Mechanical and electrical system energy consumption optimization method and system based on big data

Abstract

The application discloses a mechanical and electrical system energy consumption optimization method and system based on big data, and the method comprises the following steps: collecting energy consumption data, equipment operation parameters, environmental parameters and working condition parameters of a mechanical and electrical system to obtain multidimensional data, preprocessing the multidimensional data to obtain initial multidimensional data, clustering the working conditions of the initial multidimensional data by using a fuzzy C-means clustering algorithm based on a chaotic sequence grey wolf optimization, reducing the dimension of the data by using a uniform manifold approximation and projection method to obtain standard multidimensional data, constructing a hybrid analysis model fusing a long short-term memory network and a random forest algorithm, wherein the long short-term memory network is used for learning the time sequence characteristics of the energy consumption of the mechanical and electrical system, and the random forest algorithm model is used for identifying key factors affecting the energy consumption, and inputting the standard multidimensional data into the hybrid analysis model to output an energy consumption prediction result. Under the premise of guaranteeing the safe operation of equipment and production requirements, the energy consumption of the mechanical and electrical system is effectively reduced, and the energy utilization efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for optimizing energy consumption in electromechanical systems based on big data. Background Technology

[0002] With the acceleration of industrial intelligence and urbanization, electromechanical systems have been widely used in industrial production, buildings, transportation hubs, and other fields, becoming the core carrier of energy consumption. Their energy consumption accounts for a growing proportion of total social energy consumption, making energy conservation and emission reduction a crucial issue for green development. Currently, energy consumption optimization in electromechanical systems often adopts traditional experience-driven models, focusing solely on energy consumption data while neglecting the synergistic effects of equipment operating parameters, environmental parameters, and operating condition parameters, resulting in incomplete data support. Traditional clustering algorithms are prone to getting trapped in local optima, failing to accurately classify energy consumption characteristics under different operating conditions, making it difficult to formulate targeted optimization strategies. Single time-series models struggle to balance the temporal nature of energy consumption with the correlation of influencing factors, leading to large energy consumption prediction errors and a lack of scientific rigor in optimization instructions. Control modes are mostly static optimizations, unable to dynamically adjust based on real-time data, making it difficult to adapt to the complex and ever-changing operating states of electromechanical systems, resulting in low energy efficiency and accuracy. Summary of the Invention

[0003] The purpose of this invention is to solve the above-mentioned problems by designing a method and system for optimizing the energy consumption of electromechanical systems based on big data.

[0004] To achieve the above objectives, the technical solution of the present invention further includes the following steps in the above-mentioned big data-based electromechanical system energy consumption optimization method: Collect energy consumption data, equipment operating parameters, environmental parameters and operating condition parameters of electromechanical systems to obtain multi-dimensional data, and preprocess the multi-dimensional data to obtain initial multi-dimensional data; The initial multidimensional data was clustered by working condition using the fuzzy C-means clustering algorithm TGWO-FCM based on the Tent chaotic sequence gray wolf optimization. Then, the data dimensionality was reduced by UMAP uniform manifold approximation and projection method to obtain standard multidimensional data. A hybrid analysis model integrating LSTM (Long Short-Term Memory) network and random forest algorithm is constructed. The LSTM network is used to learn the temporal characteristics of energy consumption of electromechanical system, and the random forest algorithm model is used to identify key factors affecting energy consumption. The standard multidimensional data is input into the hybrid analysis model, and the energy consumption prediction results are output. Based on the energy consumption prediction results, dynamic optimization instructions for electromechanical equipment are generated, and the dynamic optimization instructions are transmitted to the electromechanical system for real-time control.

[0005] Furthermore, in the above-mentioned big data-based electromechanical system energy consumption optimization method, the collection of energy consumption data, equipment operating parameters, environmental parameters, and operating condition parameters of the electromechanical system yields multi-dimensional data. This multi-dimensional data is then preprocessed to obtain initial multi-dimensional data, including: Outliers in multidimensional data are identified using statistical analysis, missing values ​​are filled in using linear interpolation, and duplicate and invalid data are removed to obtain cleaned multidimensional data. The Z-score standardization method is used to convert parameters of different dimensions and orders of magnitude in the cleaned multidimensional data into dimensionless data of a unified standard, thus obtaining standard multidimensional data. Standard multidimensional data is aligned by timestamp and integrated into a unified multidimensional dataset to obtain the initial multidimensional data.

[0006] Furthermore, in the aforementioned big data-based electromechanical system energy consumption optimization method, the initial multi-dimensional data is clustered according to operating conditions using the TGWO-FCM fuzzy C-means clustering algorithm based on Tent chaotic sequence gray wolf optimization, and then the data dimensionality is reduced by UMAP uniform manifold approximation and projection methods to obtain standard multi-dimensional data, including: The fuzzy C-means clustering algorithm TGWO-FCM based on the Tent chaotic sequence gray wolf optimization is adopted. The number of clusters is determined, the cluster centers and fuzzy coefficients of the FCM algorithm are initialized, the initial value and iteration number of the Tent chaotic sequence are set, and the population size and search range of the gray wolf algorithm are initialized. The initial population of the gray wolf algorithm is optimized by utilizing the randomness and ergodicity of the Tent chaotic sequence. The cluster center of the FCM algorithm is optimized by using the predation, encirclement and search mechanism of the gray wolf algorithm, thus obtaining the TGWO-FCM algorithm. The initial multi-dimensional data is input into the optimized TGWO-FCM algorithm. Through iterative calculation, clustering results for different working conditions are obtained. Each cluster corresponds to a typical working condition. The core data features of each working condition are labeled to obtain clustered data.

