Method and device for generating control information of energy consumption resources and related equipment

Abstract

The invention provides an energy consumption resource control information generation method and device and related equipment, and the method comprises the steps: carrying out the feature extraction of energy consumption resource data in a target time period, and obtaining a plurality of pieces of feature data; evaluating the multiple pieces of feature data to obtain perception information, and inputting the multiple pieces of feature data into a deep learning model for prediction to obtain prediction information; and generating control information of the energy consumption resource based on the perception information and the prediction information. According to the method, the multiple pieces of feature data are obtained after feature extraction is performed on the energy consumption resource data in the target time period, so that the multiple pieces of feature data are evaluated and predicted, the perception information and the prediction information are obtained respectively, and the control information of the energy consumption resources is generated according to the perception information and the prediction information; therefore, the generated control information of the energy consumption resource is more fit with the energy consumption resource data, the comprehensive analysis of the control information of the energy consumption resource on the energy consumption system is realized, and the accuracy of the control information of the energy consumption resource is improved.

Description

Technical Field

[0001] The present invention relates to the field of Internet technology, and in particular to a method and device for generating control information of energy consumption resources and related equipment. Background Art

[0002] Energy and resource situational awareness involves the global, dynamic monitoring and assessment of energy (such as electricity, natural gas, and heat) and resource (such as water and raw materials) consumption through real-time monitoring, data collection, and analysis. It aims to comprehensively understand resource utilization efficiency, identify anomalies or waste, and predict future trends, thereby providing data support for optimizing resource allocation, improving energy efficiency, reducing operating costs, and achieving sustainable development goals.

[0003] The existing control information of energy consumption resources can only control single energy consumption data, and is difficult to adapt to complex and changeable energy consumption data, resulting in the problem of low accuracy of the control information of energy consumption resources generated in the existing technology. Summary of the Invention

[0004] The present invention provides a method and apparatus for generating control information of energy consumption resources and related equipment, which solve the problem of low accuracy of control information of energy consumption resources generated in the prior art.

[0005] To solve the above problems, the present invention is achieved as follows:

[0006] In a first aspect, an embodiment of the present application provides a method for generating control information of energy consumption resources, the method comprising:

[0007] Extracting features from energy consumption resource data during a target time period to obtain a plurality of feature data, wherein the energy consumption resource data is operating status data of the power system during the target time period, and different feature data correspond to different feature types.

[0008] Evaluating the plurality of feature data to obtain perception information, and inputting the plurality of feature data into a deep learning model for prediction to obtain prediction information, wherein the perception information is used to indicate a peak energy consumption of the energy consumption resource data after the target time period, and the prediction information is used to indicate a change trend of the energy consumption resource data after the target time period;

[0009] Based on the perception information and the prediction information, control information of energy consumption resources is generated, where the control information of energy consumption resources is used to control consumption of energy consumption resources of the power system after the target time period.

[0010] Optionally, before extracting features from the energy consumption resource data in the target time period to obtain a plurality of feature data, the method further includes:

[0011] Obtain multiple distributed energy consumption resource data within the target time period, where the power consumption sources of each distributed energy consumption resource data among the multiple distributed energy consumption resource data are different;

[0012] Perform data cleaning on the multiple distributed energy consumption resource data to obtain the cleaned multiple distributed energy consumption resource data, where the data cleaning is used to delete duplicate data and error data in the multiple distributed energy consumption resource data;

[0013] Perform data conversion on the cleaned multiple distributed energy consumption resource data to obtain the energy consumption resource data, where the data conversion includes at least one of the following: logarithmic transformation, detrending, and differencing operation.

[0014] Optionally, the method for extracting features from the energy consumption resource data of the target time period to obtain multiple feature data includes:

[0015] Construct an energy consumption time series model based on the physical characteristics and business requirements of the energy consumption resource data, where the energy consumption time series model is used to extract features from the input data based on the physical characteristics and the business requirements;

[0016] Input the energy consumption resource data of the target time period into the energy consumption time series model for feature extraction to obtain the multiple feature data, where the feature dimensions corresponding to the multiple feature data include at least one of the following: load rate feature of electrical equipment, seasonal electricity consumption feature, electrical equipment type feature, and electrical equipment usage time feature.

[0017] Optionally, before inputting the multiple feature data into a deep learning model for prediction to obtain prediction information, the method further includes:

[0018] Establish an initial model, where the initial model includes a long short-term memory network and a convolutional neural network, where the long short-term memory network is used to identify the time series features of the multiple feature data, and the convolutional neural network is used to identify the spatial features of the multiple feature data;

[0019] Update the initial model based on the attention mechanism and the Transformer architecture to obtain the deep learning model.

[0020] Optionally, the method for evaluating the multiple feature data to obtain perception information includes:

[0021] Obtain multiple information evaluation parameters and multiple quantization evaluation parameters, where the multiple information evaluation parameters are used to represent different information evaluation dimensions, and the multiple quantization evaluation parameters are used to represent different quantization evaluation dimensions;

[0022] Generate an information evaluation index based on the multiple information evaluation parameters, and generate a quantization evaluation index based on the multiple quantization evaluation parameters;

[0023] Evaluate the multiple feature data based on the information evaluation index and the quantization evaluation index to obtain the perception information.

[0024] Optionally, the evaluating the multiple feature data based on the information evaluation index and the quantization evaluation index to obtain the perception information includes:

[0025] Evaluate the network security of the multiple feature data to obtain multiple network security evaluation information, where the multiple feature data corresponds one-to-one with the multiple network security evaluation information, and the network security evaluation information is used to indicate whether the network security of the corresponding feature data meets the preset security requirements;

[0026] Filter the multiple feature data based on the multiple network security evaluation information to obtain multiple target feature data, and the network security evaluation results corresponding to the multiple target feature data meet the preset security requirements;

[0027] Evaluate the multiple target feature data based on the information evaluation index and the quantization evaluation index to obtain the perception information.

[0028] Optionally, the generating control information of energy consumption resources based on the perception information and the prediction information includes:

[0029] Evaluate the multiple feature data based on a data fusion quality index to obtain data evaluation information, where the data fusion quality index is used to measure the consistency and integrity after data fusion of the multiple feature data;

[0030] Generate the control information of the energy consumption resources based on the perception information, the prediction information, and the data evaluation information.

[0031] Second invention. An embodiment of the present application provides a device for generating control information of energy consumption resources, and the device includes:

[0032] An extraction module, configured to extract features from energy consumption resource data in a target time period to obtain multiple feature data, where the energy consumption resource data is the operation state data of the power system in the target time period, and among the multiple feature data, different feature data corresponds to different feature types;

[0033] An evaluation module, configured to evaluate the multiple feature data to obtain perception information, and input the multiple feature data into a deep learning model for prediction to obtain prediction information, where the perception information is used to indicate the energy consumption peak of the energy consumption resource data after the target time period, and the prediction information is used to indicate the change trend of the energy consumption resource data after the target time period;

[0034] A generation module, configured to generate control information for the energy consumption resource based on the perception information and the prediction information, where the control information for the energy consumption resource is used to control the energy consumption resource consumption of the power system after the target time period.

