Big data processing system for socket energy efficiency optimization and energy consumption prediction

Abstract

The invention relates to the technical field of data processing, and discloses a socket energy efficiency optimization and energy consumption prediction big data processing system, which comprises at least one socket and a cloud server, an edge calculation unit of the socket converts the equipment electric signals collected in real time into a behavior morpheme sequence through an electric appliance morpheme extraction module; the cloud server receives the sequence, and a sequence prediction module of the cloud server predicts future power consumption behaviors, energy consumption and a dependency relationship; the decision state generation module fuses the prediction information to construct a comprehensive state vector; and the active intervention decision module generates and issues an optimal energy efficiency intervention instruction based on the vector. According to the system, accurate energy consumption prediction and energy efficiency optimization of power utilization of the socket are realized. According to the method, through deep analysis and attribution understanding of power consumption behaviors and in combination with multi-target optimization, intelligent decision making is carried out, and a comprehensive optimal intervention strategy can be generated.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a big data processing system for socket energy efficiency optimization and energy consumption prediction. Background Art

[0002] With the development of smart grid and IoT technologies, society's demand for refined energy management is growing. As the key node connecting the power grid and electrical devices, smart sockets are an important means for collecting energy consumption data and managing electricity usage in homes and offices. By monitoring and analyzing electrical signals at the socket level, they provide the technical prerequisite for optimizing the energy efficiency of individual electrical devices.

[0003] Currently, energy-saving applications around smart sockets typically focus on providing real-time power metering and basic remote or timed control. Some solutions use collected power data to perform simple threshold checks, for example, automatically shutting off power when a device enters standby mode. While these applications improve the convenience of power management, their core technologies remain superficially based on the utilization of electrical signal data.

[0004] However, existing technologies often simplify data processing to statistics or comparisons of power values, losing the rich time-series and frequency-domain features in the original electrical signal that can characterize the specific operating status of the equipment. Therefore, they are unable to build a deep understanding of the refined behavior of the equipment, let alone effectively predict future electricity consumption behavior and energy consumption trends. Therefore, their energy efficiency management strategies can only be lagging, passive, responsive control. Ultimately, due to the lack of predictive capabilities and contextual understanding of complex electricity consumption behaviors, their optimization strategies often rely on static rules, making it difficult to strike a balance between multiple objectives such as dynamic electricity prices, user needs, and equipment operating conditions, resulting in limited overall optimization effects.

[0005] Therefore, the present invention proposes a big data processing system for socket energy efficiency optimization and energy consumption prediction to address the deficiencies of the prior art. Summary of the Invention

[0006] Existing technologies have shortcomings when it comes to optimizing socket energy efficiency and predicting energy consumption. First, at the big data processing level, they are unable to extract deep patterns that characterize the subtle behavior of devices from massive electrical signals. Second, due to a lack of data processing capabilities, they are unable to accurately predict future trends at the energy consumption forecasting level, resulting in management strategies that can only respond passively. Finally, at the energy efficiency optimization level, their strategies are based on static rules and cannot combine prediction results with electricity usage context. This makes it difficult to strike a balance between multiple objectives such as electricity prices, user preferences, and equipment operating conditions, resulting in poor optimization effects and adaptability.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] A first aspect of the present invention provides a big data processing system for socket energy efficiency optimization and energy consumption prediction, comprising: at least one socket and a cloud server, wherein the socket is provided with an edge computing unit, and the cloud server is communicatively connected to the edge computing unit;

[0009] The edge computing unit includes:

[0010] an electrical appliance morpheme extraction module, configured to collect raw electrical signals from electrical appliances connected to the socket in real time, and process the raw electrical signals into a sequence of electrical appliance morphemes arranged in time sequence that characterizes the basic electrical behavior of the electrical appliances;

[0011] The cloud server includes:

[0012] A sequence prediction module is configured to receive the electric device morpheme sequence and, based on the historical context of the electric device morpheme sequence, predict the electricity consumption behavior in the future period of time, and generate a corresponding future electric device morpheme sequence, an energy consumption trend prediction result, and an attention weight representing the dependency relationship between the historical electric device morphemes;

[0013] A decision state generation module, configured to combine the energy consumption trend prediction result and the attention weight to generate a comprehensive state vector including the electric device behavior context information;

[0014] The active intervention decision module is configured to receive the integrated state vector and output an energy efficiency optimization intervention instruction for the socket based on the integrated state vector.