[0007] Furthermore, in the aforementioned big data-based electromechanical system energy consumption optimization method, the initial multi-dimensional data is clustered according to operating conditions using the TGWO-FCM fuzzy C-means clustering algorithm based on Tent chaotic sequence gray wolf optimization, and then the data dimensionality is reduced by UMAP uniform manifold approximation and projection methods to obtain standard multi-dimensional data, including: Set the nearest neighbor number and minimum distance for the UMAP algorithm, and determine the target dimension after data dimensionality reduction; Clustered data is input into the UMAP algorithm, which uses manifold learning to preserve local and global features of the data and maps high-dimensional data to a low-dimensional space to obtain standard multidimensional data.

[0008] Furthermore, in the aforementioned big data-based energy consumption optimization method for electromechanical systems, the construction of a hybrid analysis model integrating an LSTM (Long Short-Term Memory) network and a random forest algorithm is described. The LSTM network is used to learn the temporal characteristics of the electromechanical system's energy consumption, while the random forest algorithm model is used to identify key factors affecting energy consumption, including: The LSTM (Long Short-Term Memory) network is responsible for learning the temporal characteristics of energy consumption in electromechanical systems and capturing the patterns of energy consumption changes over time. The network structure includes an input layer, hidden layers, and an output layer. The number of neurons in the input layer is consistent with the dimension of the reduced features. There are 2-3 hidden layers with 32-128 neurons in each layer, using the ReLU activation function. The output layer outputs intermediate results of energy consumption time-series prediction.

[0009] Furthermore, in the aforementioned big data-based energy consumption optimization method for electromechanical systems, the construction of a hybrid analysis model integrating an LSTM (Long Short-Term Memory) network and a random forest algorithm is described. The LSTM network is used to learn the temporal characteristics of the electromechanical system's energy consumption, while the random forest algorithm model is used to identify key factors affecting energy consumption, including: The random forest algorithm is responsible for identifying key factors affecting the energy consumption of electromechanical systems. The input is all feature parameters in standard multidimensional data, and the output is the importance weight of each feature. Construct 100-200 decision trees, generate training samples for each decision tree using the bootstrap sampling method, use the Gini coefficient as the feature splitting criterion, calculate the influence weight of each feature on energy consumption, and screen out key influencing factors. The intermediate results of energy consumption time series prediction from the LSTM module are fused with the weights of key influencing factors identified by the random forest algorithm module to obtain the energy consumption prediction results.

[0010] Furthermore, in the above-mentioned big data-based electromechanical system energy consumption optimization method, the step of inputting the standard multi-dimensional data into a hybrid analysis model, outputting energy consumption prediction results, generating dynamic optimization instructions for electromechanical equipment based on the energy consumption prediction results, and transmitting the dynamic optimization instructions to the electromechanical system for real-time control includes: Based on the comparison between the energy consumption prediction results and the optimal energy consumption threshold, and combined with key influencing factors, dynamic optimization instructions are generated and converted into a standardized instruction format that can be recognized by the electromechanical system.

[0011] Furthermore, in the big data-based electromechanical system energy consumption optimization system, the electromechanical system energy consumption optimization system includes the following modules: The electromechanical data acquisition module is used to collect energy consumption data, equipment operating parameters, environmental parameters and operating condition parameters of the electromechanical system to obtain multi-dimensional data, and to preprocess the multi-dimensional data to obtain initial multi-dimensional data. The data clustering and dimensionality reduction module is used to perform working condition clustering on the initial multidimensional data using the TGWO-FCM fuzzy C-means clustering algorithm based on the Tent chaotic sequence gray wolf optimization, and then perform data dimensionality reduction through UMAP uniform manifold approximation and projection method to obtain standard multidimensional data. The hybrid model building module is used to build a hybrid analysis model that integrates the LSTM (Long Short-Term Memory) network and the random forest algorithm. The LSTM network is used to learn the temporal characteristics of energy consumption of electromechanical systems, and the random forest algorithm model is used to identify key factors affecting energy consumption. The system energy consumption optimization module is used to input the standard multi-dimensional data into the hybrid analysis model, output energy consumption prediction results, generate dynamic optimization instructions for electromechanical equipment based on the energy consumption prediction results, and transmit the dynamic optimization instructions to the electromechanical system for real-time control.

[0012] Furthermore, in the big data-based electromechanical system energy consumption optimization system, the electromechanical data acquisition module includes the following sub-modules: The cleaning submodule is used to identify outliers in multi-dimensional data based on statistical analysis, fill in missing values ​​using linear interpolation, and delete duplicate and invalid data to obtain cleaned multi-dimensional data. The transformation submodule is used to convert parameters of different dimensions and orders of magnitude in cleaned multidimensional data into dimensionless data of a unified standard using the Z-score standardization method, thus obtaining standard multidimensional data. The integration submodule is used to align standard multidimensional data by timestamp and integrate it into a unified multidimensional dataset to obtain the initial multidimensional data.