[0035] In a third aspect, the present application further provides an electronic device, including a processor, a memory, and a computer program stored on the memory and executable on the processor, where when the computer program is executed by the processor, the steps in the method described in the first aspect above are implemented.

[0036] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored, where when the computer program is executed by a processor, the steps in the method described in the first aspect above are implemented.

[0037] In a fifth aspect, the present application further provides a computer program product, including computer instructions, where when the computer instructions are executed by a processor, the steps in the method described in the first aspect above are implemented.

[0038] The present application provides a method, an apparatus and related devices for generating control information of energy consumption resources. The method includes: extracting features from the energy consumption resource data of a target time period to obtain a plurality of feature data, where the energy consumption resource data is the operation status data of the power system during the target time period, and among the plurality of feature data, different feature data correspond to different feature types; evaluating the plurality of feature data to obtain perception information, and inputting the plurality of feature data into a deep learning model for prediction to obtain prediction information, where the perception information is used to indicate the energy consumption peak value of the energy consumption resource data after the target time period, and the prediction information is used to indicate the change trend of the energy consumption resource data after the target time period; generating control information of the energy consumption resources based on the perception information and the prediction information, where the control information of the energy consumption resources is used to control the energy consumption resource consumption of the power system after the target time period. By extracting features from the energy consumption resource data of the target time period, the present application obtains a plurality of feature data, and then evaluates and predicts the plurality of feature data to obtain perception information and prediction information respectively. Thus, the control information of the energy consumption resources is generated according to the perception information and the prediction information, so that the generated control information of the energy consumption resources is more in line with the energy consumption resource data, realizing the all-round analysis of the energy consumption system by the control information of the energy consumption resources and improving the accuracy of the control information of the energy consumption resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] To more clearly illustrate the technical solutions of the present invention, the following will briefly introduce the drawings required for the description of the present invention. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0040] Figure 1 It is a schematic flowchart of a method for generating control information of energy consumption resources provided by an embodiment of the present application;

[0041] Figure 2 It is a working flowchart provided by an embodiment of the present application;

[0042] Figure 3 It is a time series analysis diagram of energy consumption data provided by an embodiment of the present application;

[0043] Figure 4 It is a schematic diagram of an evaluation index structure provided by an embodiment of the present application;

[0044] Figure 5 It is a schematic diagram of network security detection provided by an embodiment of the present application;

[0045] Figure 6 It is a schematic diagram of the control information of the energy consumption resources provided by an embodiment of the present application;

[0046] Figure 7 This is a schematic structural diagram of a control information generation device for energy consumption resources provided by an embodiment of the present application;

[0047] Figure 8 This is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0048] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0049] Terms such as "first" and "second" in the embodiments of the present application are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices. In addition, the use of "and / or" in the present application means at least one of the connected objects. For example, A and / or B and / or C means including A alone, B alone, C alone, as well as the existence of both A and B, the existence of both B and C, the existence of both A and C, and the existence of all of A, B, and C, a total of 7 cases.

[0050] See Figure 1 , Figure 1 This is a schematic flowchart of a method for generating control information of energy consumption resources provided by an embodiment of the present application.

[0051] As Figure 1 shown, the method for generating control information of energy consumption resources may include the following steps:

[0052] Step 101: Extract features from the energy consumption resource data in the target time period to obtain a plurality of feature data. The energy consumption resource data is the operation state data of the power system in the target time period. Among the plurality of feature data, different feature data corresponds to different feature types.

[0053] In this embodiment, the energy consumption resource data can be power, gas, water resource consumption data, and the operation status data of related equipment. Exemplarily, multi-source data can be obtained from various security devices such as an energy monitoring system, an intelligent electricity meter, and a security protection device. For example, the energy monitoring system provides real-time monitoring data of the overall energy consumption, the intelligent electricity meter accurately records the details of power consumption, and the security protection device provides energy consumption-related security status information, etc. The target time period can be the current operation time period of the energy system. For example, the energy consumption resource data of the energy system for one hour or one day is obtained. Specifically, it is not specifically limited in this embodiment.

[0054] After the energy consumption resource data is obtained, feature extraction is performed on the energy consumption resource data to obtain a plurality of feature data, where different feature data correspond to different feature types. For example, they correspond to power, gas, water resource consumption data, etc. The feature data reflects a plurality of different feature types corresponding to the energy consumption resource data.

[0055] In some embodiments, the features may include lag features, time series features, and endogenous features. Among them, the lag feature uses the autoregressive idea to generate features based on past values. The time series feature extracts features from time information (such as hours, weeks, months, etc.). The endogenous feature generates features based on the data of other columns (such as the energy consumption of a fan).

[0056] Step 102: Evaluate the plurality of feature data to obtain perception information, and input the plurality of feature data into a deep learning model for prediction to obtain prediction information. The perception information is used to indicate the energy consumption peak value of the energy consumption resource data after the target time period, and the prediction information is used to indicate the change trend of the energy consumption resource data after the target time period.

[0057] In this embodiment, the perception information is used to indicate the energy consumption peak value of the energy consumption resource data after the target time period, such as the highest value and the lowest value that the energy consumption resource data will appear in the target time period, etc. The prediction information is used to indicate the change trend of the energy consumption resource data after the target time period, such as increase or decrease, etc.

[0058] Among them, when evaluating multiple feature data, it can be evaluated according to multiple evaluation indicators. For example, multiple evaluation indicators can be set in advance, and multiple feature data are evaluated according to multiple evaluation indicators. The evaluation indicators can be information evaluation indicators, quantitative evaluation indicators, etc. When predicting multiple feature data, it can be predicted through a deep learning model. The deep learning model can be an XGBoost model. The XGBoost model is an efficient and flexible machine learning algorithm, which belongs to an improved version of gradient boosting decision trees. It constructs a strong learner by integrating multiple weak learners (usually decision trees) and performs excellently in prediction tasks of structured data. It is widely used in problems such as classification, regression, and ranking. Through the XGBoost model, the change trend of energy consumption resource data after the target time period can be predicted.

[0059] Step 103: Generate control information for energy consumption resources based on the perception information and the prediction information. The control information for energy consumption resources is used to control the consumption of energy consumption resources by the power system after the target time period.

