Energy management method and system combining big data regulation and control behavior prediction

Abstract

The invention discloses an energy management method and system combining big data regulation and control behavior prediction, and relates to the technical field of intelligent energy management systems, and the method comprises the steps: setting a preset detection time period; detecting regulation and control behaviors of a plurality of regulation and control sides (users, systems and managers), and triggering at least two regulation and control sides to output corresponding data packets; performing feature vectorization processing on the data packet to obtain feature vectors, and importing the feature vectors into a prediction model for prediction; and connecting the energy management system to execute energy management according to the regulation and control prediction result. The technical problems that in an energy prediction and management scene, a traditional data processing mode is difficult to accurately perceive regulation and control intentions and reconcile behavior conflicts, consequently, the matching degree of energy scheduling and actual requirements is low, and efficient energy management requirements cannot be met are solved, and the purposes of accurately perceiving the regulation and control intentions and reconcile behavior conflicts and improving the energy management efficiency are achieved. The matching degree of energy scheduling and requirements is improved, and the technical effect of efficient energy management is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent energy management systems, and in particular to an energy management method and system combined with big data regulation behavior prediction. Background Art

[0002] In energy management, accurate prediction of control behavior is crucial for improving energy efficiency and optimizing management decisions. Existing energy management relies heavily on single-side control data or simple integration, lacking collaborative analysis of user, system, and management behaviors. Traditional methods, faced with complex interactions involving multiple sides, are unable to effectively integrate data and accurately predict control intent, resulting in a poor match between energy scheduling and actual demand, making scientific and efficient energy management difficult to achieve. Due to the lack of sufficient coordination of multi-side behavior data and the lack of a hierarchical prediction mechanism, traditional methods are unable to adapt to complex energy control scenarios and are unable to meet the needs of accurate prediction and management. Summary of the Invention

[0003] This application provides an energy management method and system that combines big data control behavior prediction to solve the technical problem that in energy prediction and management scenarios, traditional data processing methods are difficult to accurately perceive control intentions and reconcile behavioral conflicts, resulting in a low match between energy scheduling and actual needs and an inability to meet the needs of efficient energy management.

[0004] The first aspect of the present application provides an energy management method combined with big data control behavior prediction, the method comprising: setting a preset detection time period; performing control behavior detection on multiple control sides within the preset detection time period, and if it is detected that at least two of the multiple control sides trigger control behavior, outputting at least two groups of control behavior data packets corresponding to the at least two control sides, the multiple control sides including the user control side, the system control side and the management control side; performing feature vectorization processing on the at least two groups of control behavior data packets, outputting at least two groups of feature vectors, importing the at least two groups of feature vectors into a fusion behavior prediction model for prediction, and outputting control prediction results; connecting the energy management system to perform energy management with the control prediction results.

[0005] The second aspect of the present application provides an energy management system combined with big data control behavior prediction, the system including: a time period setting module, used to set a preset detection time period; a behavior detection execution module, used to perform control behavior detection on multiple control sides within the preset detection time period, and if it is detected that at least two of the multiple control sides trigger control behavior, output at least two groups of control behavior data packets corresponding to the at least two control sides, the multiple control sides include a user control side, a system control side and a management control side; a vectorization processing module, used to perform feature vectorization processing on the at least two groups of control behavior data packets, output at least two groups of feature vectors, import the at least two groups of feature vectors into a fusion behavior prediction model for prediction, and output the control prediction result; an energy management execution module, used to connect the energy management system to perform energy management with the control prediction result.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application collects the control behavior data of the user side, system side and management side in the energy management scenario, generates feature vectors through feature extraction, standardization and embedding of side information, and uses intention perception and three-level fusion modules to Figure 1 The consistency index is used to predict the control behavior in different collaborative scenarios, and the prediction results are adjusted in combination with the amplification, demodulation and other strategies of each module, so as to accurately predict the multi-side control behavior, provide a reliable basis for the energy management system to execute the strategy, improve the scientificity and efficiency of energy management, and achieve the technical effect of accurately perceiving the control intention, reconciling behavioral conflicts, improving the matching degree of energy scheduling and demand, and realizing efficient energy management. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0008] Figure 1 This is a flow chart of an energy management method combined with big data regulation behavior prediction provided in an embodiment of the present application.

[0009] Figure 2 It is a structural diagram of an energy management system combined with big data regulation behavior prediction provided in an embodiment of the present application.

[0010] Description of the accompanying drawings: time period setting module 1, behavior detection execution module 2, vectorization processing module 3, energy management execution module 4. DETAILED DESCRIPTION

[0011] This application provides an energy management method and system that combines big data control behavior prediction to solve the technical problem that in energy prediction and management scenarios, traditional data processing methods are difficult to accurately perceive control intentions and reconcile behavioral conflicts, resulting in a low match between energy scheduling and actual needs and an inability to meet the needs of efficient energy management.

[0012] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0013] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0014] Example 1, as Figure 1 As shown, an energy management method combining big data regulation behavior prediction, wherein the method includes: Step A100: Set a preset detection time period.

[0015] Specifically, the setting of the preset detection time period needs to determine the cycle length in combination with the actual needs of the energy management scenario. For example, it can be set to 15 minutes, 1 hour or 1 day, etc. The specific duration is flexibly adjusted by technical personnel in this field according to the response frequency of the energy system and data update requirements.

[0016] By setting a preset detection time period, this solution can realize the regular collection and phased processing of multi-control-side behavior data. With the help of standardized time windows, it provides a unified time benchmark for subsequent feature vector quantization, fusion prediction and energy management execution, effectively improving the efficiency and quality of data collection, reducing invalid computing overhead, and laying the foundation for accurately capturing multi-side behavior coordination patterns and improving the accuracy of energy control predictions.

[0017] Step A200: Perform control behavior detection on multiple control sides within the preset detection time period. If it is detected that at least two control sides among the multiple control sides trigger control behavior, output at least two groups of control behavior data packets corresponding to the at least two control sides. The multiple control sides include a user control side, a system control side, and a management control side.