[0013] Furthermore, in the big data-based electromechanical system energy consumption optimization system, the data clustering and dimensionality reduction module includes the following sub-modules: The initialization submodule is used to determine the number of clusters, initialize the cluster centers and fuzzy coefficients of the FCM algorithm, set the initial value and number of iterations of the Tent chaotic sequence, and initialize the population size and search range of the gray wolf algorithm using the fuzzy C-means clustering algorithm based on the Tent chaotic sequence gray wolf optimization. The optimization submodule is used to optimize the initial population of the Grey Wolf algorithm by utilizing the randomness and ergodicity of the Tent chaotic sequence. Through the predation, encirclement and search mechanism of the Grey Wolf algorithm, the cluster center of the FCM algorithm is optimized to obtain the TGWO-FCM algorithm. The calculation submodule is used to input the initial multi-dimensional data into the optimized TGWO-FCM algorithm. Through iterative calculation, it obtains the clustering results of different working conditions. Each cluster corresponds to a typical working condition. The core data features of each working condition are labeled to obtain the clustered data.

[0014] Its beneficial effects lie in the fact that by collecting energy consumption, equipment operation, environmental and operating condition parameters from multiple dimensions, and combining cleaning, standardization and integration processing, it ensures the comprehensiveness, accuracy and usability of the data, providing solid data support for subsequent optimization analysis and solving the problems of single data dimension and poor data quality in traditional methods. The operating condition clustering and dimensionality reduction effects are excellent. The TGWO-FCM algorithm is used to optimize clustering accuracy, avoid local optima, accurately classify typical operating conditions, and combine it with UMAP dimensionality reduction technology to remove data redundancy, improving the efficiency of subsequent model processing while retaining core features, solving the problems of low clustering accuracy and analysis lag caused by data redundancy in traditional methods. The hybrid analysis model has accurate predictions, combining the advantages of LSTM and random forest, capturing the temporal variation pattern of energy consumption and accurately identifying key influencing factors, improving the accuracy of energy consumption prediction, providing a scientific basis for optimization instruction generation, and making up for the limitations of single models. It achieves dynamic closed-loop control, generates targeted optimization instructions based on prediction results, transmits them to the electromechanical system in real time and feeds back operating data, and continuously iterates and optimizes. Under the premise of ensuring safe operation of equipment and production needs, it effectively reduces the energy consumption of electromechanical system, improves energy utilization efficiency, extends equipment service life, and reduces operation and maintenance costs. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0016] Figure 1 This is a schematic diagram of the first embodiment of the energy consumption optimization method for electromechanical systems based on big data in this invention. Figure 2 This is a schematic diagram of the second embodiment of the energy consumption optimization method for electromechanical systems based on big data in this invention. Figure 3 This is a schematic diagram of the first embodiment of the electromechanical system energy consumption optimization system based on big data in this invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms "one," "an," and "this" used herein may also include the plural forms. It should be further understood that the terminology used in this specification includes the presence of features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0019] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 As shown, a big data-based energy consumption optimization method for electromechanical systems includes the following steps:

[0020] Step 101: Collect energy consumption data, equipment operating parameters, environmental parameters and operating condition parameters of the electromechanical system to obtain multi-dimensional data. Preprocess the multi-dimensional data to obtain initial multi-dimensional data. Specifically, in this embodiment, outliers in multidimensional data are identified based on statistical analysis, missing values ​​are filled in using linear interpolation, and duplicate and invalid data are deleted to obtain cleaned multidimensional data. The Z-score standardization method is used to convert parameters of different dimensions and orders of magnitude in the cleaned multidimensional data into dimensionless data of a unified standard to obtain standard multidimensional data. The standard multidimensional data is aligned by timestamp and integrated into a unified multidimensional dataset to obtain the initial multidimensional data.

[0021] It covers the entire electromechanical system chain, including core production equipment such as motors, water pumps, and fans, auxiliary control equipment, energy supply modules, and surrounding environmental monitoring areas; ① Energy consumption data: real-time power consumption, water consumption, and gas consumption of each device; peak / valley energy consumption per unit time; cumulative energy consumption, etc.; ② Equipment operating parameters: motor speed, voltage, current, power factor; inlet and outlet pressure and flow rate of water pumps / fans; equipment running time, number of start-stop cycles, fault alarm information, etc.; ③ Environmental parameters: ambient temperature, humidity, air pressure, light intensity, outdoor wind speed, etc.; ④ Operating condition parameters: system load rate, production cycle time, work shifts, equipment operating mode, manual / automatic, etc. The system adopts a three-level acquisition mode of "sensor + data acquisition unit + edge gateway". The sensor collects various parameters in real time, the data acquisition unit completes data aggregation and preliminary conversion, and the edge gateway realizes data encryption and transmission to avoid data loss. The data collection frequency should be dynamically adjusted according to the importance of the parameters. Energy consumption data and core equipment operating parameters, such as voltage and current, should be collected at a frequency of no less than once per minute. Environmental parameters and operating condition parameters should be collected at a frequency of no less than once per 10 minutes to ensure that the temporal change characteristics of the data are captured.