[0015] As a preferred technical solution, the electrical appliance morpheme extraction module operates in the following manner: collecting the original electrical signal at a preset frequency to form a signal window; performing a time-frequency transformation on the signal window to generate a time-frequency spectrum; and analyzing and classifying the time-frequency spectrum using a pre-set neural network, which is a convolutional neural network, to map each signal window to an electrical appliance morpheme. This process can be represented by the following formula:

[0016] m t =CNN(S t );

[0017] Where S t is the time-frequency spectrum generated at time t; CNN(·) represents the classification function performed by the convolutional neural network; m t The output of the corresponding time of the electrical equipment morpheme. Thus, the electrical equipment morpheme sequence {m1,m2,...,m t}.

[0018] As a preferred technical solution, the sequence prediction module runs through an encoder-decoder model based on the Transformer architecture. The encoder part of the model processes the historical context of the electrical device morpheme sequence, and the decoder part of the model generates the future electrical device morpheme sequence. The historical context of the electrical device morpheme sequence is a subsequence of electrical device morphemes of a preset length before the time point of prediction. The model further combines the future electrical device morpheme sequence with the mapping relationship between the preset electrical device morphemes and power consumption to generate the energy consumption trend prediction result. The attention weight is calculated by the multi-head self-attention mechanism built into the encoder-decoder model.

[0019] As a preferred technical solution, the decision state generation module operates specifically in the following manner: based on the attention weight, attribution analysis is performed on the historical electric equipment morphemes in the electric equipment morpheme sequence to identify the key historical electric equipment morphemes that play a decisive role in the current state; based on the results of the attribution analysis, a contextual criticality vector is generated that quantifies the degree of influence of the key historical electric equipment morphemes; the contextual criticality vector is spliced ​​with the energy consumption trend prediction result to generate the comprehensive state vector including the context information of the electric equipment behavior.

[0020] As a further preferred technical solution, the decision state generation module is further configured to splice an external environment information vector and a user preference vector when generating the comprehensive state vector. The external environment information vector includes at least one of real-time electricity price information and grid load status, and the user preference vector includes at least one of historical intervention instruction acceptance rates and user-preset mode preferences. This splicing process can be represented by the following formula:

[0021] S t =concat(V ctx ,R pred ,V env ,V user );

[0022] Where S t is the comprehensive state vector generated at time t; concat(·) represents the vector concatenation operation; V ctx is the context key vector; R pred is the energy consumption trend prediction result; V env is the external environment information vector; V user is the user preference vector.

[0023] As a preferred technical solution, the active intervention decision module uses a reinforcement learning agent that is a pre-trained deep Q-network model to run, taking the comprehensive state vector as the current state input, and the reinforcement learning agent selects the action that can maximize the long-term cumulative reward to generate the energy efficiency optimization intervention instruction. The action selection process can be represented by the following formula:

[0024] a t =argmax a Q(S t ,a;θ);

[0025] Where a t is the action selected at time t, i.e., the generated energy efficiency optimization intervention instruction; a is an action in the set of all possible actions; Q(S t ,a;θ) is the depth Q network in state S t The value function of executing action a under θ is the parameter of the network.

[0026] As a further preferred technical solution, the active intervention decision module outputs the energy efficiency optimization intervention instruction based on a multi-objective reward function, which includes at least three evaluation dimensions: expected energy savings, user satisfaction, and device health. This reward function can be represented by the following formula:

[0027] R t =w e ·E t +w u ·U t +w d ·D t ;

[0028] Where R t is the immediate reward at time t; E t 、U t 、D t represent the quantitative values ​​of expected energy savings, user satisfaction, and equipment health, respectively; w e 、w u 、w d is the corresponding weight coefficient.