[0018] In the embodiments of this application, the term "control side" refers to the three types of entities involved in energy management: user control side, system control side, and management control side. Control actions refer to energy management-related actions triggered by the control side within a preset detection time period, such as user-side power usage mode adjustments, system-side automatic control instructions execution, and management-side policy updates.

[0019] Optionally, at the beginning of the preset detection time period, the main scheduler sends behavior detection requests to multiple control sides to collect control behavior data packets from the user side, system side and management side, and performs content change analysis on the data packets. If the control behavior is not triggered, the current time period is skipped and the next time period is entered. If it is triggered by a single control side, the corresponding control behavior data packet is cached. The specific steps are described in detail in A210-A230.

[0020] Furthermore, in an industrial park management system, for example, user-side electricity meters upload data every 15 minutes, system-side load data is updated every 5 minutes, and management-side policies are issued daily. These three systems have inconsistent time bases, resulting in timestamp deviations of 3-5 minutes when analyzing cross-behavior correlations. This increases the error rate of the prediction model due to time series confusion. By using a one-hour preset time window and a synchronous API collection mechanism, a multi-side data collection process is established within a unified time window, addressing the coordination issues between time series and data types.

[0021] At the start of a preset detection period (e.g., a 1-hour cycle, triggered on the hour), the main scheduler, as the core control unit, simultaneously initiates behavior detection requests to the three types of control sides through a standardized API interface: User control side: Call the smart meter API to collect power consumption (for example, the current value is 2.5kW, which fluctuates by ±1.2kW compared to the previous period) and equipment start / stop status (for example, machine tool M01 in the workshop started at 14:00). The data is returned in structured JSON format, including the device ID, behavior type (power adjustment / device start / stop), and timestamp (accurate to the moment).

[0022] System control side: Access the power monitoring system API to obtain grid voltage (such as a 10kV bus voltage fluctuation of ±0.3kV) and load factor (such as a feeder load increasing from 60% to 75%). The data is transmitted in the form of a real-time data stream, with the monitoring point number, behavior type (load warning / voltage anomaly), and timestamp attached.

[0023] Management and control side: connect to the policy release platform API to capture energy efficiency index adjustments (such as the unit energy consumption threshold of the park from 1.2kWh / Reduced to 1.1kWh / ), peak-valley electricity price updates (such as the valley electricity price of 0.3 yuan / kWh is adjusted to 0.28 yuan / kWh), the data is returned in the form of command text, including the command type, control object (the entire workshop / specific production line), and timestamp (the time when the policy takes effect).

[0024] After the three types of requests are initiated, the main scheduler monitors responses through an asynchronous parallel mechanism, ensuring that all data packets are collected within 5 seconds (user-side response time ≤ 2 seconds, system-side ≤ 1.5 seconds, and management-side ≤ 1.5 seconds). This generates a data set for the time period, including timestamps, side identifiers, and raw data fields. After collection is complete, the system first verifies time synchronization: if the timestamp deviation of the data on the three sides is ≤ 5 seconds, it is considered to be effectively synchronized within the time period. If the deviation exceeds this, a time alignment algorithm is triggered. For example, using the user-side timestamp as a reference, the system-side and management-side data are interpolated and corrected to ensure time consistency in subsequent analysis.

[0025] When it is detected that at least two control sides trigger behavior within the time period, such as the power change on the user side ≥5% and the load rate change on the system side ≥3%, or the management side releases a new policy and the user side equipment starts and stops in response, it is immediately marked as a multi-side collaborative event, and the original data packet of the corresponding side (retaining the API return format characteristics) and the time-aligned metadata are output. This mechanism provides a multi-source data foundation with unified timing and type adaptation for subsequent feature extraction and model input.

[0026] By synchronously collecting and collaboratively triggering behavioral data on the user side, system side, and management side within a preset detection time period, multi-dimensional perception of energy regulation behavior is achieved, and valid data packets are output only when there is substantial correlation between data on multiple sides. This not only avoids the one-sided defects of single-side data, but also provides a high-quality collaborative data set for subsequent feature vector quantization and fusion prediction, thereby improving the scientific nature and response efficiency of energy management.

[0027] Step A300: performing feature vectorization processing on the at least two groups of control behavior data packets, outputting at least two groups of feature vectors, importing the at least two groups of feature vectors into the fusion behavior prediction model for prediction, and outputting the control prediction results.

[0028] In an embodiment of the present application, the prediction model is a hierarchical model for analyzing feature vectors of multiple control sides and outputting control prediction results.

[0029] In one embodiment of the present application, the behavior fields of at least two groups of control behavior data packets are extracted to obtain the behavior type, control object, control content and behavior timestamp, and then standardized; the control side embedding vector combination is introduced to standardize the behavior field, and at least two groups of feature vectors are output. The specific steps are described in detail in A310-A330.

[0030] The fusion behavior prediction model includes an intention perception module and three fusion modules. The feature vector is input into the intention perception module for analysis to obtain the control intention. Figure 1 Consistency index; based on this index, the corresponding fusion module prediction is activated and the corresponding control prediction results are output. The specific steps are described in detail in A340-A370.

[0031] Step A400: connecting to an energy management system to perform energy management using the control prediction result.

[0032] In an embodiment of the present application, the energy management system is a system for receiving the control prediction results output by the fusion behavior prediction model and performing energy management accordingly.

[0033] Specifically, the output of the fused behavior prediction model is first obtained: the intent perception module analyzes the semantic consistency of feature vectors from multiple sides and activates the corresponding fusion module to generate the control prediction results. After receiving the control prediction results, the energy management system first performs policy parsing and device mapping, translating the abstract prediction into specific execution instructions. For example, a 1.5kW power increase on the user side corresponds to adjusting the power threshold of the smart meter, while the activation of reactive power compensation on the system side triggers the compensation device of the feeder.

