly distributed. The control command update module is used to generate and update control commands for multiple injection molding machines based on the balanced power adjustment vector.

8. An electronic device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the energy-saving control method for injection molding machines based on multi-source data as described in any one of claims 1 to 6.

Images