[0029] As a preferred technical solution, the cloud server also includes a dynamic morpheme library update module, which is used to generate new electrical equipment morpheme definitions when the sequence prediction module identifies a new electrical equipment behavior pattern that cannot be accurately described by the existing electrical equipment morphemes, and send the new electrical equipment morpheme definitions to the electrical appliance morpheme extraction module to dynamically expand the electrical equipment morpheme library.

[0030] A second aspect of the present invention provides a big data processing method for socket energy efficiency optimization and energy consumption prediction, which is applied to the aforementioned system and includes the following steps:

[0031] Collecting original electrical signals of electrical devices connected to the socket in real time, and processing the original electrical signals into a time-series sequence of electrical device morphemes representing basic electrical behavior of the electrical devices;

[0032] Receive the electric device morpheme sequence, and based on the historical context of the electric device morpheme sequence, predict the electricity consumption behavior in a future period of time, and generate a corresponding future electric device morpheme sequence, an energy consumption trend prediction result, and an attention weight representing the dependency relationship between the historical electric device morphemes;

[0033] Combining the energy consumption trend prediction result and the attention weight, generating a comprehensive state vector including the electric device behavior context information;

[0034] The integrated state vector is received, and an energy efficiency optimization intervention instruction for the socket is output based on the integrated state vector.

[0035] The present invention provides a big data processing system for socket energy efficiency optimization and energy consumption prediction. It has the following beneficial effects:

[0036] 1. This invention utilizes an appliance morpheme extraction module and a sequence prediction module to deeply process the raw electrical signal data stream collected by sockets, converting it into a structured morpheme sequence of electrical devices and performing time series prediction. This enables the system to accurately predict future energy consumption trends, transitioning from passive response to active prediction and providing a key decision-making basis for subsequent energy efficiency optimization and scheduling.

[0037] 2. This invention utilizes a decision-state generation module to perform attribution analysis on historical electrical device morphemes using attention weights, thereby generating a comprehensive state vector containing precise contextual information. This enables the system to not only predict the future but also understand the key historical behaviors that led to those predictions, significantly improving the relevance and effectiveness of subsequent energy efficiency optimization interventions.

[0038] 3. This invention utilizes a reinforcement learning-based active intervention decision-making module and configures a multi-objective reward function that incorporates expected energy savings, user satisfaction, and device health. This enables the system to automatically balance multiple dimensions when making energy efficiency optimization decisions, finding the optimal intervention strategy overall. This surpasses traditional simple rule-based systems that focus solely on energy conservation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a diagram showing the architecture of a big data processing system for socket energy efficiency optimization and energy consumption prediction according to the present invention;

[0040] Figure 2 This is a flowchart of electrical appliance morpheme extraction according to the present invention;

[0041] Figure 3 A schematic diagram of the internal structure of a sequence prediction module according to the present invention;

[0042] Figure 4 A schematic diagram of the internal processing flow of the decision state generation module according to the present invention;

[0043] Figure 5 A schematic diagram of the operation flow of the active intervention decision module according to the present invention;

[0044] Figure 6 A schematic diagram of the operation flow of the dynamic morpheme library updating module according to the present invention;

[0045] Figure 7 This is a flow chart of the big data processing method for socket energy efficiency optimization and energy consumption prediction according to the present invention.

[0046] Among them, 10, edge computing unit; 110, electrical appliance morpheme extraction module; 20, cloud server; 210, sequence prediction module; 220, decision state generation module; 230, active intervention decision module; 240, dynamic morpheme library update module. DETAILED DESCRIPTION

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0048] Refer to the attached Figure 1 , Figure 1 This diagram illustrates the architecture of a big data processing system for socket energy efficiency optimization and energy consumption prediction according to one embodiment of the present invention. The present invention provides a big data processing system for socket energy efficiency optimization and energy consumption prediction, comprising at least one socket (not shown) and a cloud server 20. The socket is equipped with an edge computing unit 10, which establishes a communication connection with the cloud server 20 via methods such as Wi-Fi, Zigbee, or a mobile communication network.

[0049] The edge computing unit 10 includes: an electrical appliance morpheme extraction module 110 .