[0034] Then enter the priority execution stage: for high consistency prediction results (such as control intention Figure 1 The consistency index is ≥0.7), and the system executes the enhanced strategy after weighted fusion with a response speed of 0.5 seconds, such as amplifying the energy saving to 15%. For the compromise strategy in the conflict scenario, for example, the user increases the load by 5% and activates the energy storage. The system executes in steps according to the response priority (user demand 0.6, system stability 0.4), ensuring that the load fluctuation is controlled within ±8%.

[0035] Finally, a closed-loop correction process is implemented through real-time monitoring and feedback. The system continuously collects real-time data after execution, such as actual user power consumption and system load factor, compares it with the predicted results, and generates a deviation value (e.g., a tolerance of ±3%). If a threshold is exceeded, the model is recalculated. For example, if the load factor remains at 85% after the energy-saving strategy is implemented, the system automatically calls the second fusion module to generate a supplementary strategy, such as an additional 5% load reduction, forming a dynamic optimization chain of prediction, execution, and correction.

[0036] By deeply coupling the fusion prediction results with the equipment control logic of the energy management system, a complete closed loop is constructed from multi-source data perception to precise strategy execution. This process solves the static execution defects of traditional systems through a three-level mechanism of model prediction result-driven strategy generation, dynamic priority scheduling, and real-time feedback correction, thereby improving the match between the actual effect of the energy regulation strategy and the prediction target, and enhancing the system's adaptability and management efficiency for complex energy scenarios.

[0037] Furthermore, step A200 in the method provided in the embodiment of the present application includes: A210: When the preset detection time period begins, the main scheduler sends a behavior detection request to the multiple control sides, and collects user side control behavior data packets, system control side control behavior data packets, and management control side control behavior data packets.

[0038] A220: Perform content change analysis on the user-side control behavior data packet, the system-side control behavior data packet, and the management-side control behavior data packet. If it is detected that no control behavior is triggered in the multiple control sides, skip the current preset detection time period and enter the next preset detection time period.

[0039] A230: If it is detected that the multiple control sides trigger a control behavior for a single control side, cache the control behavior data packet corresponding to the single control side.

[0040] Specifically, at the start of a preset detection time period, the master dispatcher, acting as the system's core control unit, first sends a behavior detection request to the user, system, and management sides simultaneously. For example, taking a preset time period of one hour as an example, at every hour, the master dispatcher initiates a data acquisition request through an API interface to the user-side smart meter (collecting data such as power consumption and equipment startup and shutdown), the system-side power monitoring system (collecting parameters such as grid voltage and load factor), and the management-side policy publishing platform (collecting instructions such as energy efficiency index adjustment and peak and valley electricity price updates). Within five seconds, the master dispatcher completes the collection of three types of control behavior data packets, forming a raw data set containing fields such as timestamps, behavior types, and parameter values.

[0041] After the collection is completed, the three types of data packets are analyzed for content changes. This is achieved by comparing the difference between the current cycle data and the previous cycle data: Assuming that the power fluctuation amplitude on the user side is less than 5%, the load rate change on the system side is less than 3%, and there are no new instructions pushed on the management side (the specific difference threshold can be determined by technical personnel in this field based on the actual needs of the energy management scenario), it is determined that no control behavior has been triggered. In this case, the current cycle is directly skipped to avoid meaningless processing of static data; if the power fluctuation amplitude on the user side is ≥10%, such as when the user turns on a high-power appliance, and the data on the system side and the management side remain unchanged, it is determined to be a single control-side trigger, and the user-side data packet is stored in the cache queue, marked with a cache time, and waits for the subsequent cycle data combination.

[0042] Through the timed request triggering of the main scheduler, the synchronous collection of multi-side data and the intelligent filtering mechanism based on content changes, the resource consumption of invalid data processing is reduced. At the same time, by caching single-side data, the data foundation is retained for cross-cycle multi-side behavior collaborative analysis, thereby improving operational efficiency while laying the foundation for the accuracy of subsequent multi-source data fusion prediction.

[0043] Furthermore, step A240 in the method provided in the embodiment of the present application includes: A241: When no control behavior is triggered by another control side different from the cache control side at the end of the preset detection time period, the energy management system performs energy management according to the control behavior data packet corresponding to the cache control side.

[0044] A242: When the preset detection time period ends, at least one other control side different from the cache control side is detected to trigger a control behavior, the control behavior data packet corresponding to the cache control side is extracted and combined with the control behavior data packet corresponding to the at least one other control side, and at least two groups of control behavior data packets are output.

[0045] In the embodiment of the present application, the cache control side is the corresponding control side that caches the control behavior data packets corresponding to a single control side.

[0046] Optionally, after the aforementioned control behavior detection of multiple control sides within the preset detection time period, the corresponding trigger conditions are shown in Table 1. When only a single control side trigger behavior is detected within the preset detection time period (such as set to 1 hour), such as the management side issues a 10% load reduction instruction for the afternoon period at 14:00, the system stores the management side data packet in the cache queue, marks the cache time as 14:00-15:00, and starts dynamic monitoring before the end of the cycle. At the end of the period (15:00), the system first scans other control sides (user side, system side) to see if there are new trigger behaviors: Scenario without other-side triggering: If the user-side power fluctuation is less than 5% and the system-side load rate is stable, it is determined to be a valid instruction on a single side. The energy management system directly executes the cached management-side instructions, such as adjusting the power supply thresholds of each area. This process does not require starting the fusion prediction module, and computing resource consumption is reduced compared to traditional full-process processing.

[0047] There are other-side triggering scenarios: If it is detected at 14:55 that a high-power device in a workshop on the user side is started, and the trigger power change is ≥15%, the system immediately extracts the cached management-side data packet and the newly collected user-side data packet, combines them into a collaborative data set of management-side load reduction instructions and user-side load increase requirements, and outputs them for subsequent feature vectorization processing.

[0048] By caching data packets on a single control side and triggering detection across cycles, the blindness of direct data processing on a single side in the traditional model is avoided, and data combination conditions are created for potential multi-side collaboration through delayed decision-making. This improves the efficiency of computing resource utilization while ensuring the immediacy of energy management strategies in scenarios dominated by a single side and the accuracy of multi-side collaboration scenarios.