[0060] In this embodiment, by comprehensively considering the perception information and the prediction information, control information for energy consumption resources is generated to control the consumption of energy consumption resources after the target time period. Among them, the control information for energy consumption resources is the control strategy for energy consumption resources. Exemplarily, when it is predicted that an energy consumption peak is approaching and the current load of the system is close to saturation, the control strategy for energy consumption resources automatically generates an equipment scheduling plan, starts standby energy equipment in advance or adjusts the energy distribution strategy to ensure the stability and security of energy supply. At the same time, according to different scenarios and requirements, multiple alternative plans are formulated, and evaluated and selected in combination with the actual situation to ensure the rationality and effectiveness of the decision-making.

[0061] In this embodiment, after generating the control information of the energy consumption resources, the control information of the energy consumption resources is executed and displayed. By strengthening the visualization display and interaction functions, an intuitive and convenient energy consumption situation perception and management interface is provided for users. Through forms such as visual charts and maps, the distribution, trends, correlation relationships, and situation assessment results of multi-source heterogeneous energy consumption data are displayed. Users can, through interactive operations, select different time intervals, regional ranges, data types, etc. for in-depth data analysis and query. At the same time, users can also, by setting thresholds and other means, conduct real-time monitoring and early warning of the energy consumption data. When the energy consumption exceeds the set threshold, the system automatically issues an alarm to remind the user to handle it. The user can click on a specific area on the map to view the detailed energy consumption data and situation analysis report of that area, including specific values of energy consumption, equipment operation status, prediction trends, and other information, so as to promptly discover abnormal situations and take corresponding measures. Through this application, not only the distribution, trends, etc. of multi-source heterogeneous energy consumption data are displayed in conventional forms such as visual charts and maps, and users can intuitively view the energy consumption data and analysis results. It also supports users to select different time intervals, regional ranges, data types, etc. through interactive operations for in-depth data analysis and query, and can also set thresholds for real-time monitoring and early warning.

[0062] After extracting features from the energy consumption resource data within the target time period in this application, multiple feature data are obtained, and then the multiple feature data are evaluated and predicted to respectively obtain perception information and prediction information. Based on this, control information of the energy consumption resources is generated, so that the generated control information of the energy consumption resources is more in line with the energy consumption resource data, realizing a comprehensive analysis of the energy consumption system by the control information of the energy consumption resources and improving the accuracy of the control information of the energy consumption resources.

[0063] In some feasible implementation manners, optionally, before extracting features from the energy consumption resource data of the target time period to obtain multiple feature data, the method further includes:

[0064] Obtain multiple distributed energy consumption resource data within the target time period, where the power consumption sources of each distributed energy consumption resource data among the multiple distributed energy consumption resource data are different;

[0065] Perform data cleaning on the multiple distributed energy consumption resource data to obtain the multiple cleaned distributed energy consumption resource data, where the data cleaning is used to delete duplicate data and error data in the multiple distributed energy consumption resource data;

[0066] Perform data conversion on the multiple cleaned distributed energy consumption resource data to obtain the energy consumption resource data, where the data conversion includes at least one of the following: logarithmic transformation, detrending, and differencing operation.

[0067] In this embodiment, such asFigure 2 As shown Figure 2 is the workflow diagram in this application. First, multiple distributed energy consumption resource data in multiple target time periods are obtained. Exemplarily, demand-side distributed energy consumption resource data are widely collected, such as electricity, gas, water resource consumption data, and related equipment operation status data. Multi-source data are obtained from various security devices such as energy monitoring systems, smart meters, and security protection equipment. For example, the energy monitoring system provides real-time overall energy consumption monitoring data, the smart meter accurately records the details of electricity consumption, and the security protection equipment provides energy consumption-related security status information. To ensure the stability and timeliness of data transmission, the collected various data are transmitted to the data processing center through a reliable network connection, ensuring data integrity and timeliness and laying a foundation for subsequent processing. Edge computing capabilities are introduced at the data collection terminal to preliminarily filter and simply analyze the original energy consumption data, reducing unnecessary data transmission and reducing network bandwidth pressure. At the same time, combined with the fog computing architecture, a small data processing center is established at the local network node close to the data source to achieve distributed storage and collaborative processing of data, further improving the efficiency and timeliness of data collection and preliminary processing. Using the decentralized and immutable characteristics of the blockchain, the collected energy consumption data are encrypted and stored on the chain. Each data block contains a timestamp, a hash value of the data content, and a hash link to the previous data block, ensuring that the data source is traceable and the content is immutable, providing a secure and trustworthy data basis for subsequent data processing and analysis.

[0068] This application widely collects various energy consumption and equipment operation status data such as electricity, gas, and water resources, covering multi-source devices such as energy monitoring systems, smart meters, and security protection equipment, comprehensively covering demand-side distributed energy consumption resources to ensure data integrity.

[0069] After obtaining multiple distributed energy consumption resource data, perform data cleaning on the multiple distributed energy consumption resource data to obtain the cleaned multiple distributed energy consumption resource data. Data cleaning is used to delete duplicate data and error data in the multiple distributed energy consumption resource data. Specifically, for image data: Use image preprocessing algorithms to remove noise and enhance image quality. For example, use filtering algorithms to remove salt-and-pepper noise in the image, etc., to make the image clearer for subsequent analysis. For noise processing in image data, in addition to traditional filtering algorithms, introduce generative adversarial network (GAN) technology. The generator learns the feature patterns of normal images, and the discriminator distinguishes real images from generated images. Through the adversarial training of the two, it can more effectively remove complex noise in the image while retaining the key detail information of the image and improving the image quality. For text data: Perform natural language processing, extract key information, and remove irrelevant characters and redundant content. For example, perform lexical and syntactic analysis on device operation logs to extract key content such as device operation status and fault information. In text data preprocessing, use the pre-trained language model based on the Transformer architecture to perform semantic understanding and enhancement of the text. It can not only extract key information more accurately but also supplement missing or ambiguous information through generative methods, further improving the usability and analysis value of text data. For numerical data: Standardize and normalize it according to its physical meaning and dimension. For example, uniformly standardize power consumption data in kilowatts to facilitate comparison and analysis between different data.

[0070] Perform data transformation on the cleaned multiple distributed energy consumption resource data to obtain energy consumption resource data. The data transformation includes at least one of the following: logarithmic transformation, detrending, and differencing operation. That is, transform the cleaned multiple distributed energy consumption resource data. The transformation methods include logarithmic transformation, detrending, and differencing operation to improve model performance and avoid the problem of non-stationarity. Preprocess the energy consumption target value through a Target Transformer, such as logarithmic transformation and trend elimination, aiming to eliminate abnormal fluctuations or trend changes in the data and improve the model's ability to capture residuals. At the same time, this module also provides an inverse transformation function, which can restore the prediction results of the model to the original scale, ensuring the interpretability and practical significance of the prediction results.