[0050] The cloud server 20 includes a sequence prediction module 210 , a decision state generation module 220 , an active intervention decision module 230 , and a dynamic morpheme library update module 240 .

[0051] Refer to the attached Figure 2 , Figure 2 FIG2 is a flowchart of an electrical appliance morpheme extraction process according to an embodiment of the present invention. The process is executed by the electrical appliance morpheme extraction module 110 in the edge computing unit 10 disposed in the socket.

[0052] In a specific embodiment, the function of the electrical morpheme extraction module 110 is to convert the continuous, unstructured original electrical signal generated by the electrical equipment into a discrete, structured time series data, namely, the electrical equipment morpheme sequence, to provide high-quality input for the in-depth analysis and prediction of the cloud server 20.

[0053] The module first uses built-in acquisition circuitry to collect raw electrical signals from electrical devices flowing through the socket in real time at a preset sampling frequency (e.g., 10kHz or higher). The raw electrical signals can be current, voltage, or a combination of both. The high sampling frequency ensures that transient characteristics that characterize device state transitions can be captured.

[0054] The collected continuous signal stream is then segmented into time-sequential signal windows of fixed duration (e.g., 2 seconds). For each signal window, the module performs a time-frequency domain transform to generate a time-spectrogram. In one specific implementation, the time-frequency domain transform method used is the short-time Fourier transform (STFT). The resulting time-spectrogram is a two-dimensional matrix that shows how the frequency components of the signal change over time within the signal window.

[0055] Subsequently, the appliance morpheme extraction module 110 uses a pre-configured convolutional neural network (CNN) neural network model to analyze and classify the spectrogram generated in the previous step. Because the spectrogram is structurally similar to image data, the CNN effectively extracts local and global texture features. These features are highly correlated with the specific operating states of the electrical equipment (e.g., start, stop, constant power, variable frequency, etc.).

[0056] The convolutional neural network model receives a time-frequency spectrum as input and outputs a classification result, which is defined as an electrical equipment morpheme. This process can be represented by the following formula:

[0057] m t =CNN(S t );

[0058] Where S tis the time-frequency spectrum generated at time t; CNN(·) represents the classification function performed by the convolutional neural network; m t The output morpheme of the electric device at the corresponding time is, for example, one of "compressor start", "standby state", "heating module constant power operation" or "motor idling".

[0059] By sequentially executing the above-mentioned acquisition, transformation and classification processes on the continuous signal windows, the electrical appliance morpheme extraction module 110 finally generates a time-sequential electrical appliance morpheme sequence {m1, m2, ..., m t The sequence is then sent to the cloud server 20 for subsequent processing via the communication link between the edge computing unit 10 and the cloud server 20 .

[0060] Refer to the attached Figure 3 , Figure 3 2 is a schematic diagram of the internal structure of a sequence prediction module according to an embodiment of the present invention. The sequence prediction module 210 provided in the cloud server 20 is configured to receive the electrical device morpheme sequence generated by the electrical device morpheme extraction module 110 and perform sequence-to-sequence prediction.

[0061] In a specific embodiment, the core of the sequence prediction module 210 is an encoder-decoder model based on the Transformer architecture. This module first converts the received discrete morpheme sequence of electrical devices into a sequence of continuous real number vectors through a word embedding layer for easy processing by the neural network model.

[0062] The encoder portion of the model is responsible for processing the historical context of the morpheme sequence of electrical devices. This historical context is defined as a subsequence of morphemes of electrical devices of a preset length preceding the prediction time point. The encoder, through its internal multi-layer stacking structure, encodes the input historical context vector sequence and generates a set of intermediate representation vectors that represent the deep semantic information of the historical context.

[0063] The core computational unit within the encoder is the Multi-Head Self-Attention mechanism. This mechanism allows the encoder to calculate the degree of association between any morpheme in a sequence and all other morphemes in the sequence. This association is quantified as an attention weight. The calculation process of the attention weight can be represented by the following formula:

[0064]

[0065] Where Q, K, and V represent the query matrix, key matrix, and value matrix, respectively, which are all obtained by linear transformation of the input vector sequence. k is the dimension of the key vector. The result of this formula is the weighted sum of the value vectors, which reflects the importance of different parts of the sequence to the current part. The attention score matrix generated in this calculation process That is, the attention weight representing the dependency relationship between the historical electrical equipment morphemes is output to the decision state generation module 220.