[0049] Table 1: Multi-side data acquisition and trigger conditions Furthermore, step A300 in the method provided in the embodiment of the present application includes: A310: Extracting the behavior fields of the at least two groups of control behavior data packets respectively, and outputting the behavior type, control object, control content, and behavior timestamp.

[0050] A320: Perform standardization processing according to the behavior type, control object, control content, and behavior timestamp to obtain at least two groups of standardized behavior fields.

[0051] A330: Introduce a regulation side embedding vector to combine the at least two groups of standardized behavior fields, and output at least two groups of feature vectors.

[0052] In an embodiment of the present application, the regulation side embedding vector is a vector used to combine standardized behavior fields. By introducing this vector, at least two groups of standardized behavior fields can be combined to output at least two groups of feature vectors.

[0053] Specifically, we first perform action field extraction: We parse the core fields of the control action data packets from the user, system, and management sides. For example, on the user side, we extract the action type (e.g., device start / stop), control object (e.g., smart meter ID-001), control content (e.g., power increase from 2.5kW to 3.8kW), and timestamp from the JSON-formatted data packet. On the system side, we extract the action type (e.g., voltage fluctuation), control object (e.g., feeder L-03), control content (e.g., voltage reduction from 10.2kV to 9.8kV), and timestamp from the data stream. On the management side, we extract the action type (e.g., energy efficiency index adjustment), control object (e.g., all users), control content (e.g., carbon emission standard reduction from 0.5t / month to 0.4t / month), and timestamp (e.g., policy effective date) from the policy text.

[0054] Next, when performing feature vectorization processing on at least two groups of control behavior data packets, regular expression matching technology is first used to accurately extract behavior type, control object, control content, timestamp and other fields from structured data (such as JSON format data packets on the user side and real-time data streams on the system side); for unstructured text data on the management side, natural language processing (NLP) technology is used to perform semantic analysis to extract key information such as instruction type and control object, so as to achieve accurate extraction of fields of structured and unstructured data.

[0055] Standardization then occurs: extracted fields are unified in type and their values ​​are normalized. For example, behavior types are mapped to enumerated values, such as device start / stop → 01, voltage fluctuation → 02, and energy efficiency adjustment → 03. Control objects are encoded as unique identifiers, such as smart meter ID-001 → OBJ-001. Control content values ​​are normalized according to the system range, such as dividing power by rated power and voltage by standard voltage. Timestamps are converted to relative times within a preset time period, such as decimals between 0 and 1 within a 1-hour period. For example, for user-side power data, at a rated power of 5kW, 3.8kW is normalized to 0.76; for a system-side voltage of 10kV, 9.8kV is normalized to 0.98. After standardization, multi-source data dimensions are unified into four core fields, ultimately resulting in at least two sets of standardized behavior fields. This improves data format consistency and lays the foundation for model input.

[0056] Finally, we introduce the side-specific embedding vector: We define one-hot encoding vectors for the user side, system side, and management side, such as [1, 0, 0] for the user side, [0, 1, 0] for the system side, and [0, 0, 1] for the management side. These are then concatenated with the standardized behavior field to generate a multidimensional feature vector containing side-specific characteristics. For example, a piece of user-side data is processed to form the feature vector: [01, OBJ-001, 0.76, 0.25, 1, 0, 0], where the first four dimensions represent the standardized behavior field and the last three dimensions represent the side-specific embedding vector. This vector not only preserves behavioral details but also explicitly expresses the data source, enabling the fusion model to distinguish the influence weights of different regulatory sides.

[0057] Through the above steps, the heterogeneous control behavior data on the user side, system side, and management side are converted into feature vectors with unified structure and clear semantics, which solves the problems of chaotic data fields and implicit side information in traditional methods, provides high-quality feature expression for the subsequent fusion behavior prediction model, and significantly enhances the accuracy and reliability of energy control behavior prediction.

[0058] Furthermore, step A300 in the method provided in the embodiment of the present application includes: A340: The fused behavior prediction model includes an intention perception module, a first fused behavior prediction module, a second fused behavior prediction module and a third fused behavior prediction module.

[0059] A350: Input the at least two sets of feature vectors into the intention perception module for analysis to obtain the control intention Figure 1 Consistency indicators, if the regulatory intent Figure 1 If the consistency index is greater than or equal to a preset consistency index threshold, the first fusion behavior prediction module is activated to predict the at least two groups of feature vectors and output a first control prediction result.

[0060] A360: If the control Figure 1 If the consistency index is less than the preset consistency index threshold and the returned value is not empty, the second fusion behavior prediction module is activated to predict the at least two groups of feature vectors and output a second control prediction result.

[0061] A370: If the control Figure 1 If the consistency index is returned as null, the third fusion behavior prediction module is activated to predict the at least two groups of feature vectors and output a third control prediction result.

[0062] In the embodiment of the present application, the intention perception module is used to perform semantic analysis on each feature vector in at least two sets of feature vectors, obtain at least two sets of semantic vectors, and calculate the semantic similarity of the at least two sets of semantic vectors respectively, output the mean value of the semantic similarity, and obtain the control intention with the mean value of the semantic similarity. Figure 1The fusion behavior prediction module is a model that includes the intention perception module, the first fusion behavior prediction module, the second fusion behavior prediction module, and the third fusion behavior prediction module.

[0063] Specifically, the core process of the fusion behavior prediction model is the semantic analysis of the intention perception module: at least two sets of feature vectors are input into the module, and the behavior fields in each vector (such as power regulation and load warning) are semantically parsed through natural language processing (NLP) technology to generate corresponding semantic vectors, such as the user-side vector [energy saving, equipment, +15%] and the system-side vector [stability, power grid, -10%]. The cosine similarity of each set of semantic vectors is then calculated, and the mean of the semantic similarity is output as the control intention. Figure 1 Consistency index (value range 0-1). For example, when both the user side and the management side involve energy-saving semantics, the average similarity can reach above 0.8; if the user side is increasing the load and the system side is reducing the load, the similarity may be lower than 0.3.

