An energy prediction system based on quantum particle swarm federal space-time coupling

Abstract

This invention provides an energy prediction system based on quantum particle swarm optimization (QPSO) federated spatiotemporal coupling, comprising an edge sensing layer, a federated computing layer, a quantum optimization layer, and a spatiotemporal prediction layer. The edge sensing layer collects operational data from distributed energy nodes and performs data cleaning and feature encoding. The federated computing layer jointly models the spatial relationships between energy nodes and the temporal evolution characteristics of operational data, generating model update information. The quantum optimization layer triggers a federated aggregation process when preset conditions are met and updates the model aggregation weights based on the QPSO mechanism, generating global model parameters. The spatiotemporal prediction layer predicts and analyzes the operating status of the energy system based on the global model parameters and outputs risk assessment information. This invention achieves multi-node collaborative modeling without centralizing raw energy data, which is beneficial for improving the adaptability and application feasibility of the energy prediction system in complex operating scenarios.

Description

Technical Field

[0001] This invention belongs to the field of energy system operation status prediction technology, specifically relating to an energy prediction system based on quantum particle swarm federation spatiotemporal coupling. Background Technology

[0002] New energy power generation equipment mainly includes wind turbines, photovoltaic inverters, energy storage systems, and their supporting electrical units. These are key infrastructure components for new energy power generation systems to achieve power production and grid connection. With the continuous expansion of new energy installed capacity, the number of devices is growing exponentially, and operating scenarios are becoming increasingly complex. Operation and maintenance management systems need to perform real-time monitoring, maintenance scheduling, and status assessment of widely distributed, diverse, and environmentally varied equipment to ensure the safety, economy, and continuity of the system. Against this backdrop, how to achieve comprehensive perception of equipment operating status, dynamic assessment of health levels, and scientific scheduling of operation and maintenance tasks has become an important research direction for the intelligent operation and maintenance of new energy power plants.

[0003] Existing new energy operation and maintenance management models mostly adopt periodic inspections and manual task assignment, with the dispatch center allocating tasks based on alarm information or experience. This approach has the following problems: First, task generation relies on manual judgment, lacking quantitative analysis of equipment health status, easily leading to delayed maintenance or repetitive work; second, dispatch strategies are usually based on fixed schedules or single priority rules, failing to comprehensively consider multi-dimensional factors such as weather, grid operation, traffic conditions, and resource constraints, resulting in low task execution efficiency; third, existing systems are mostly hierarchical management structures, with delayed information transmission and a lack of dynamic replanning capabilities for emergencies, making it difficult to adapt to the rapidly changing operating environment of new energy power plants. These shortcomings lead to inaccurate equipment health management, low utilization of operation and maintenance resources, and untimely dispatch response, affecting the safe and stable operation of new energy power generation systems.

[0004] To address the issues of insufficient intelligence in task generation, lack of global optimization in scheduling strategies, and inability to respond to operational disturbances in the existing operation and maintenance of new energy power generation equipment, there is an urgent need for a technical solution that can comprehensively assess equipment health, external operating environment, and operation and maintenance resource conditions to achieve proactive generation, intelligent scheduling, and dynamic optimization of operation and maintenance tasks, thereby improving the accuracy of operation and maintenance decisions and the efficiency of execution for new energy power generation equipment. Summary of the Invention

[0005] To address the problems of existing technologies, embodiments of the present invention provide an energy prediction system based on quantum particle swarm federation spatiotemporal coupling. The technical solution is as follows: On the one hand, an energy prediction system based on quantum particle swarm federation spatiotemporal coupling is provided, including an edge sensing layer, a federated computing layer, a quantum optimization layer and a spatiotemporal prediction layer; The edge sensing layer is used to collect the operating data of distributed energy nodes, clean the operating data and generate feature representations. The federated computing layer is used to receive the feature representation, model the spatial correlation between energy nodes, and model the temporal evolution characteristics of energy operation data to form a local spatiotemporal joint model and generate model update information. The quantum optimization layer is used to receive model update information from multiple federated computing layers, trigger the federated aggregation process when preset conditions are met, and update the federated aggregation weights based on the quantum particle swarm optimization mechanism to generate global model parameters. The spatiotemporal prediction layer is used to predict and analyze the operating status of the energy system within a future time window based on the global model parameters, and output the prediction results and corresponding risk assessment information.

[0006] Furthermore, the edge sensing layer includes a data acquisition unit, a data cleaning unit, and a feature encoding unit; The data acquisition unit is used to collect electrical parameter data, environmental parameter data, and energy output data of the energy equipment. The data cleaning unit is used to perform abnormal data removal, missing data filling or marking, and time alignment processing on the collected data. The feature encoding unit is used to map the cleaned data into a feature representation of a uniform dimension.

[0007] Furthermore, the federated computing layer includes a spatial feature modeling unit, a temporal feature modeling unit, and a gradient generation and privacy protection unit; The spatial feature modeling unit models the spatial relationships between energy nodes based on the topology of the energy system. The time feature modeling unit models the time series characteristics of energy operation data based on the spatial feature modeling results. The gradient generation and privacy protection unit is used to generate model update information and encrypt the model update information after performing privacy protection processing.

[0008] Furthermore, before encrypting the model update information, the gradient generation and privacy protection unit performs noise perturbation processing on the model update information to reduce the risk of privacy leakage in the model update information.