[0022] Statistical analysis is used to identify outliers, such as voltage and current data that exceed the normal operating range of the equipment. Interpolation methods, such as linear interpolation and polynomial interpolation, are used to fill in missing values ​​and delete duplicate and invalid data, such as garbled data caused by sensor failure. The Z-score standardization method is used to convert parameters of different dimensions and orders of magnitude, such as power consumption, temperature and speed, into dimensionless data of a unified standard, eliminating the influence of dimensions and facilitating subsequent algorithm calculations. After cleaning and standardizing, various types of data are aligned by timestamp and integrated into a unified multi-dimensional dataset. The data is labeled with the corresponding device number, collection time, and operating condition type to form initial multi-dimensional data, which is then stored in a big data database to provide a data foundation for subsequent cluster analysis.

[0023] Step 102: Use the fuzzy C-means clustering algorithm TGWO-FCM based on Tent chaotic sequence gray wolf optimization to perform working condition clustering on the initial multidimensional data, and then use UMAP uniform manifold approximation and projection method to reduce the data dimensionality to obtain standard multidimensional data. Specifically, this embodiment employs the TGWO-FCM fuzzy C-means clustering algorithm optimized based on the Tent chaotic sequence gray wolf algorithm. The number of clusters is determined, and the cluster centers and fuzzy coefficients of the FCM algorithm are initialized. The initial value and iteration count of the Tent chaotic sequence are set, and the population size and search range of the gray wolf algorithm are initialized. The initial population of the gray wolf algorithm is optimized using the randomness and ergodicity of the Tent chaotic sequence. Through the predation, encirclement, and search mechanisms of the gray wolf algorithm, the cluster centers of the FCM algorithm are optimized, resulting in the TGWO-FCM algorithm. The initial multi-dimensional data is input into the optimized TGWO-FCM algorithm, and through iterative calculation, clustering results for different working conditions are obtained. Each cluster corresponds to a typical working condition, and the core data features of each working condition are labeled to obtain the clustered data.

[0024] Set the nearest neighbor number and minimum distance for the UMAP algorithm, and determine the target dimension after data dimensionality reduction; input the clustered data into the UMAP algorithm, and through the manifold learning method, preserve the local and global features of the data, mapping the high-dimensional data to the low-dimensional space to obtain standard multi-dimensional data.

[0025] A fuzzy C-means clustering algorithm based on Tent chaotic sequence optimization (TGWO-FCM) is adopted to classify the initial multidimensional data according to the working conditions and determine the number of clusters. Based on the common working conditions of electromechanical systems, the optimal number of clusters is determined by combining the elbow rule, which is generally 3-8. The cluster centers and fuzzy coefficients of the FCM algorithm are initialized, the initial value and iteration number of the Tent chaotic sequence are set, and the population size and search range of the gray wolf algorithm are initialized. By leveraging the randomness and ergodicity of the Tent chaotic sequence, the initial population of the Grey Wolf algorithm is optimized to avoid the algorithm getting trapped in local optima and improve search accuracy. The clustering centers of the FCM algorithm are optimized through the predation, encirclement, and search mechanisms of the Grey Wolf algorithm. The initial multi-dimensional data is input into the optimized TGWO-FCM algorithm. Through iterative calculation, clustering results for different operating conditions are obtained. Each cluster corresponds to a typical operating condition, such as high load, low load, and standby conditions. The core data features of each type of operating condition are labeled. The silhouette coefficient method is used to verify the clustering effect. If the silhouette coefficient is greater than 0.7, it indicates that the clustering effect is good. If it is less than 0.5, the number of clusters and algorithm parameters are adjusted, and the clustering operation is re-executed until the requirements are met.

[0026] The UMAP uniform manifold approximation and projection method are used to reduce the dimensionality of the initial multi-dimensional data after clustering, thereby reducing data redundancy and improving the training efficiency and prediction accuracy of subsequent models. The specific operation is as follows. Set the number of nearest neighbors for the UMAP algorithm, typically 15-30, and the minimum distance, typically 0.1-0.5. Determine the target dimension after dimensionality reduction, and reduce it to 2-5 dimensions depending on the degree of data redundancy. The initial multidimensional data after clustering is input into the UMAP model. Through the manifold learning method, the local and global features of the data are preserved, and the high-dimensional data is mapped to a low-dimensional space to obtain the dimensionality-reduced dataset. Verify the feature retention rate of the dimensionality-reduced data, ensuring that the retention rate of core energy consumption features and operating condition features is not less than 90%. Remove abnormal data generated during the dimensionality reduction process, and match the dimensionality-reduced dataset with the operating condition clustering labels to obtain standard multi-dimensional data for subsequent training and testing of hybrid models.

[0027] Step 103: Construct a hybrid analysis model that integrates the LSTM (Long Short-Term Memory) network and the random forest algorithm. The LSTM network is used to learn the temporal characteristics of energy consumption of the electromechanical system, and the random forest algorithm model is used to identify key factors affecting energy consumption. Specifically, in this embodiment, the LSTM (Long Short-Term Memory) network is responsible for learning the temporal characteristics of the energy consumption of the electromechanical system and capturing the pattern of energy consumption changes over time. The network structure includes an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is consistent with the dimension of the reduced feature. The hidden layer has 2-3 layers, with 32-128 neurons in each layer, and uses the ReLU activation function. The output layer outputs the intermediate results of the temporal prediction of energy consumption.