[0066] The decoder of this model receives the intermediate representation vectors output by the encoder and, using an autoregressive approach, generates the morphemes of electrical devices for future time steps one by one. At each time step, the decoder combines the encoder output with the morphemes generated at the previous time step to predict the morpheme most likely to appear at the current time step, ultimately constructing a complete sequence of morphemes for future electrical devices.

[0067] After generating a future sequence of electrical device morphemes, the sequence prediction module 210 further combines this future sequence with a preset mapping between electrical device morphemes and power consumption to generate an energy consumption trend forecast. This mapping can be a lookup table storing the average power consumption value corresponding to each morpheme, or a small regression model. By converting each morpheme in the future sequence into a corresponding power consumption value, the module outputs a detailed energy consumption curve for a period of time in the future, i.e., the energy consumption trend forecast.

[0068] Refer to the attached Figure 4 , Figure 4 Figure 2 is a schematic diagram of the internal processing flow of a decision state generation module according to one embodiment of the present invention. Decision state generation module 220, provided in cloud server 20, integrates prediction information from sequence prediction module 210 with contextual information, and combines it with external data to construct a comprehensive, integrated state vector for subsequent decision making.

[0069] In a specific embodiment, the decision state generation module 220 first receives two core data output from the sequence prediction module 210: energy consumption trend prediction result and attention weight.

[0070] This module performs attribution analysis on historical consumer morphemes in the consumer morpheme sequence based on attention weights. Because the attention weight matrix quantifies the contribution of each morpheme in the historical sequence to generating the prediction result, analyzing this matrix can identify one or more key historical consumer morphemes that play a decisive role in the current and future states.

[0071] Based on the results of the attribution analysis, the module generates a contextual criticality vector. This vector quantitatively describes key historical electrical equipment morphemes and their impact. For example, if the analysis finds that the historical morpheme "compressor start" has the highest contribution to the predicted high energy consumption for a period of time, the contextual criticality vector will have high values ​​in the corresponding dimensions, thus explicitly representing this key causal relationship in the state vector.

[0072] After generating the context criticality vector, the decision state generation module 220 performs a vector concatenation operation to generate a final integrated state vector. In a specific implementation, the module concatenates the context criticality vector with the energy consumption trend prediction result.

[0073] Furthermore, to provide a more comprehensive basis for decision-making, the module can also obtain an external environmental information vector and a user preference vector from an external data interface and concatenate them. The external environmental information vector includes at least one of real-time electricity price information and grid load status. The user preference vector includes at least one of the intervention command acceptance rate calculated from historical interactions and the user's preset mode preference (e.g., energy-saving mode or performance mode).

[0074] The complete splicing process can be characterized by the following formula:

[0075] S t =concat(V ctx ,R pred ,V env ,V user );

[0076] Where S t is the comprehensive state vector generated at time t, which will serve as the input of the next module, namely the active intervention decision module 230; concat(·) represents the vector concatenation operation; V ctx is the context key vector; R pred is the energy consumption trend prediction result; V env is the external environment information vector; V user is the user preference vector.

[0077] Refer to the attached Figure 5 , Figure 5 Figure 2 is a schematic diagram of the operational flow of the active intervention decision module according to one embodiment of the present invention. Active intervention decision module 230, provided in cloud server 20, is configured to receive the comprehensive state vector generated by decision state generation module 220 and, based on the vector, output a specific energy efficiency optimization intervention instruction for the socket.

[0078] In a specific embodiment, the module is run by a reinforcement learning agent that is a pre-trained deep Q network (DQN) model. The reinforcement learning agent converts the comprehensive state vector S output by the decision state generation module 220 into t As the current state input.