[0064] Based on intention Figure 1 Consistency indicators, the model dynamically triggers different fusion modules: High consistency scenario (indicator ≥ threshold, such as 0.7): The first fusion behavior prediction module obtains a first fused feature vector by weighted averaging and fusing at least two sets of feature vectors. After introducing the first amplification factor to enhance the feature, it predicts the energy control behavior and outputs the first control prediction result. The specific steps are described in detail in A351-A353.

[0065] Conflict or low consistency scenario (0.2 < index < 0.7): The second fusion behavior prediction module first divides at least two groups of feature vectors into consistent and conflicting feature vector groups, performs weighted average fusion, and then uses the conflict demodulation network to reconcile the conflicting groups. The second control prediction result is output based on the fused and reconciled feature vectors. The specific steps are described in detail in A361-A365.

[0066] Data is missing or unrelated (index = empty): The third fusion behavior prediction module predicts at least two sets of feature vectors separately, and uses the output control behavior of each side as the third control prediction result. The specific steps are described in detail in A371-A372.

[0067] By constructing a hierarchical prediction model including intention perception and three-level fusion modules, Figure 1 The consistency indicators dynamically adapt to different collaborative scenarios, achieve accurate prediction of multi-side control behaviors, and improve the scientific nature of energy management strategies and system stability.

[0068] Furthermore, step A350 in the method provided in the embodiment of the present application includes: A351: Wherein, the first fusion behavior prediction module includes a first amplification coefficient.

[0069] A352: Activate the first fusion behavior prediction module to perform weighted average fusion on the at least two groups of feature vectors to obtain a first fused feature vector.

[0070] A353: Introduce the first amplification coefficient to perform feature enhancement on the first fused feature vector, output a first enhanced fused feature vector, predict energy control behavior on the first enhanced fused feature vector, and output a first control prediction result.

[0071] Specifically, first, after the first fusion behavior prediction module is activated, a weighted average fusion is performed on at least two sets of feature vectors. Based on preset side-by-side weights, such as 0.4 for the user side, 0.3 for the system side, and 0.3 for the management side, a weighted sum is taken of the behavior type, control object, and other dimensions of each vector to generate the first fused feature vector. For example, after weighted fusion of the user-side feature vector [energy saving, air conditioning, +10%, 0.5, 1, 0, 0] and the management-side vector [energy saving, all users, -15%, 0.5, 0, 0, 1], the fused vector [energy saving, air conditioning / all users, -5%, 0.5, 0.4, 0, 0.6] is obtained, which comprehensively reflects the synergy between the two sides in achieving the energy-saving goal.

[0072] Then, the first magnification factor (such as 1.2) is introduced to enhance the features of the fusion vector. By linearly amplifying the collaborative feature dimension, for example, the semantic weight related to energy saving is amplified from -5% to -6%, highlighting the multi-sided meaning. Figure 1 The first enhanced fusion feature vector is generated based on the impact of high consistency. After enhancement, the proportion of collaborative features in the vector increases, making the model more sensitive to high-consistency scenarios. Finally, energy control behavior prediction is performed based on the enhanced vector, and the first control prediction result is output.

[0073] Through the combined strategy of weighted average fusion and feature enhancement, the first fusion behavior prediction module achieves in-depth mining and enhanced expression of multi-sided highly consistent control intentions, thereby improving the execution strength and synergy effect of energy control strategies while ensuring prediction accuracy.

[0074] Furthermore, step A360 in the method provided in the embodiment of the present application includes: A361: Wherein, the second fusion behavior prediction module includes a conflict demodulation network.

[0075] A362: Activate the second fusion behavior prediction module to divide the at least two groups of feature vectors, and output a consistent feature vector group and a conflicting feature vector group.

[0076] A363: Perform weighted averaging fusion on the at least two groups of feature vectors to obtain a fused feature vector.

[0077] A364: Introducing the conflict demodulation network to reconcile the conflict feature vector group to obtain a reconciled feature vector, wherein the reconciliation target includes the response priority and the system load size.

[0078] A365: Predict energy control behavior based on the fused feature vector and the harmonized feature vector, and output a second control prediction result.

[0079] In the embodiment of the present application, the conflict demodulation network is a component in the second fusion behavior prediction module for reconciling conflicting feature vector groups.

[0080] In one embodiment, after the second fusion behavior prediction module is activated, it first performs a consistency analysis on at least two sets of feature vectors. Using a preset conflict threshold, such as a semantic similarity of less than 0.5, it determines a conflict and divides the feature vectors into a consistent feature vector group (e.g., fields related to equipment operating status on both sides) and a conflicting feature vector group (e.g., opposing fields representing a user-side load increase of +15% and a system-side load reduction of -10%). For example, in an industrial park scenario, the user-side feature vectors are [load increase, machine tool, +15%, 0.6, 1, 0, 0], and the system-side feature vectors are [load reduction, feeder, -10%, 0.6, 0, 1, 0]. After division, the consistent feature vector group becomes [equipment, time 0.6], and the conflicting feature vector group becomes [load increase +15%, load reduction -10%].

[0081] Subsequently, all original input feature vectors processed by the second fusion behavior prediction module, before any partitioning, are obtained. This represents the initial set of at least two feature vectors used for consistency analysis and subsequent weighted average fusion. Weighted average fusion is performed on these vectors to generate a basic fused feature vector, such as a combined power change of +5%. Simultaneously, a conflict demodulation network is activated for the conflicting feature vector groups. This network reconstructs the conflicting fields using a linear combination algorithm, using response priority (e.g., production demand priority 0.7, system stability priority 0.3) and system load size (e.g., triggering a load reduction priority strategy when the current load rate exceeds 80%) as reconciliation targets. For example, in the aforementioned conflict scenario, the demodulated output is a reconciled feature vector [power change +3%, priority 0.7:0.3], which preserves some user-side demand while keeping the system load within a safe threshold.