[0028] The Random Forest algorithm is responsible for identifying key factors affecting the energy consumption of electromechanical systems. The input is all feature parameters in the standard multidimensional data, and the output is the importance weight of each feature. 100-200 decision trees are constructed. Training samples for each decision tree are generated by the bootstrap sampling method. The Gini coefficient is used as the feature splitting criterion to calculate the influence weight of each feature on energy consumption and screen out the key influencing factors. The intermediate results of energy consumption time series prediction from the LSTM module are fused with the weights of the key influencing factors identified by the Random Forest algorithm module to obtain the energy consumption prediction result.

[0029] The LSTM (Long Short-Term Memory) network module is responsible for learning the temporal characteristics of energy consumption in electromechanical systems and capturing the patterns of energy consumption changes over time. The input consists of time-related parameters from standard multi-dimensional data, such as historical energy consumption data and time-series sequences of equipment operating parameters. The network structure includes an input layer, hidden layers, and an output layer. The number of neurons in the input layer is consistent with the dimension of the reduced features. The hidden layer has 2-3 layers, with 32-128 neurons in each layer, using the ReLU activation function. The output layer outputs intermediate results of energy consumption temporal prediction. By using the forget gate, input gate, and output gate of LSTM, the vanishing gradient problem in training with long sequence data is solved, accurately capturing the long-term temporal dependencies of energy consumption.

[0030] The random forest model module is responsible for identifying key factors affecting the energy consumption of electromechanical systems. The input is all feature parameters in the standard multi-dimensional data, including time-series features, operating condition features, and environmental features. The output is the importance weight of each feature. 100-200 decision trees are constructed. Training samples for each decision tree are generated by the bootstrap sampling method. The Gini coefficient is used as the feature splitting criterion to calculate the influence weight of each feature on energy consumption and screen out key influencing factors, such as equipment load rate, ambient temperature, and motor speed.

[0031] The intermediate time-series prediction results of the LSTM module are fused with the weights of key factors identified by the random forest module, and a weighted summation method is used to obtain the final energy consumption prediction result. The fusion weights are determined through cross-validation to ensure the prediction accuracy of the model.

[0032] Standard multidimensional data was divided into training, validation, and test sets in a 7:2:1 ratio for model training, parameter tuning, and performance validation. The LSTM module was trained using the training set data with the Adam optimizer, a learning rate of 0.001-0.01, and 100-200 iterations. Mean squared error (MSE) was used as the loss function, and the network parameters were adjusted through backpropagation until the loss function of the model converged on the validation set. The random forest module is trained using the training set data. Parameters such as the number of decision trees, maximum depth, and minimum number of sample splits are adjusted. The feature recognition accuracy of the model is verified using the validation set to ensure that the accuracy of key factor recognition is not less than 85%. Hybrid model fusion training involves training both modules and then adjusting the fusion weights based on the validation set data to minimize the prediction error of the hybrid model, thus completing the model training.

[0033] The grid search method is used to optimize the key parameters of the hybrid model, including the number of hidden layer neurons and learning rate of LSTM, the number of decision trees and maximum depth of random forest, and fusion weights. The optimal parameter combination is selected through cross-validation to ensure the model's generalization ability and prediction accuracy.

[0034] Step 104: Input the standard multi-dimensional data into the hybrid analysis model, output the energy consumption prediction results, generate dynamic optimization instructions for electromechanical equipment based on the energy consumption prediction results, and transmit the dynamic optimization instructions to the electromechanical system for real-time control.

[0035] Specifically, in this embodiment, based on the comparison between the energy consumption prediction results and the optimal energy consumption threshold, and combined with key influencing factors, dynamic optimization instructions are generated, and the dynamic optimization instructions are converted into a standardized instruction format that can be recognized by the electromechanical system.

[0036] Standard multi-dimensional data is input into the trained hybrid analysis model in real time. The model outputs the energy consumption prediction results of the electromechanical system for a future period, such as 10 minutes or 1 hour, including real-time energy consumption prediction, peak energy consumption prediction, and energy consumption trend. At the same time, it outputs the key factors and weights that affect the current energy consumption, providing a basis for optimizing instruction generation.

[0037] The system compares predicted energy consumption with actual energy consumption data in real time, calculates prediction error (MAE) and RMSE, and adjusts model parameters and retrains the model in a timely manner if the error exceeds a preset threshold, such as 5%, to ensure prediction accuracy. Based on the prediction results, the system analyzes energy consumption trends and identifies energy consumption anomalies, such as a sudden increase in predicted energy consumption, and provides early warnings to optimize control.

[0038] Based on energy consumption prediction results, key influencing factors, and the operational constraints of electromechanical systems, such as equipment rated parameters and production requirements, targeted dynamic optimization instructions are generated to minimize the energy consumption of electromechanical systems, reduce energy consumption peaks, and achieve energy consumption balance under the premise of meeting production requirements and ensuring safe equipment operation. Based on the comparison between predicted energy consumption and the optimal energy consumption threshold, and combined with key influencing factors, specific instructions are generated. For example, when the predicted energy consumption is too high and the equipment load rate is too high as the main influencing factor, an instruction to "reduce the equipment operating load rate to a reasonable range" is generated; when the ambient temperature is too high, causing the air conditioning system energy consumption to increase, an instruction to "adjust the air conditioning set temperature and optimize the fan speed" is generated; when the equipment is in a low-load, high-energy-consumption state, an instruction to "adjust the equipment operating mode and merge work shifts" is generated. The generated optimized instructions are converted into a standardized instruction format that can be recognized by the electromechanical system, such as Modbus protocol instructions, specifying the instruction execution object, execution parameters, and execution time to ensure that the instructions can be directly issued.