[0079] Based on this state input, the agent uses its internal value function network to evaluate the long-term cumulative reward that can be obtained by executing each possible action in the current state, and selects the action that can maximize the expected value of this reward to generate energy-efficient intervention instructions. This action selection process can be represented by the following formula:

[0080] a t =argmax a Q(S t ,a;θ);

[0081] Where a t is the action selected at time t, i.e., the generated energy efficiency optimization intervention instruction, such as “delayed start”, “reduced power operation” or “immediate shutdown”; a is an action in the set of all possible actions; Q(S t ,a;θ) is the depth Q network in state S t The value function of executing action a under θ is the parameter of the network.

[0082] The reinforcement learning agent's decision-making is based on a multi-objective reward function used during its pre-training process. This reward function is used during the training phase to provide quantitative feedback signals for the immediate results of the agent performing an action in a specific state. In a specific implementation, this reward function includes at least three evaluations: expected energy savings, user satisfaction, and device health. This reward function can be represented by the following formula:

[0083] R t =w e ·E t +w u ·U t +w d ·D t ;

[0084] Where R t is the immediate reward at time t; E t 、U t 、D t represent the quantitative values ​​of expected energy savings, user satisfaction, and equipment health, respectively; w e 、w u 、w d is the corresponding weight coefficient, which is used to adjust the relative importance between different optimization objectives so that the final decision can achieve a balance between multiple dimensions.

[0085] Refer to the attached Figure 6 , Figure 6 The following is a schematic diagram of the operation flow of the dynamic morpheme library update module according to one embodiment of the present invention. The dynamic morpheme library update module 240 provided in the cloud server 20 is used to implement adaptive expansion of the morpheme library of electrical devices to ensure that the system can identify and process emerging electricity usage behavior patterns.

[0086] In one specific embodiment, the module's update process is triggered by specific conditions. When the sequence prediction module 210 processes a sequence of electrical device morphemes, if its predicted output for future morphemes has low confidence across all known morpheme categories, or if the edge computing unit 110 reports a time-frequency spectrum that its convolutional neural network cannot effectively classify, the system determines that a new electricity usage behavior pattern has emerged that cannot be accurately described by the existing electrical device morphemes. This triggers the execution of the dynamic morpheme library update module 240.

[0087] Once triggered, the module first retrieves the raw signal window data and its spectrogram corresponding to the new electricity usage behavior pattern from the data log. The module accumulates multiple spectrogram samples marked as unknown. The module then applies an unsupervised clustering algorithm (e.g., DBSCAN or K-Means) to this accumulated set of unknown spectrogram samples, grouping samples with similar characteristics. Each resulting stable cluster is considered a candidate new electricity device morpheme.

[0088] For each candidate new electric device morpheme, the module generates a new morpheme definition for it. This definition includes assigning a unique new morpheme identifier and extracting a feature representation of the cluster (e.g., the centroid of all time-spectrograms within the cluster or a representative sample).

[0089] The module then uses these sample data with new morpheme identifiers to update or fine-tune a baseline convolutional neural network model deployed on the cloud server 20. This process integrates new electricity usage behavior patterns and their corresponding morpheme definitions into the capabilities of the classification model.

[0090] After completing the model update and verification in the cloud, the dynamic morpheme library update module 240 sends the updated convolutional neural network model parameters to one or more designated edge computing units 10 via the cloud-edge communication link. After receiving the new model parameters, the appliance morpheme extraction module 110 within the edge computing unit 10 replaces its local preset neural network.

[0091] Through the above steps, the morpheme library of electrical equipment is dynamically expanded. Subsequently, when an electrical equipment exhibits this new behavior pattern again, the edge computing unit 10 can accurately identify it as a new electrical equipment morpheme using the updated model and report it to the cloud server 20, thus ensuring the entire big data processing system's adaptability to new equipment or new equipment operating conditions and its long-term operational accuracy.