[0082] Finally, based on the combination of the fusion eigenvector and the harmonization eigenvector, such as the basic power +5% superimposed on the harmonized +3%, a total adjustment amount of +8% is formed, and the second control prediction result is output, thereby improving the system's ability to balance conflicting control demands.

[0083] Through the structured division, weighted fusion and directional coordination of multi-side feature vectors and conflict demodulation networks, the second fusion behavior prediction module achieves refined processing of control conflicts, ensuring system stability while maximizing the satisfaction of multi-side needs, thereby improving the scientificity and reliability of energy management strategies in complex scenarios.

[0084] Furthermore, step A370 in the method provided in the embodiment of the present application includes: A371: Activate the third fusion behavior prediction module to predict the at least two groups of feature vectors separately, and output the first-side regulation behavior and the second-side regulation behavior.

[0085] A372: The first-side control behavior and the second-side control behavior are output as the third control prediction result.

[0086] Optional, when regulating Figure 1 If the consistency index is returned as null, that is, the mean semantic similarity of the feature vectors on each side is lower than the preset threshold, such as <0.2, the third fusion behavior prediction module is activated and performs independent prediction processes on at least two sets of feature vectors. Take the unrelated bilateral data on the user side and the system side as an example: User-side feature vector processing: Taking [device on / off, air conditioner, +1, 0.3, 1, 0, 0] as an example, the module first extracts the device on / off behavior type, the air conditioner control object, the +1 control content (indicating the on state), and the 0.3 timestamp (at 30% of the time period) from the behavior field. After normalization, this is combined with the user-side embedding vector [1, 0, 0] to generate a feature vector. The module analyzes this vector using a time series prediction algorithm. Based on the temporal correlation between air conditioner on-time and power changes in historical data, for example, if similar devices have historically experienced an average power increase of 1.2-1.8 kW when on at 30% of the time period, the module predicts that the user-side power will increase by 1.5 kW. This algorithm achieves this prediction by identifying periodic and trending features in the time series.

[0087] System-side feature vector processing: Taking [voltage fluctuation, feeder A, -3%, 0.3, 0, 1, 0] as an example, the module extracts the voltage fluctuation behavior type, feeder A control object, -3% control content (voltage drop amplitude), and the 0.3 timestamp. This is combined with the system-side embedded vector [0, 1, 0] to generate a feature vector. The module analyzes the load model, which includes built-in voltage thresholds and load regulation rules. For example, a voltage drop of 2% or greater triggers reactive power compensation. Based on the current feeder load factor, assuming a 90% probability of exceeding 85% when the voltage drops by 3% based on historical data, the output system needs to initiate an independent strategy for reactive power compensation. This model simulates the dynamic relationship between voltage and load to predict and regulate system stability.

[0088] The above-mentioned side-by-side prediction results are directly output in parallel as the third control prediction results without the need for fusion processing.

[0089] By independently analyzing and parallelizing the feature vectors of multiple control sides, the third fusion behavior prediction module maintains the effectiveness of the prediction and the independence of the strategy in scenarios with sparse or unrelated data. This not only avoids errors caused by invalid fusion, but also ensures that the energy management system can quickly respond to unilateral control needs under complex working conditions, thereby improving the robustness and practicality of the overall prediction model.

[0090] Furthermore, step A350 in the method provided in the embodiment of the present application includes: A354: The intention perception module is used to perform semantic analysis on each feature vector in the at least two groups of feature vectors to obtain at least two groups of semantic vectors.

[0091] A355: Calculate semantic similarity for the at least two groups of semantic vectors respectively, output the semantic similarity mean, and obtain the control meaning with the semantic similarity mean. Figure 1 Consistency indicators.

[0092] In one embodiment, the intent perception module first performs vector-by-vector semantic analysis on at least two sets of feature vectors. Using natural language processing (NLP) techniques, it performs word segmentation, part-of-speech tagging, and entity recognition on text fields such as behavior type and control content in the feature vectors. For example, a power adjustment of +15% in the user-side feature vector is parsed into the semantic vector [user demand, load increase, 15%], a system-side load warning of -10% is parsed into [system status, load reduction, 10%], and a management-side peak-valley electricity price adjustment is parsed into [policy directive, price incentive, none]. These semantic fields are mapped into vectors in a high-dimensional space using a word vector model (such as Word2Vec), forming at least two sets of semantic vectors.

[0093] Then, the cosine similarity of each set of semantic vectors is calculated: taking the user-side and system-side vectors as an example, the cosine value of the angle between the two in the semantic space is calculated. If the value is 0.3 (indicating that the semantics of load increase and load reduction are opposite), it is judged as low consistency; if the similarity between the management-side and user-side vectors is 0.8 (such as energy saving and price incentives), it is judged as high consistency. All pairwise similarity values ​​are averaged and the semantic similarity mean is output, which is used as the control intention. Figure 1 Consistency index (value range 0-1).

[0094] Through the semantic analysis and similarity calculation process of the intent perception module, the system can accurately quantify the degree of coordination of behavioral intentions across multiple control sides, providing a decision-making basis for the fusion behavior prediction model. This step addresses the shortcomings of traditional methods in analyzing semantic associations across multi-source heterogeneous data, elevating the assessment of control intentions from simple numerical matching to a deeper understanding at the semantic level. This improves the accuracy of multi-side behavioral coordination predictions and lays the foundation for the dynamic triggering of hierarchical fusion strategies.

[0095] In summary, the energy management method combined with big data regulation behavior prediction provided by the embodiments of the present application has the following technical effects: This application detects the control behavior of the user control side, the system control side and the management control side within a preset detection time period, collects the control behavior data packets of multiple sides, obtains the feature vector through feature vectorization processing, and imports the fusion behavior prediction model to analyze the control intention. Figure 1 Consistency indicators are used to activate different modules for prediction based on the indicator results, and energy management is performed in combination with the control prediction results output by each module, thereby realizing energy management combined with big data control behavior prediction, making the energy control prediction results more comprehensive and accurate, achieving the technical effect of accurately perceiving control intentions, reconciling behavioral conflicts, improving the matching degree of energy scheduling and demand, and achieving efficient energy management.