[0071] This proposal adopts advanced and comprehensive processing technologies for different types of multi-source heterogeneous data. For example, for time series data, use logarithmic transformation, detrending, and differencing operation, etc., to improve the model's ability to capture residuals through a Target Transformer; use generative adversarial network technology to process complex noise in image data and retain key details; use the pre-trained language model based on the Transformer architecture to perform semantic understanding and enhancement of text data and supplement missing information to improve data usability.

[0072] Optionally, performing feature extraction on the energy consumption resource data for the target time period to obtain multiple feature data, including:

[0073] Based on the physical characteristics and business requirements of the energy consumption resource data, constructing an energy consumption time series model, where the energy consumption time series model is used to perform feature extraction on the input data based on the physical characteristics and the business requirements;

[0074] Inputting the energy consumption resource data for the target time period into the energy consumption time series model for feature extraction to obtain the multiple feature data, where the feature dimensions corresponding to the multiple feature data include at least one of the following: the load rate feature of the electrical equipment, the seasonal electricity consumption feature, the electrical equipment type feature, and the electrical equipment usage time feature.

[0075] In this embodiment, according to the physical characteristics and business requirements of the energy consumption resource data, an energy consumption time series model is constructed. Specifically, the multi-dimensional time lag features, time series features, and endogenous features are comprehensively utilized to construct the energy consumption time series model features, which can better capture the internal patterns and correlations of the data. The features are screened in combination with domain knowledge and expert experience. On the basis of screening features in combination with domain knowledge and expert experience, a causal inference method is introduced to mine the causal relationship features between the energy consumption data. By calculating the causal effects between variables, the features that have a real causal impact on the energy consumption situation perception and prediction are found, avoiding the feature selection bias caused by relying solely on correlation analysis, and improving the model's ability to understand and utilize the internal patterns of the data.

[0076] Through time series analysis plotting, including time series plots, distribution histograms, autocorrelation functions (ACF), and partial autocorrelation functions (PACF), representing the time series data analysis of the energy consumption data. The time series analysis diagram of the energy consumption data is as Figure 3 shown`. According to the physical characteristics and business requirements of the energy consumption data, the features that have an important impact on the situation perception and prediction are screened. For the electric power consumption data, factors such as the type, usage time, and season of the electrical equipment are considered, and targeted features are constructed, such as the equipment load rate feature, the seasonal electricity consumption feature, etc., to enhance the model's ability to understand and utilize the data features. The different types of energy consumption data and the relevant equipment operation state data are constructed into a graph structure, where the nodes represent the data elements and the edges represent the relationships between the data. The graph neural network is used to perform feature extraction and fusion on the graph structure data, fully considering the complex associations between multi-source data, mining richer feature information, and improving the model performance.

[0077] This application structures multi-dimensional time-lag features, time series features, and endogenous features, combining domain knowledge and expert experience to construct an energy consumption time series model feature to better capture the inherent patterns and correlations in the data. Causal inference methods are introduced to explore the causal relationship characteristics between energy consumption data, avoiding the feature selection bias caused by relying solely on correlation analysis, improving the model's ability to understand and utilize the inherent patterns in the data. Graph neural networks are also used to extract and fuse features from the graph structure constructed from multi-source data to mine richer information.

[0078] Optionally, before inputting the plurality of feature data into a deep learning model for prediction and obtaining prediction information, the method further includes:

[0079] Establishing an initial model, the initial model including a long short-term memory network and a convolutional neural network, wherein the long short-term memory network is used to identify the time series features of the plurality of feature data, and the convolutional neural network is used to identify the spatial features of the plurality of feature data;

[0080] The initial model is updated based on the attention mechanism and Transformer architecture to obtain the deep learning model.

[0081] In this embodiment, the deep learning model is an XGBoost model. The initial model includes a long short-term memory network and a convolutional neural network, and the initial model is updated through an attention mechanism and a Transformer architecture to obtain the deep learning model. Specifically, the XGBoost model is used to construct a model for predicting future energy consumption. An automated hyperparameter optimization technique is adopted to ensure the optimal performance of the model. A fusion model is constructed for accurate prediction. Based on the fusion model of deep learning and ensemble learning for prediction, in addition to ensemble learning methods such as random forests and gradient boosting decision trees, a deep combination model of a long short-term memory network (LSTM) and a convolutional neural network (CNN) is introduced. The LSTM is used to process the time series characteristics of energy consumption data and capture long-term time dependencies. The CNN is used to extract the spatial characteristics of the data. By combining the advantages of both, more accurate prediction of multi-source heterogeneous energy consumption data is achieved. A dynamic model update strategy is adopted to adapt to the dynamic changes and real-time requirements of energy consumption data. According to the continuous influx of new data, the prediction model is regularly retrained and updated. The attention mechanism and the Transformer architecture are introduced into the deep combination model of the long short-term memory network and the convolutional neural network. The attention mechanism enables the model to automatically focus on key time steps and feature dimensions when processing time series data, better capturing the long-term dependencies and local features in the energy consumption data, and further improving the prediction accuracy of the model. For example, the model is fine-tuned every hour or every day based on the latest energy consumption data to ensure that the model can promptly capture the changes in energy consumption trends and the impact of emergencies on the prediction results, improving the real-time adaptability and prediction accuracy of the model. The transfer learning technique is adopted to transfer the model parameters trained in other similar energy consumption scenarios or related fields to the current prediction model, accelerating the model convergence speed and improving the model generalization ability. At the same time, combined with the meta-learning method, the model can quickly adapt to the new energy consumption data distribution and change rules, automatically adjust the model parameters and hyperparameters, and realize the optimization of the dynamic model update strategy, improving the real-time adaptability and prediction accuracy of the model.

[0082] In this application, the XGBoost model is combined with an automated hyperparameter optimization technique to construct a fusion model based on deep learning and ensemble learning. A deep combination model of a long short-term memory network and a convolutional neural network is introduced, and the attention mechanism and the Transformer architecture are also introduced into it to improve the model prediction accuracy. A dynamic model update strategy is adopted, combined with transfer learning and meta-learning methods, to retrain and update the model in real time according to new data, quickly adapt to the energy consumption data distribution and change rules, and improve the real-time adaptability and prediction accuracy of the model, which is not available in the prior art.

[0083] Optionally, the evaluating the multiple feature data to obtain the perception information includes:

[0084] Obtain a plurality of information evaluation parameters and a plurality of quantitative evaluation parameters, where the plurality of information evaluation parameters are used to represent different information evaluation dimensions, and the plurality of quantitative evaluation parameters are used to represent different quantitative evaluation dimensions;

[0085] Generate an information evaluation index based on the plurality of information evaluation parameters, and generate a quantitative evaluation index based on the plurality of quantitative evaluation parameters;

[0086] Evaluate the plurality of feature data based on the information evaluation index and the quantitative evaluation index to obtain the perception information.