[0039] Dynamic optimization commands are transmitted to the control terminal of the electromechanical system via communication modules such as industrial Ethernet and 5G. The control terminal executes the commands, adjusts the equipment operating status, and realizes real-time control of energy consumption optimization. At the same time, the actual energy consumption data and equipment operating parameters after the electromechanical system executes the commands are collected in real time and fed back to the big data database for subsequent iterative optimization of the model, forming a closed-loop control and continuously improving the energy consumption optimization effect.

[0040] Its beneficial effects lie in the fact that by collecting energy consumption, equipment operation, environmental and operating condition parameters from multiple dimensions, and combining cleaning, standardization and integration processing, it ensures the comprehensiveness, accuracy and usability of the data, providing solid data support for subsequent optimization analysis and solving the problems of single data dimension and poor data quality in traditional methods. The operating condition clustering and dimensionality reduction effects are excellent. The TGWO-FCM algorithm is used to optimize clustering accuracy, avoid local optima, accurately classify typical operating conditions, and combine UMAP dimensionality reduction technology to remove data redundancy, improving the efficiency of subsequent model processing while retaining core features, solving the problems of low clustering accuracy and analysis lag caused by data redundancy in traditional methods. The hybrid analysis model has accurate predictions, combining the advantages of LSTM and random forest, capturing the temporal variation pattern of energy consumption and accurately identifying key influencing factors, improving energy consumption prediction accuracy by 10%, providing a scientific basis for optimization instruction generation and making up for the limitations of single models. It achieves dynamic closed-loop control, generates targeted optimization instructions based on prediction results, transmits them to the electromechanical system in real time and feeds back operating data, and continuously iterates and optimizes. Under the premise of ensuring safe operation of equipment and production needs, it effectively reduces the energy consumption of the electromechanical system by 10%, improves energy utilization efficiency by 29%, extends the service life of equipment, and reduces operation and maintenance costs.

[0041] Please see Figure 2 In the energy consumption optimization method for electromechanical systems based on big data, the TGWO-FCM fuzzy C-means clustering algorithm based on Tent chaotic sequence gray wolf optimization is used to perform working condition clustering on the initial multidimensional data. Then, the data dimensionality is reduced by UMAP uniform manifold approximation and projection methods to obtain standard multidimensional data. The steps include the following steps.

[0042] Step 201: Using the fuzzy C-means clustering algorithm TGWO-FCM based on the Tent chaotic sequence and optimized by the gray wolf algorithm, determine the number of clusters, initialize the cluster centers and fuzzy coefficients of the FCM algorithm, set the initial value and iteration number of the Tent chaotic sequence, and initialize the population size and search range of the gray wolf algorithm. Step 202: Optimize the initial population of the gray wolf algorithm by utilizing the randomness and ergodicity of the Tent chaotic sequence. Optimize the cluster centers of the FCM algorithm through the predation, encirclement, and search mechanisms of the gray wolf algorithm to obtain the TGWO-FCM algorithm. Step 203: Input the initial multi-dimensional data into the optimized TGWO-FCM algorithm. Through iterative calculation, obtain the clustering results of different working conditions. Each cluster corresponds to a typical working condition. Label the core data features of each working condition to obtain the clustered data.

[0043] The above describes embodiments of the energy consumption optimization method for electromechanical systems based on big data according to the present invention. Please refer to [link / reference]. Figure 3 In the big data-based electromechanical system energy consumption optimization system, the electromechanical system energy consumption optimization system includes the following modules: The electromechanical data acquisition module is used to collect energy consumption data, equipment operating parameters, environmental parameters and operating condition parameters of electromechanical systems to obtain multi-dimensional data. The multi-dimensional data is preprocessed to obtain initial multi-dimensional data. The data clustering and dimensionality reduction module is used to perform working condition clustering on the initial multidimensional data using the TGWO-FCM fuzzy C-means clustering algorithm based on the Tent chaotic sequence gray wolf optimization, and then perform data dimensionality reduction through UMAP uniform manifold approximation and projection method to obtain standard multidimensional data. The hybrid model building module is used to build a hybrid analysis model that integrates the LSTM (Long Short-Term Memory) network and the random forest algorithm. The LSTM network is used to learn the temporal characteristics of energy consumption of electromechanical systems, and the random forest algorithm model is used to identify key factors affecting energy consumption. The system energy consumption optimization module is used to input standard multi-dimensional data into the hybrid analysis model, output energy consumption prediction results, generate dynamic optimization instructions for electromechanical equipment based on the energy consumption prediction results, and transmit the dynamic optimization instructions to the electromechanical system for real-time control.