[0096] Example 2, as Figure 2 As shown, based on the same inventive concept as in the aforementioned embodiment 1, the embodiment of the present application provides an energy management system combined with big data control behavior prediction, the system comprising: The time period setting module 1 is used to set a preset detection time period.

[0097] Behavior detection execution module 2, the behavior detection execution module 2 is used to perform control behavior detection on multiple control sides within the preset detection time period. If it is detected that at least two control sides among the multiple control sides trigger control behavior, at least two groups of control behavior data packets corresponding to the at least two control sides are output. The multiple control sides include user control side, system control side and management control side.

[0098] The vectorization processing module 3 is used to perform feature vectorization processing on the at least two groups of control behavior data packets, output at least two groups of feature vectors, import the at least two groups of feature vectors into the fusion behavior prediction model for prediction, and output the control prediction results.

[0099] The energy management execution module 4 is used to connect to the energy management system to execute energy management based on the control prediction result.

[0100] Furthermore, the behavior detection execution module 2 is used to perform the following steps: At the beginning of the preset detection time period, the main scheduler sends a behavior detection request to the multiple control sides, collects user-side control behavior data packets, system control side control behavior data packets and management control side control behavior data packets; performs content change analysis on the user-side control behavior data packets, system control side control behavior data packets and management control side control behavior data packets, and if it is detected that no control behavior is triggered in the multiple control sides, skips the current preset detection time period and enters the next preset detection time period; if it is detected that the multiple control sides are a single control side that triggers the control behavior, caches the control behavior data packet corresponding to the single control side.

[0101] Furthermore, the behavior detection execution module 2 is used to perform the following steps: At the end of the preset detection time period, no other control side different from the cache control side is detected to trigger a control behavior, and the energy management system performs energy management according to the control behavior data packet corresponding to the cache control side; at the end of the preset detection time period, at least one other control side different from the cache control side is detected to trigger a control behavior, and the control behavior data packet corresponding to the cache control side is extracted and combined with the control behavior data packet corresponding to the at least one other control side, and at least two groups of control behavior data packets are output.

[0102] Furthermore, the vectorized processing module 3 is configured to perform the following steps: The behavior fields of the at least two groups of control behavior data packets are extracted respectively, and the behavior type, control object, control content and behavior timestamp are output; standardization is performed according to the behavior type, control object, control content and behavior timestamp to obtain at least two groups of standardized behavior fields; a control side embedding vector is introduced to combine the at least two groups of standardized behavior fields, and at least two groups of feature vectors are output.

[0103] Furthermore, the vectorized processing module 3 is configured to perform the following steps: The fusion behavior prediction model includes an intention perception module, a first fusion behavior prediction module, a second fusion behavior prediction module and a third fusion behavior prediction module; the at least two sets of feature vectors are input into the intention perception module for analysis to obtain the control intention. Figure 1 Consistency indicators, if the regulatory intent Figure 1 If the consistency index is greater than or equal to the preset consistency index threshold, the first fusion behavior prediction module is activated to predict the at least two groups of feature vectors and output a first control prediction result; if the control intention Figure 1 If the consistency index is less than the preset consistency index threshold and the return value is not empty, the second fusion behavior prediction module is activated to predict the at least two groups of feature vectors and output a second control prediction result; if the control intention Figure 1 If the consistency index is returned as null, the third fusion behavior prediction module is activated to predict the at least two groups of feature vectors and output a third control prediction result.

[0104] Furthermore, the vectorized processing module 3 is configured to perform the following steps: Among them, the first fusion behavior prediction module includes a first amplification coefficient; the first fusion behavior prediction module is activated to perform weighted average fusion on the at least two groups of feature vectors to obtain a first fusion feature vector; the first amplification coefficient is introduced to perform feature enhancement on the first fusion feature vector, and a first enhanced fusion feature vector is output; the energy control behavior prediction is performed on the first enhanced fusion feature vector, and a first control prediction result is output.

[0105] Furthermore, the vectorized processing module 3 is configured to perform the following steps: Among them, the second fusion behavior prediction module includes a conflict demodulation network; the second fusion behavior prediction module is activated to divide the at least two groups of feature vectors, and output a consistent feature vector group and a conflict feature vector group; the at least two groups of feature vectors are weighted averaged and fused to obtain a fused feature vector; the conflict demodulation network is introduced to reconcile the conflict feature vector group to obtain a reconciled feature vector, wherein the reconciliation target includes the response priority and the system load size; energy control behavior prediction is performed based on the fused feature vector and the reconciled feature vector, and a second control prediction result is output.

[0106] Furthermore, the vectorized processing module 3 is configured to perform the following steps: Activate the third fusion behavior prediction module to predict the at least two groups of feature vectors respectively, and output the first-side regulation behavior and the second-side regulation behavior; output the first-side regulation behavior and the second-side regulation behavior as the third regulation prediction result.

[0107] Furthermore, the vectorized processing module 3 is configured to perform the following steps: The intention perception module is used to perform semantic analysis on each feature vector in the at least two groups of feature vectors to obtain at least two groups of semantic vectors; perform semantic similarity calculation on the at least two groups of semantic vectors respectively, output the semantic similarity mean, and obtain the control intention with the semantic similarity mean. Figure 1 Consistency indicators.

[0108] The energy management system combined with big data regulation behavior prediction provided by the embodiment of the present invention can execute the energy management method combined with big data regulation behavior prediction provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0109] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention.