[0087] In this embodiment, as Figure 4 shown, Figure 4 is a schematic diagram of the evaluation index structure, which includes an information evaluation index and a quantitative evaluation index. The information evaluation index is the evaluation index of the information acquisition layer, and the quantitative evaluation index is the quantitative index of the quantitative perception layer. Considering the characteristics of the monitoring data of the industrial control system, according to the specific requirements of different systems, the deviation between the measured data and the normal data or the fluctuation degree of the measured data is set as the standard for stage division.

[0088] The situation elements extracted by the information acquisition layer are used as the evaluation indexes of the evaluation system, and the situation evaluation results of the to-be-measured system are obtained by combining the current system. If the to-be-measured system is denoted as S, the evaluation method is mapped as →, and the evaluation index is indicators, then the joint evaluation structure of network security and energy consumption situation can be further expressed as:

[0089] indicators×S→results

[0090] The data sources are diverse, and the same index may be obtained from multiple channels. Some data can be directly used as evaluation indexes. The data that needs to be processed may need to be converted through quantization or labeling to meet the input requirements.

[0091] This application introduces image vision analysis technology to enrich the information dimension, combines artificial intelligence for joint situation evaluation, also introduces quantum machine learning algorithms to accelerate the analysis of massive energy consumption data and complex system states, and combines digital twin technology to achieve all-round and immersive perception and analysis of the energy consumption system. Compared with the prior art, the analysis is more comprehensive, accurate and efficient.

[0092] Optionally, the evaluating the plurality of feature data based on the information evaluation index and the quantitative evaluation index to obtain the perception information includes:

[0093] Evaluate the network security of the multiple feature data to obtain multiple network security evaluation information, where the multiple feature data correspond one-to-one to the multiple network security evaluation information, and the network security evaluation information is used to indicate whether the network security of the corresponding feature data meets the preset security requirements;

[0094] Based on the multiple network security evaluation information, screen the multiple feature data to obtain multiple target feature data, and the network security evaluation results corresponding to the multiple target feature data meet the preset security requirements;

[0095] Based on the information evaluation index and the quantitative evaluation index, evaluate the multiple target feature data to obtain the perception information.

[0096] In this embodiment, evaluate the network security of the multiple feature data to obtain multiple network security evaluation information. The network security evaluation information will be fed back to the data acquisition layer and the quantitative perception layer to provide a reliable basis for their information processing, thereby improving the system performance, as Figure 5 shown.

[0097] By obtaining multiple network security evaluation information to screen the multiple feature data, multiple target feature data are obtained. Thus, based on the information evaluation index and the quantitative evaluation index, the multiple target feature data are evaluated to obtain the perception information.

[0098] Specifically, when evaluating the network security, establish a database of known attack behaviors. When an abnormality occurs in the monitoring node, compare it with the attack patterns in the database. If the abnormality is highly similar to the attack patterns in the database, it is identified as an attack behavior. By comparing the feature differences between known attack types and normal events, set corresponding threat levels according to the degree of deviation.

[0099] In traditional information security, the attack process is typically divided into four phases: scanning, intrusion, privilege acquisition, and target achievement. However, not all attacks strictly follow these steps. Therefore, it is necessary to assume that all attacks may have multiple stages, making the attack path difficult to exhaust. This complicates the reconstruction of attack paths based on fixed patterns. To address this, this proposal proposes a similarity calculation method based on Euclidean distance for dynamic reconstruction of attack trajectories. Within any time window, attack behaviors are not isolated but rather comprise multiple, continuously evolving attack intentions, indicating correlations between security events. By calculating the distance between the features at each time point and the features of known attack types, the combination of known attack patterns can be used to describe the attack situation facing the current system, providing a method for intuitively understanding the system's security status. Specifically, by setting a sliding window, the security assessment value P at each time point is calculated based on the data within the current window. The data features of each sliding window are compared with the features of the known attack types, and the similarity between the test data within the current window and the known attack types is calculated. This results in a normalized security similarity vector {a1,...,an}, which serves as the weight for each known attack type. The distance value in the security similarity vector reflects the similarity between attack events: the smaller the distance, the higher the similarity, and vice versa. This proposal not only considers the frequency of attacks, but more importantly, highlights the impact of high-threat attacks.

[0100] Compared to the number of security incidents, the threat attributes of an attack itself pose a more serious threat to the system. Therefore, the assessment method proposed in this study multiplies the threat attribute of each known attack by the number of attacks using an exponential function of 2. This balances the number of attacks and the threat attribute, generating an assessment result that accurately reflects the actual security status of the data within the sliding window. By analyzing the changes in the elements, the current attack trajectory of the variable can be determined. Based on the unique threat factors of the variable, the situation value of the variable can be determined.

[0101]

[0102] Among them, P j represents the situation assessment results at each attack moment, n represents the number of attack categories suffered by the detection node, a i Indicates the normalized similarity between the event to be tested and the known attack event, EA i Represents the threat attributes of each attack time, N i Indicates the number of attacks.

[0103] This application introduces an adversarial assessment strategy and constructs a real-time perception and analysis platform for energy consumption status that integrates multiple functional modules and uses big data technology to achieve efficient storage and real-time processing of massive data to meet all-round management needs.

[0104] Optionally, generating control information for energy consumption resources based on the perception information and the prediction information includes:

[0105] Evaluating the multiple feature data based on a data fusion quality metric to obtain data evaluation information, where the data fusion quality metric is used to measure the consistency and integrity of the multiple feature data after data fusion;

[0106] Generating the control information for the energy consumption resources based on the perception information, the prediction information, and the data evaluation information.