[0044] Its beneficial effects lie in the fact that by collecting energy consumption, equipment operation, environmental and operating condition parameters from multiple dimensions, and combining cleaning, standardization and integration processing, it ensures the comprehensiveness, accuracy and usability of the data, providing solid data support for subsequent optimization analysis and solving the problems of single data dimension and poor data quality in traditional methods. The operating condition clustering and dimensionality reduction effects are excellent. The TGWO-FCM algorithm is used to optimize clustering accuracy, avoid local optima, accurately classify typical operating conditions, and combine UMAP dimensionality reduction technology to remove data redundancy, improving the efficiency of subsequent model processing while retaining core features, solving the problems of low clustering accuracy and analysis lag caused by data redundancy in traditional methods. The hybrid analysis model has accurate predictions, combining the advantages of LSTM and random forest, capturing the temporal variation pattern of energy consumption and accurately identifying key influencing factors, improving energy consumption prediction accuracy by 10%, providing a scientific basis for optimization instruction generation and making up for the limitations of single models. It achieves dynamic closed-loop control, generates targeted optimization instructions based on prediction results, transmits them to the electromechanical system in real time and feeds back operating data, and continuously iterates and optimizes. Under the premise of ensuring safe operation of equipment and production needs, it effectively reduces the energy consumption of the electromechanical system by 10%, improves energy utilization efficiency by 29%, extends the service life of equipment, and reduces operation and maintenance costs.

[0045] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended technical solutions and equivalents.

Claims

1. A method for optimizing energy consumption in electromechanical systems based on big data, characterized in that, The electromechanical system energy consumption optimization method includes the following steps: Collect energy consumption data, equipment operating parameters, environmental parameters and operating condition parameters of electromechanical systems to obtain multi-dimensional data, and preprocess the multi-dimensional data to obtain initial multi-dimensional data; The initial multidimensional data was clustered by working condition using the fuzzy C-means clustering algorithm TGWO-FCM based on the Tent chaotic sequence gray wolf optimization. Then, the data dimensionality was reduced by UMAP uniform manifold approximation and projection method to obtain standard multidimensional data. A hybrid analysis model integrating LSTM (Long Short-Term Memory) network and random forest algorithm is constructed. The LSTM network is used to learn the temporal characteristics of energy consumption of electromechanical system, and the random forest algorithm model is used to identify key factors affecting energy consumption. The standard multidimensional data is input into the hybrid analysis model, and the energy consumption prediction results are output. Based on the energy consumption prediction results, dynamic optimization instructions for electromechanical equipment are generated, and the dynamic optimization instructions are transmitted to the electromechanical system for real-time control.

2. The energy consumption optimization method for electromechanical systems based on big data as described in claim 1, characterized in that, The energy consumption data, equipment operating parameters, environmental parameters, and operating condition parameters of the collected electromechanical system are used to obtain multi-dimensional data. This multi-dimensional data is then preprocessed to obtain initial multi-dimensional data, including: Outliers in multidimensional data are identified using statistical analysis, missing values ​​are filled in using linear interpolation, and duplicate and invalid data are removed to obtain cleaned multidimensional data. The Z-score standardization method is used to convert parameters of different dimensions and orders of magnitude in the cleaned multidimensional data into dimensionless data of a unified standard, thus obtaining standard multidimensional data. Standard multidimensional data is aligned by timestamp and integrated into a unified multidimensional dataset to obtain the initial multidimensional data.

3. The energy consumption optimization method for electromechanical systems based on big data as described in claim 1, characterized in that, The initial multidimensional data is clustered according to working conditions using the TGWO-FCM fuzzy C-means clustering algorithm based on Tent chaotic sequence gray wolf optimization. Then, the data dimensionality is reduced using UMAP uniform manifold approximation and projection methods to obtain standard multidimensional data, including: The fuzzy C-means clustering algorithm TGWO-FCM based on the Tent chaotic sequence gray wolf optimization is adopted. The number of clusters is determined, the cluster centers and fuzzy coefficients of the FCM algorithm are initialized, the initial value and iteration number of the Tent chaotic sequence are set, and the population size and search range of the gray wolf algorithm are initialized. The initial population of the gray wolf algorithm is optimized by utilizing the randomness and ergodicity of the Tent chaotic sequence. The cluster center of the FCM algorithm is optimized by using the predation, encirclement and search mechanism of the gray wolf algorithm, thus obtaining the TGWO-FCM algorithm. The initial multi-dimensional data is input into the optimized TGWO-FCM algorithm. Through iterative calculation, clustering results for different working conditions are obtained. Each cluster corresponds to a typical working condition. The core data features of each working condition are labeled to obtain clustered data.

4. The energy consumption optimization method for electromechanical systems based on big data as described in claim 3, characterized in that, The initial multidimensional data is clustered according to working conditions using the TGWO-FCM fuzzy C-means clustering algorithm based on Tent chaotic sequence gray wolf optimization. Then, the data dimensionality is reduced using UMAP uniform manifold approximation and projection methods to obtain standard multidimensional data, including: Set the nearest neighbor number and minimum distance for the UMAP algorithm, and determine the target dimension after data dimensionality reduction; Clustered data is input into the UMAP algorithm, which uses manifold learning to preserve local and global features of the data and maps high-dimensional data to a low-dimensional space to obtain standard multidimensional data.