[0092] Refer to the attached Figure 7 , Figure 7 This is a flow chart of a big data processing method for socket energy efficiency optimization and energy consumption prediction according to an embodiment of the present invention. The method provided by the present invention, applied to the system of the aforementioned embodiment 1, may include the following steps:

[0093] S10 collects raw electrical signals from electrical devices connected to the socket in real time and processes them into a time-series sequence of electrical device morphemes that characterize the basic electrical behavior of the electrical devices. This step is performed by the appliance morpheme extraction module 110 within the edge computing unit 10. Specifically, this step includes: collecting raw electrical signals at a preset frequency to form a signal window; performing a time-frequency transformation on the signal window to generate a time-frequency spectrum; and classifying the time-frequency spectrum using a preset convolutional neural network to map each signal window to an electrical device morpheme, thereby generating an electrical device morpheme sequence.

[0094] S20, receiving the morpheme sequence of electric equipment, and based on the historical context of the morpheme sequence of electric equipment, predicting the electricity consumption behavior in the future period, and generating the corresponding future morpheme sequence of electric equipment, energy consumption trend prediction results, and attention weights that characterize the dependency relationship between historical electric equipment morphemes. This step is performed by the sequence prediction module 210 in the cloud server 20. Specifically, this step utilizes an encoder-decoder model based on the Transformer architecture, with its encoder part processing the historical context and its decoder part generating the future morpheme sequence of electric equipment. At the same time, the attention weight is calculated through the multi-head self-attention mechanism built into the model, and the energy consumption trend prediction result is generated by combining the future sequence with the preset power consumption mapping relationship.

[0095] S30 combines the energy consumption trend prediction results and attention weights to generate a comprehensive state vector containing contextual information about the electrical device's behavior. This step is performed by the decision state generation module 220 within the cloud server 20. Specifically, this step includes: performing attribution analysis on historical electrical device morphemes based on attention weights to identify key historical electrical device morphemes and generate a contextual key vector; and concatenating the contextual key vector with the energy consumption trend prediction results, as well as the optional external environment information vector and user preference vector, to generate a comprehensive state vector.

[0096] S40: Receive the integrated state vector and output energy efficiency optimization intervention instructions for the socket based on the integrated state vector. This step is performed by the active intervention decision module 230 within the cloud server 20. Specifically, this step utilizes a reinforcement learning agent, a pre-trained deep Q-network model, which takes the integrated state vector as the current state input. The reinforcement learning agent selects an action that maximizes the long-term cumulative reward based on a multi-objective reward function (which evaluates at least expected energy savings, user satisfaction, and device health) to generate energy efficiency optimization intervention instructions.

[0097] In a further embodiment, the method may further comprise the following steps:

[0098] S50: When a new electricity usage behavior pattern is identified that cannot be accurately described by existing electrical device morphemes, a new electrical device morpheme definition is generated and sent to the appliance morpheme extraction module to dynamically expand the electrical device morpheme library. This step is performed by the dynamic morpheme library update module 240 within the cloud server 20 to ensure the system's adaptability to new devices and new operating conditions.

[0099] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A big data processing system for socket energy efficiency optimization and energy consumption prediction, characterized in that: include: At least one socket, wherein an edge computing unit is provided in the socket, and the edge computing unit includes an electrical appliance morpheme extraction module; The electrical appliance morpheme extraction module is used to collect original electrical signals from the electrical appliances connected to the socket in real time, and process the original electrical signals into a sequence of electrical appliance morphemes arranged in time sequence that characterizes the basic electrical consumption behavior of the electrical appliances; A cloud server is in communication with the edge computing unit, and includes a sequence prediction module, a decision state generation module, and an active intervention decision module: The sequence prediction module is used to receive the electric device morpheme sequence and, based on the historical context of the electric device morpheme sequence, predict the electricity consumption behavior in the future period of time, and generate the corresponding future electric device morpheme sequence, energy consumption trend prediction results, and attention weights representing the dependency relationship between historical electric device morphemes; The decision state generation module is used to combine the energy consumption trend prediction result and the attention weight to generate a comprehensive state vector including the electric device behavior context information; The active intervention decision module is used to receive the comprehensive state vector and output an energy efficiency optimization intervention instruction for the socket based on the comprehensive state vector.