[0107] In this embodiment, the multiple feature data are evaluated based on a data fusion quality metric to obtain data evaluation information, where the data fusion quality metric is used to measure the consistency and integrity of the multiple feature data after data fusion. Specifically, the schematic diagram of the control information for energy consumption resources is as Figure 6 shown. A more comprehensive and detailed automated evaluation system is constructed. In addition to traditional evaluation metrics such as accuracy, recall rate, root mean square error (RMSE), and mean absolute error (MAE), evaluation metrics specific to the characteristics of multi-source heterogeneous data are introduced, such as the data fusion quality metric and the feature extraction effectiveness metric. The data fusion quality metric can measure the consistency and integrity of the data after fusion from different data sources. The higher the correlation coefficient, the stronger the correlation of the data after fusion. The lower the information entropy, the better the consistency and integrity of the data. The feature extraction effectiveness metric can evaluate the contribution of feature engineering to the improvement of model performance by comparing the improvement in the accuracy of the model before and after using different feature engineering methods to measure the effectiveness of feature extraction. According to the evaluation results, the operation of the energy system is optimized and adjusted, and then suggestions for equipment upgrade or replacement are proposed to achieve refined management and sustainable development of the energy system. During the evaluation process, an adversarial evaluation strategy is introduced. The generator in the generative adversarial network (GAN) generates false data, and the evaluation metrics and the model judge and identify these false data. Through adversarial training, the evaluation metrics and the model are continuously optimized to enable them to more accurately identify real data and false data, improving the reliability and robustness of the evaluation. At the same time, the evaluation results can also provide a reference for subsequent data processing and model optimization to continuously improve the system performance. For the characteristics of multi-source heterogeneous data, in addition to the data fusion quality metric and the feature extraction effectiveness metric, multi-modal data fusion evaluation metrics are proposed, comprehensively considering factors such as semantic consistency and information complementarity between different modal data to comprehensively evaluate the effect of multi-source data fusion and its improvement on model performance. At the same time, this module supports visual analysis of the energy consumption trend within different time ranges to facilitate users to understand the change trend of energy consumption.

[0108] This application constructs a more comprehensive and detailed automated evaluation system. In addition to traditional evaluation indicators, evaluation indicators targeting the characteristics of multi-source heterogeneous data are introduced, such as data fusion quality indicators, feature extraction effectiveness indicators, multi-modal data fusion evaluation indicators, etc., to comprehensively evaluate the effect of multi-source data fusion and its improvement on model performance.

[0109] Adversarial evaluation strategy: An adversarial evaluation strategy is introduced. The generator in the generative adversarial network (GAN) is used to generate false data to optimize the evaluation indicators and the model, improving the reliability and robustness of the evaluation.

[0110] Specifically, RMSE (root mean square error) and MAPE (mean absolute percentage error) are used for the performance evaluation and analysis of the algorithm model for energy consumption situation awareness and prediction.

[0111] RMSE (root mean square error): It is used to measure the gap between the model's predicted values and the actual values, and calculates the square root of the average of the squared prediction errors. RMSE is more sensitive to larger errors and is suitable for use when large error penalties are concerned. The smaller the value, the better the prediction performance of the model. Specifically, as shown below:

[0112]

[0113] Among them, n represents the activation function, y i is the true value, is the predicted value.

[0114] MAPE (mean absolute percentage error): It calculates the percentage of the absolute error between the predicted value and the actual value to the actual value, expressed in percentage form. MAPE can intuitively reflect the proportion of the error to the actual value. Specifically, as shown in the following formula:

[0115]

[0116] After generating the control information of energy consumption resources, an intelligent operation management decision support system is established. Based on the situation awareness, prediction results, and evaluation indicators, it provides intelligent decision-making suggestions for energy management. When it is predicted that an energy consumption peak is approaching and the current system load is close to saturation, an equipment scheduling plan is automatically generated to start standby energy equipment in advance or adjust the energy distribution strategy to ensure the stability and security of energy supply. At the same time, according to different scenarios and requirements, multiple alternative plans are formulated, and evaluated and selected in combination with the actual situation to ensure the rationality and effectiveness of the decision. The swarm intelligence algorithms, such as ant colony algorithm, particle swarm optimization algorithm, etc., are introduced to simulate the intelligent behavior of biological groups, optimize the search and group decision-making for multiple alternative decision-making plans. Through the collaborative effect of the swarm intelligence algorithms, different decision-making factors and constraints are fully considered to explore better energy management decision-making plans and improve the rationality and effectiveness of the decision.

[0117] This application establishes an intelligent operation management decision support system, which provides intelligent decision-making suggestions for energy management based on situation awareness, prediction results, and evaluation indicators. When situations such as energy consumption peaks are predicted, equipment scheduling plans are automatically generated to ensure the stable and safe supply of energy. Group intelligence algorithms, such as ant colony algorithms and particle swarm optimization algorithms, are introduced to optimize the search and group decision-making for multiple alternative decision-making plans, fully considering different decision-making factors and constraints, mining better decision-making plans, and improving the rationality and effectiveness of decision-making.

[0118] After extracting the feature data from the energy consumption resource data within the target time period in this application, multiple feature data are obtained, and then the multiple feature data are evaluated and predicted to obtain perception information and prediction information respectively. Based on this, control information for energy consumption resources is generated according to the perception information and prediction information, so that the generated control information for energy consumption resources is more in line with the energy consumption resource data, realizing a comprehensive analysis of the energy consumption system by the control information of energy consumption resources and improving the accuracy of the control information of energy consumption resources.

[0119] See Figure 7 , Figure 7 which is the structural diagram of the control information generation device for energy consumption resources provided by the embodiment of this application. As Figure 7 shown, the control information generation device 700 for energy consumption resources includes:

[0120] An extraction module 710, configured to extract feature data from the energy consumption resource data within the target time period to obtain multiple feature data. The energy consumption resource data is the operation status data of the power system within the target time period, and among the multiple feature data, different feature data corresponds to different feature types;

[0121] An evaluation module 720, configured to evaluate the multiple feature data to obtain perception information, and input the multiple feature data into a deep learning model for prediction to obtain prediction information. The perception information is used to indicate the energy consumption peak of the energy consumption resource data after the target time period, and the prediction information is used to indicate the change trend of the energy consumption resource data after the target time period;

[0122] A generation module 730, configured to generate control information for energy consumption resources based on the perception information and the prediction information. The control information for energy consumption resources is used to control the energy consumption resource consumption of the power system after the target time period.

[0123] Optionally, it further includes:

[0124] A collection module, configured to obtain multiple distributed energy consumption resource data within the target time period, and the power consumption sources of each distributed energy consumption resource data among the multiple distributed energy consumption resource data are different;

[0125] A cleaning module, configured to clean the multiple distributed energy consumption resource data to obtain the cleaned multiple distributed energy consumption resource data, where the data cleaning is used to delete duplicate data and error data in the multiple distributed energy consumption resource data;

[0126] A conversion module, configured to perform data conversion on the cleaned multiple distributed energy consumption resource data to obtain the energy consumption resource data, where the data conversion includes at least one of the following: logarithmic transformation, detrending, and differencing operations.

[0127] Optionally, the extraction module 710 includes:

[0128] A construction sub-module, configured to construct an energy consumption time series model based on the physical characteristics and service requirements of the energy consumption resource data, where the energy consumption time series model is used to extract features from the input data based on the physical characteristics and the service requirements;

[0129] An extraction sub-module, configured to input the energy consumption resource data in the target time period into the energy consumption time series model for feature extraction to obtain the multiple feature data, where the feature dimensions corresponding to the multiple feature data include at least one of the following: load rate feature of the electrical equipment, seasonal electricity consumption feature, electrical equipment type feature, and electrical equipment usage time feature.