5. The energy consumption optimization method for electromechanical systems based on big data as described in claim 3, characterized in that, The proposed hybrid analysis model integrates an LSTM (Long Short-Term Memory) network and a random forest algorithm. The LSTM network is used to learn the temporal characteristics of energy consumption in the electromechanical system, while the random forest algorithm is used to identify key factors affecting energy consumption, including: The LSTM (Long Short-Term Memory) network is responsible for learning the temporal characteristics of energy consumption in electromechanical systems and capturing the patterns of energy consumption changes over time. The network structure includes an input layer, hidden layers, and an output layer. The number of neurons in the input layer is consistent with the dimension of the reduced features. There are 2-3 hidden layers with 32-128 neurons in each layer, using the ReLU activation function. The output layer outputs intermediate results of energy consumption time-series prediction.

6. The energy consumption optimization method for electromechanical systems based on big data as described in claim 3, characterized in that, The proposed hybrid analysis model integrates an LSTM (Long Short-Term Memory) network and a random forest algorithm. The LSTM network is used to learn the temporal characteristics of energy consumption in the electromechanical system, while the random forest algorithm is used to identify key factors affecting energy consumption, including: The random forest algorithm is responsible for identifying key factors affecting the energy consumption of electromechanical systems. The input is all feature parameters in standard multidimensional data, and the output is the importance weight of each feature. Construct 100-200 decision trees, generate training samples for each decision tree using the bootstrap sampling method, use the Gini coefficient as the feature splitting criterion, calculate the influence weight of each feature on energy consumption, and screen out key influencing factors. The intermediate results of energy consumption time series prediction from the LSTM module are fused with the weights of key influencing factors identified by the random forest algorithm module to obtain the energy consumption prediction results.

7. The energy consumption optimization method for electromechanical systems based on big data as described in claim 3, characterized in that, The process of inputting the standard multi-dimensional data into a hybrid analysis model, outputting energy consumption prediction results, generating dynamic optimization instructions for electromechanical equipment based on the energy consumption prediction results, and transmitting the dynamic optimization instructions to the electromechanical system for real-time control includes: Based on the comparison between the energy consumption prediction results and the optimal energy consumption threshold, and combined with key influencing factors, dynamic optimization instructions are generated and converted into a standardized instruction format that can be recognized by the electromechanical system.

8. A big data-based electromechanical system energy consumption optimization system, characterized in that, The electromechanical system energy consumption optimization system includes the following modules: The electromechanical data acquisition module is used to collect energy consumption data, equipment operating parameters, environmental parameters and operating condition parameters of the electromechanical system to obtain multi-dimensional data, and to preprocess the multi-dimensional data to obtain initial multi-dimensional data. The data clustering and dimensionality reduction module is used to perform working condition clustering on the initial multidimensional data using the TGWO-FCM fuzzy C-means clustering algorithm based on the Tent chaotic sequence gray wolf optimization, and then perform data dimensionality reduction through UMAP uniform manifold approximation and projection method to obtain standard multidimensional data. The hybrid model building module is used to build a hybrid analysis model that integrates the LSTM (Long Short-Term Memory) network and the random forest algorithm. The LSTM network is used to learn the temporal characteristics of energy consumption of electromechanical systems, and the random forest algorithm model is used to identify key factors affecting energy consumption. The system energy consumption optimization module is used to input the standard multi-dimensional data into the hybrid analysis model, output energy consumption prediction results, generate dynamic optimization instructions for electromechanical equipment based on the energy consumption prediction results, and transmit the dynamic optimization instructions to the electromechanical system for real-time control.

9. The electromechanical system energy consumption optimization system based on big data as described in claim 8, characterized in that, The electromechanical data acquisition module includes the following sub-modules: The cleaning submodule is used to identify outliers in multi-dimensional data based on statistical analysis, fill in missing values ​​using linear interpolation, and delete duplicate and invalid data to obtain cleaned multi-dimensional data. The transformation submodule is used to convert parameters of different dimensions and orders of magnitude in cleaned multidimensional data into dimensionless data of a unified standard using the Z-score standardization method, thus obtaining standard multidimensional data. The integration submodule is used to align standard multidimensional data by timestamp and integrate it into a unified multidimensional dataset to obtain the initial multidimensional data.

10. The electromechanical system energy consumption optimization system based on big data as described in claim 8, characterized in that, The data clustering and dimensionality reduction module includes the following sub-modules: The initialization submodule is used to determine the number of clusters, initialize the cluster centers and fuzzy coefficients of the FCM algorithm, set the initial value and number of iterations of the Tent chaotic sequence, and initialize the population size and search range of the gray wolf algorithm using the fuzzy C-means clustering algorithm based on the Tent chaotic sequence gray wolf optimization. The optimization submodule is used to optimize the initial population of the Grey Wolf algorithm by utilizing the randomness and ergodicity of the Tent chaotic sequence. Through the predation, encirclement and search mechanism of the Grey Wolf algorithm, the cluster center of the FCM algorithm is optimized to obtain the TGWO-FCM algorithm. The calculation submodule is used to input the initial multi-dimensional data into the optimized TGWO-FCM algorithm. Through iterative calculation, it obtains the clustering results of different working conditions. Each cluster corresponds to a typical working condition. The core data features of each working condition are labeled to obtain the clustered data.

Images