2. A big data processing system for socket energy efficiency optimization and energy consumption prediction according to claim 1, characterized in that: The electrical appliance morpheme extraction module is specifically used for: Sampling the original electrical signal at a preset frequency to form a signal window; Performing time-frequency domain transformation on the signal window to generate a time-frequency spectrum diagram; The time-frequency spectrum is analyzed and classified using a preset neural network to map each signal window into an electrical equipment morpheme, thereby generating the electrical equipment morpheme sequence arranged in time sequence and representing the basic electrical consumption behavior of the electrical equipment. The preset neural network is a convolutional neural network.

3. The big data processing system for socket energy efficiency optimization and energy consumption prediction according to claim 1, characterized in that: The sequence prediction module is specifically used for: The encoder-decoder model based on the Transformer architecture processes the historical context of the electric device morpheme sequence for use by its encoder part, and generates the future electric device morpheme sequence by its decoder part; The model further combines the future electric device morpheme sequence with a preset mapping relationship between electric device morphemes and power consumption to generate the energy consumption trend prediction result; The attention weight is calculated by the multi-head self-attention mechanism built into the encoder-decoder model; The historical context of the electric device morpheme sequence is an electric device morpheme subsequence of a preset length before the time point of the prediction.

4. The big data processing system for socket energy efficiency optimization and energy consumption prediction according to claim 1, characterized in that: The decision state generation module is specifically used for: Based on the attention weights, performing attribution analysis on historical electric device morphemes in the electric device morpheme sequence to identify key historical electric device morphemes that play a decisive role in the current state; generating, based on the result of the attribution analysis, a contextual criticality vector quantifying the degree of influence of the key historical electrical equipment morphemes; The context criticality vector is concatenated with the prediction result to generate the comprehensive state vector including the electric device behavior context information.

5. The big data processing system for socket energy efficiency optimization and energy consumption prediction according to claim 4, characterized in that: When generating the comprehensive state vector, the decision state generation module is further used to splice an external environment information vector and a user preference vector, wherein the external environment information vector includes at least one of real-time electricity price information and grid load status, and the user preference vector includes at least one of a historical intervention instruction acceptance rate and a user-preset mode preference.

6. The big data processing system for socket energy efficiency optimization and energy consumption prediction according to claim 1, characterized in that: The active intervention decision module is specifically used to: A reinforcement learning agent, which is a pre-trained deep Q-network model, takes the comprehensive state vector as the current state input, and selects an action that can maximize the long-term cumulative reward by the reinforcement learning agent to generate the energy efficiency optimization intervention instruction.

7. The big data processing system for socket energy efficiency optimization and energy consumption prediction according to claim 6, characterized in that: The decision basis for the active intervention decision module to output the energy efficiency optimization intervention instruction is a multi-objective reward function, and the multi-objective reward function includes at least an evaluation of three dimensions: expected energy savings, user satisfaction, and equipment health.

8. The big data processing system for socket energy efficiency optimization and energy consumption prediction according to claim 1, characterized in that: The cloud server also includes: The dynamic morpheme library update module is used to generate a new electric device morpheme definition when the sequence prediction module identifies a new electric device behavior pattern that cannot be accurately described by the existing electric device morphemes, and send the new electric device morpheme definition to the electrical appliance morpheme extraction module to dynamically expand the electric device morpheme library.

9. A big data processing method for socket energy efficiency optimization and energy consumption prediction, applied to a big data processing system for socket energy efficiency optimization and energy consumption prediction according to any one of claims 1 to 8, characterized in that: The following steps are involved: Collecting original electrical signals of electrical devices connected to the socket in real time, and processing the original electrical signals into a time-series sequence of electrical device morphemes representing basic electrical behavior of the electrical devices; Receive the electric device morpheme sequence, and based on the historical context of the electric device morpheme sequence, predict the electricity consumption behavior in a future period of time, and generate a corresponding future electric device morpheme sequence, an energy consumption trend prediction result, and an attention weight representing the dependency relationship between the historical electric device morphemes; Combining the energy consumption trend prediction result and the attention weight, generating a comprehensive state vector including the electric device behavior context information; The integrated state vector is received, and an energy efficiency optimization intervention instruction for the socket is output based on the integrated state vector.

Images