[0130] Optionally, it further includes:

[0131] An establishment module, configured to establish an initial model, where the initial model includes a long short-term memory network and a convolutional neural network, where the long short-term memory network is used to identify the time series features of the multiple feature data, and the convolutional neural network is used to identify the spatial features of the multiple feature data;

[0132] An update module, configured to update the initial model based on the attention mechanism and the Transformer architecture to obtain the deep learning model.

[0133] Optionally, the evaluation module 720 includes:

[0134] An acquisition sub-module, configured to acquire multiple information evaluation parameters and multiple quantization evaluation parameters, where the multiple information evaluation parameters are used to represent different information evaluation dimensions, and the multiple quantization evaluation parameters are used to represent different quantization evaluation dimensions;

[0135] A first generation sub-module, configured to generate an information evaluation index based on the multiple information evaluation parameters and generate a quantization evaluation index based on the multiple quantization evaluation parameters;

[0136] The first evaluation sub-module is used to evaluate the multiple feature data based on the information evaluation index and the quantitative evaluation index to obtain the perception information.

[0137] Optionally, the evaluation sub-module includes:

[0138] An evaluation unit is used to evaluate the network security of the multiple feature data to obtain multiple network security evaluation information. The multiple feature data correspond to the multiple network security evaluation information one by one. The network security evaluation information is used to indicate whether the network security of the corresponding feature data meets the preset security requirements;

[0139] A screening unit is used to screen the multiple feature data based on the multiple network security evaluation information to obtain multiple target feature data. The network security evaluation results corresponding to the multiple target feature data meet the preset security requirements;

[0140] Evaluate the multiple target feature data based on the information evaluation index and the quantitative evaluation index to obtain the perception information.

[0141] Optionally, the generation module 730 includes:

[0142] A second evaluation sub-module is used to evaluate the multiple feature data based on the data fusion quality index to obtain data evaluation information. The data fusion quality index is used to measure the consistency and integrity after the multiple feature data are fused;

[0143] A second generation sub-module is used to generate the control information of the energy consumption resources based on the perception information, the prediction information, and the data evaluation information.

[0144] In this application, after extracting features from the energy consumption resource data in the target time period, multiple feature data are obtained, and then the multiple feature data are evaluated and predicted to obtain the perception information and the prediction information respectively. Thus, the control information of the energy consumption resources is generated according to the perception information and the prediction information, so that the generated control information of the energy consumption resources is more in line with the energy consumption resource data, realizing a comprehensive analysis of the energy consumption system by the control information of the energy consumption resources and improving the accuracy of the control information of the energy consumption resources.

[0145] This application embodiment also provides an electronic device. Please refer to Figure 8 , the electronic device may include a processor 801, a memory 802, and a program 8021 stored in the memory 802 and executable on the processor 801.

[0146] When the program 8021 is executed by the processor 801, it can implement Figure 1 any step in the corresponding method embodiment:

[0147] Extract features from the energy consumption resource data for the target time period to obtain multiple feature data. The energy consumption resource data is the operation status data of the power system during the target time period. Among the multiple feature data, different feature data correspond to different feature types;

[0148] Evaluate the multiple feature data to obtain perception information, and input the multiple feature data into a deep learning model for prediction to obtain prediction information. The perception information is used to indicate the energy consumption peak of the energy consumption resource data after the target time period, and the prediction information is used to indicate the change trend of the energy consumption resource data after the target time period;

[0149] Generate control information for the energy consumption resource based on the perception information and the prediction information. The control information for the energy consumption resource is used to control the energy consumption resource consumption of the power system after the target time period.

[0150] Optionally, before extracting features from the energy consumption resource data for the target time period to obtain multiple feature data, the method further includes:

[0151] Obtain multiple distributed energy consumption resource data during the target time period. The power consumption sources of each distributed energy consumption resource data among the multiple distributed energy consumption resource data are different;

[0152] Perform data cleaning on the multiple distributed energy consumption resource data to obtain the cleaned multiple distributed energy consumption resource data. The data cleaning is used to delete duplicate data and error data in the multiple distributed energy consumption resource data;

[0153] Perform data conversion on the cleaned multiple distributed energy consumption resource data to obtain the energy consumption resource data. The data conversion includes at least one of the following: logarithmic transformation, detrending, and differencing operation.

[0154] Optionally, extracting features from the energy consumption resource data for the target time period to obtain multiple feature data includes:

[0155] Construct an energy consumption time series model based on the physical characteristics and business requirements of the energy consumption resource data. The energy consumption time series model is used to extract features from the input data based on the physical characteristics and the business requirements;

[0156] Input the energy consumption resource data of the target time period into the energy consumption time series model for feature extraction to obtain the multiple feature data. The feature dimensions corresponding to the multiple feature data include at least one of the following: load rate feature of electrical equipment, seasonal electricity consumption feature, electrical equipment type feature, and electrical equipment usage time feature.

[0157] Optionally, before inputting the multiple feature data into the deep learning model for prediction to obtain prediction information, the method further includes:

[0158] Establish an initial model, where the initial model includes a long short-term memory network and a convolutional neural network. Among them, the long short-term memory network is used to identify the time series features of the multiple feature data, and the convolutional neural network is used to identify the spatial features of the multiple feature data;

[0159] Update the initial model based on the attention mechanism and the Transformer architecture to obtain the deep learning model.

[0160] Optionally, the evaluating the multiple feature data to obtain perception information includes:

[0161] Obtain a plurality of information evaluation parameters and a plurality of quantization evaluation parameters, where the plurality of information evaluation parameters are used to represent different information evaluation dimensions, and the plurality of quantization evaluation parameters are used to represent different quantization evaluation dimensions;

[0162] Generate an information evaluation index based on the plurality of information evaluation parameters, and generate a quantization evaluation index based on the plurality of quantization evaluation parameters;

[0163] Evaluate the multiple feature data based on the information evaluation index and the quantization evaluation index to obtain the perception information.

[0164] Optionally, the evaluating the multiple feature data based on the information evaluation index and the quantization evaluation index to obtain the perception information includes:

[0165] Evaluate the network security of the multiple feature data to obtain a plurality of network security evaluation information, where the multiple feature data correspond to the plurality of network security evaluation information one by one, and the network security evaluation information is used to indicate whether the network security of the corresponding feature data meets the preset security requirements;

[0166] Filter the multiple feature data based on the plurality of network security evaluation information to obtain a plurality of target feature data, and the network security evaluation results corresponding to the plurality of target feature data meet the preset security requirements;

[0167] Evaluate the multiple target feature data based on the information evaluation index and the quantization evaluation index to obtain the perception information.