[0110] The above specific embodiments do not constitute a limitation to the scope of protection of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of this application should be included in the scope of protection of this application. In some cases, the actions or steps recorded in this application can be performed in an order different from that in the embodiments and can still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. An energy management method combining big data control behavior prediction, characterized in that: The method comprises: Set a preset detection time period; Performing control behavior detection on multiple control sides within the preset detection time period, and if it is detected that at least two control sides among the multiple control sides trigger control behavior, outputting at least two groups of control behavior data packets corresponding to the at least two control sides, the multiple control sides including the user control side, the system control side, and the management control side; Performing feature vectorization processing on the at least two groups of control behavior data packets, outputting at least two groups of feature vectors, importing the at least two groups of feature vectors into a fusion behavior prediction model for prediction, and outputting a control prediction result; The energy management system is connected to perform energy management based on the control prediction result.

2. The method according to claim 1, wherein The method of detecting control behaviors of multiple control sides within the preset detection time period includes: At the beginning of the preset detection time period, the main scheduler sends a behavior detection request to the multiple control sides, and collects the user side control behavior data packet, the system control side control behavior data packet and the management control side control behavior data packet; Performing content change analysis on the user-side control behavior data packet, the system-side control behavior data packet, and the management-side control behavior data packet; if it is detected that no control behavior is triggered in the multiple control sides, skipping the current preset detection time period and entering the next preset detection time period; If it is detected that the multiple control sides trigger the control behavior for a single control side, the control behavior data packet corresponding to the single control side is cached.

3. The method according to claim 2, wherein After caching the control action data packet corresponding to the single control side, the method further includes: When no other control side different from the cache control side triggers a control behavior at the end of the preset detection time period, the energy management system performs energy management according to the control behavior data packet corresponding to the cache control side; At the end of the preset detection time period, at least one other control side different from the cache control side is detected to trigger a control behavior, the control behavior data packet corresponding to the cache control side is extracted and combined with the control behavior data packet corresponding to the at least one other control side, and at least two groups of control behavior data packets are output.

4. The method according to claim 1, wherein Performing feature vectorization processing on the at least two groups of control behavior data packets, the method comprising: Extracting the behavior fields of the at least two groups of control behavior data packets respectively, and outputting the behavior type, control object, control content, and behavior timestamp; Performing standardization processing according to the behavior type, control object, control content, and behavior timestamp to obtain at least two groups of standardized behavior fields; A regulatory side embedding vector is introduced to combine the at least two groups of standardized behavior fields, and at least two groups of feature vectors are output.

5. The method according to claim 1, wherein The at least two sets of feature vectors are imported into the fusion behavior prediction model for prediction, and the control prediction result is output. include: The fusion behavior prediction model includes an intention perception module, a first fusion behavior prediction module, a second fusion behavior prediction module and a third fusion behavior prediction module; Inputting the at least two sets of feature vectors into the intention perception module for analysis to obtain a control intention consistency index; if the control intention consistency index is greater than or equal to a preset consistency index threshold, activating the first fusion behavior prediction module to predict the at least two sets of feature vectors and outputting a first control prediction result; If the control intention consistency index is less than the preset consistency index threshold and the returned value is not empty, activating the second fusion behavior prediction module to predict the at least two groups of feature vectors and outputting a second control prediction result; If the control intention consistency index is returned as null, the third fusion behavior prediction module is activated to predict the at least two groups of feature vectors and output a third control prediction result.

6. The method according to claim 5, wherein Activate the first fusion behavior prediction module to predict the at least two groups of feature vectors and output a first control prediction result. include: Wherein, the first fusion behavior prediction module includes a first amplification coefficient; activating the first fusion behavior prediction module to perform weighted average fusion on the at least two groups of feature vectors to obtain a first fused feature vector; The first amplification coefficient is introduced to perform feature enhancement on the first fused feature vector, and a first enhanced fused feature vector is output. Energy control behavior is predicted on the first enhanced fused feature vector, and a first control prediction result is output.

7. The method according to claim 5, wherein Activate the second fusion behavior prediction module to predict the at least two groups of feature vectors and output a second control prediction result. include: Wherein, the second fusion behavior prediction module includes a conflict demodulation network; activating the second fusion behavior prediction module to divide the at least two groups of feature vectors and output a consistent feature vector group and a conflicting feature vector group; Performing weighted averaging fusion on the at least two groups of feature vectors to obtain a fused feature vector; Introducing the conflict demodulation network to reconcile the conflict feature vector group to obtain a reconciled feature vector, wherein the reconciliation target includes a response priority and a system load size; Energy regulation behavior prediction is performed based on the fused feature vector and the harmonized feature vector, and a second regulation prediction result is output.

8. The method according to claim 5, wherein Activating the third fusion behavior prediction module to predict the at least two groups of feature vectors and outputting a third control prediction result, the method includes: activating the third fusion behavior prediction module to predict the at least two groups of feature vectors respectively, and outputting a first-side regulation behavior and a second-side regulation behavior; The first-side control behavior and the second-side control behavior are output as the third control prediction result.

9. The method according to claim 5, wherein The intention perception module is used to perform semantic analysis on each feature vector in the at least two groups of feature vectors to obtain at least two groups of semantic vectors; Semantic similarity is calculated for each of the at least two groups of semantic vectors, and a semantic similarity mean is output, and a control intention consistency index is obtained using the semantic similarity mean.

10. An energy management system that combines big data control behavior prediction is characterized by: The energy management method combined with big data regulation behavior prediction according to any one of claims 1 to 9 is implemented, wherein the system comprises: The time period setting module is used to set the preset detection time period; a behavior detection execution module, configured to perform control behavior detection on multiple control sides within the preset detection time period, and output at least two groups of control behavior data packets corresponding to the at least two control sides if it is detected that at least two control sides among the multiple control sides trigger control behavior, wherein the multiple control sides include a user control side, a system control side, and a management control side; a vectorization processing module, configured to perform feature vectorization processing on the at least two groups of control behavior data packets, output at least two groups of feature vectors, import the at least two groups of feature vectors into a fusion behavior prediction model for prediction, and output a control prediction result; The energy management execution module is used to connect to the energy management system to execute energy management based on the regulation and prediction results.

Images