[0168] Optionally, the generating control information of energy consumption resources based on the perception information and the prediction information includes:

[0169] Evaluate the multiple feature data based on a data fusion quality metric to obtain data evaluation information, where the data fusion quality metric is used to measure the consistency and integrity after data fusion of the multiple feature data;

[0170] Generate control information for the energy consumption resources based on the perception information, the prediction information, and the data evaluation information.

[0171] In this application, after extracting features from the energy consumption resource data within a target time period, multiple feature data are obtained. Then, the multiple feature data are evaluated and predicted to obtain perception information and prediction information respectively. Thus, control information for the energy consumption resources is generated based on the perception information and the prediction information, making the generated control information for the energy consumption resources more consistent with the energy consumption resource data, achieving a comprehensive analysis of the energy consumption system by the control information for the energy consumption resources, and improving the accuracy of the control information for the energy consumption resources.

[0172] This application embodiment also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements each process of the above-mentioned embodiment of the method for generating control information for energy consumption resources and can achieve the same technical effects. To avoid repetition, it will not be elaborated here. Among them, the computer-readable storage medium, such as a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disc, etc.

[0173] This application embodiment further provides a computer program product, which is stored in a storage medium. The computer program product is executed by at least one processor to implement each process of the above-mentioned embodiment of the method for generating control information for energy consumption resources and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0174] It should be noted that in this article, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device including a series of elements not only includes those elements but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article, or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article, or device including that element.

[0175] Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to enable a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of the present application.

[0176] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the present application and the scope protected by the claims, and all of them belong to the protection scope of the present application.

Claims

1. A method for generating control information of energy consumption resources, characterized in that, The method includes: Performing feature extraction on the energy consumption resource data for the target time period to obtain multiple feature data. The energy consumption resource data is the operation status data of the power system during the target time period. Among the multiple feature data, different feature data correspond to different feature types; Evaluating the multiple feature data to obtain perception information, and inputting the multiple feature data into a deep learning model for prediction to obtain prediction information. The perception information is used to indicate the energy consumption peak of the energy consumption resource data after the target time period, and the prediction information is used to indicate the change trend of the energy consumption resource data after the target time period; Generating control information for the energy consumption resource based on the perception information and the prediction information. The control information for the energy consumption resource is used to control the energy consumption resource consumption of the power system after the target time period.

2. The method according to claim 1, wherein Before performing feature extraction on the energy consumption resource data for the target time period to obtain multiple feature data, the method further includes: Obtaining multiple distributed energy consumption resource data during the target time period. The power consumption sources of each distributed energy consumption resource data among the multiple distributed energy consumption resource data are different; Performing data cleaning on the multiple distributed energy consumption resource data to obtain the cleaned multiple distributed energy consumption resource data. The data cleaning is used to delete duplicate data and error data in the multiple distributed energy consumption resource data; Performing data conversion on the cleaned multiple distributed energy consumption resource data to obtain the energy consumption resource data. The data conversion includes at least one of the following: logarithmic transformation, detrending, and differencing operation.

3. The method according to claim 2, wherein Performing feature extraction on the energy consumption resource data for the target time period to obtain multiple feature data includes: Constructing an energy consumption time series model based on the physical characteristics and business requirements of the energy consumption resource data. The energy consumption time series model is used to perform feature extraction on the input data based on the physical characteristics and the business requirements; Inputting the energy consumption resource data for the target time period into the energy consumption time series model for feature extraction to obtain the multiple feature data. The feature dimensions corresponding to the multiple feature data include at least one of the following: load rate feature of the electrical equipment, seasonal electricity consumption feature, electrical equipment type feature, and electrical equipment usage time feature.

4. The method according to claim 1, wherein Before inputting the multiple feature data into a deep learning model for prediction to obtain prediction information, the method further includes: Establishing an initial model. The initial model includes a long short-term memory network and a convolutional neural network. Among them, the long short-term memory network is used to identify the time series features of the multiple feature data, and the convolutional neural network is used to identify the spatial features of the multiple feature data; Updating the initial model based on the attention mechanism and the Transformer architecture to obtain the deep learning model.

5. The method according to claim 1, wherein Evaluating the multiple feature data to obtain perception information includes: Obtain a plurality of information evaluation parameters and a plurality of quantitative evaluation parameters, where the plurality of information evaluation parameters are used to represent different information evaluation dimensions, and the plurality of quantitative evaluation parameters are used to represent different quantitative evaluation dimensions; Generate an information evaluation index based on the plurality of information evaluation parameters, and generate a quantitative evaluation index based on the plurality of quantitative evaluation parameters; Evaluate the plurality of feature data based on the information evaluation index and the quantitative evaluation index to obtain the perceived information.

6. The method according to claim 5, characterized in that, The evaluating the plurality of feature data based on the information evaluation index and the quantitative evaluation index to obtain the perceived information includes: Evaluate the network security of the plurality of feature data to obtain a plurality of network security evaluation information, where the plurality of feature data and the plurality of network security evaluation information are in one-to-one correspondence, and the network security evaluation information is used to indicate whether the network security of the corresponding feature data meets the preset security requirements; Filter the plurality of feature data based on the plurality of network security evaluation information to obtain a plurality of target feature data, and the network security evaluation results corresponding to the plurality of target feature data meet the preset security requirements; Evaluate the plurality of target feature data based on the information evaluation index and the quantitative evaluation index to obtain the perceived information.

7. The method according to claim 1, characterized in that The generating the control information of the energy consumption resources based on the perceived information and the prediction information includes: Evaluate the plurality of feature data based on the data fusion quality index to obtain data evaluation information, where the data fusion quality index is used to measure the consistency and integrity after the plurality of feature data are fused; Generate the control information of the energy consumption resources based on the perceived information, the prediction information, and the data evaluation information.

8. A control information generation device for energy consumption resources, characterized in that, The apparatus includes: An extraction module, configured to extract features from the energy consumption resource data of the target time period to obtain a plurality of feature data, where the energy consumption resource data is the operation state data of the power system during the target time period, and among the plurality of feature data, different feature data correspond to different feature types; An evaluation module, configured to evaluate the plurality of feature data to obtain perceived information, and input the plurality of feature data into a deep learning model for prediction to obtain prediction information, where the perceived information is used to indicate the energy consumption peak of the energy consumption resource data after the target time period, and the prediction information is used to indicate the change trend of the energy consumption resource data after the target time period; A generation module, configured to generate control information of the energy consumption resources based on the perceived information and the prediction information, where the control information of the energy consumption resources is used to control the energy consumption resource consumption of the power system after the target time period.

9. An electronic device, characterized in that, Includes: A processor, a memory, and a program stored on the memory and executable on the processor, where when the program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

11. A computer program product, characterized in that, Comprising computer instructions which, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.

Images