[0009] Furthermore, the quantum optimization layer includes a gradient decryption unit, a quantum particle encoding unit, a federated aggregation and parameter optimization unit, and a model generation and distribution unit; The gradient decryption unit is used to decrypt the encrypted model update information. The quantum particle encoding unit is used to map federated aggregation weights into quantum particle form; The federated aggregation and parameter optimization unit is used to update the federated aggregation weights based on the quantum particle swarm optimization mechanism; The model generation and distribution unit is used to generate global model parameters and distribute them.

[0010] Furthermore, the quantum optimization layer triggers the federated aggregation process when the model update information meets the preset threshold condition, and performs federated aggregation and parameter update operations after triggering.

[0011] Furthermore, the spatiotemporal prediction layer includes a feature fusion unit, a prediction modeling unit, and a risk assessment unit; The feature fusion unit is used to fuse spatial features, temporal features, and environmental features. The predictive modeling unit is used to predict the operating status of the energy system based on fused features and global model parameters; The risk assessment unit is used to assess the risk of the prediction results and output risk assessment information.

[0012] Furthermore, the edge perception layer, federated computing layer, quantum optimization layer, and spatiotemporal prediction layer interact with each other via communication connections to exchange data and model parameters.

[0013] Furthermore, the feature fusion unit performs splicing processing on spatial features, temporal features, and environmental features to form a fused feature representation for prediction modeling.

[0014] Furthermore, after completing one predictive analysis and outputting the results, the system enters the next round of operation, repeatedly executing the processes of data collection, local modeling, federated aggregation, model updating, and predictive analysis to achieve continuous updates of model parameters and prediction results.

[0015] The beneficial effects of the technical solution provided by the embodiments of the present invention are as follows: This invention provides an energy prediction system based on quantum particle swarm federation spatiotemporal coupling. By collecting, cleaning, and feature-encoding the operating data of distributed energy nodes at the edge, and completing multi-node collaborative modeling without centralizing the original data, it realizes the distributed execution of data processing and model training in the prediction of the operating status of the energy system, effectively reducing the communication pressure and data security risks brought about by centralized data processing.

[0016] This invention improves the stability and adaptability of the model in complex energy system scenarios by jointly modeling the spatial correlation between energy nodes and the temporal evolution characteristics of operational data in the federated computing layer, and by introducing a quantum particle swarm optimization mechanism in the quantum optimization layer to dynamically update the model aggregation weights. This enables the aggregation process of model parameters to be adaptively adjusted according to the operational characteristics of different nodes.

[0017] Furthermore, by distributing global model parameters to the prediction layer for operational status prediction and risk assessment, this invention achieves unified analysis and output of the future operational status of the energy system. This enables the system to maintain the consistency and reliability of prediction results under conditions of multi-node collaboration and continuous updates, thereby enhancing the feasibility of energy prediction systems in actual operating environments. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the system architecture of an energy prediction system based on quantum particle swarm federation spatiotemporal coupling according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the operation flow of an energy prediction system based on quantum particle swarm federation spatiotemporal coupling according to an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0021] The following is in conjunction with the appendix Figure 1 and attached Figure 2 This embodiment describes the structural composition and operation of an energy prediction system based on quantum particle swarm federation spatiotemporal coupling. This embodiment is only for illustrating the technical solution of the present invention and does not constitute a limitation on the scope of protection.

[0022] I. Overall System Structure As attached Figure 1 As shown, the energy prediction system in this embodiment adopts a hierarchical collaborative architecture, which includes, from bottom to top, an edge sensing layer, a federated computing layer, a quantum optimization layer, and a spatiotemporal prediction layer.

[0023] Among them, a communication connection is established between the edge perception layer and the federated computing layer to upload the feature vector or quantum state representation after feature encoding; A communication connection is established between the federated computing layer and the quantum optimization layer to upload encrypted model gradient information; A communication connection is established between the quantum optimization layer and the federated computing layer to distribute updated global model parameters; The spatiotemporal prediction layer performs predictive analysis based on the global model parameters and outputs the prediction results and risk assessment results through the result output interface.

[0024] The aforementioned layers work together through data communication and interaction with model parameters, forming a multi-layer collaborative data processing and model update system.

[0025] II. Structure and Function of the Edge Sensing Layer The edge perception layer is deployed on the distributed energy node side to complete the collection, preprocessing and feature encoding of energy system operation data. It includes a data acquisition unit, a data cleaning unit and a feature encoding unit, which are connected in sequence.

[0026] (a) Data Acquisition Unit The data acquisition unit is used to collect operational data from distributed energy nodes. The collected data includes electrical parameter data, environmental parameter data, and energy output data. The electrical parameter data includes current, voltage, and power parameters; the environmental parameter data includes temperature, humidity, and irradiance parameters; and the energy output data characterizes the output power of photovoltaic or wind power equipment. All data types are recorded using a unified timestamp.

[0027] (ii) Data Cleaning Unit The data cleaning unit is used to preprocess and clean the collected raw data. The preprocessing and cleaning includes at least the removal of abnormal data, the completion or marking of missing data, and the time alignment of multi-source data to form a continuous data sequence that can be used for subsequent feature encoding.

[0028] (III) Feature Coding Unit The feature encoding unit is used to perform feature encoding processing on the cleaned multi-source data, mapping the original data into a unified latent space feature representation. After processing by the feature encoding unit, the output feature vector is represented as follows: in, This represents characteristic information related to the spatial structure of energy nodes. This indicates characteristic information about how energy operating status changes over time. It represents characteristic information related to environmental parameters.

[0029] For feature data involving the operating status of energy equipment, the edge sensing layer performs quantum state preprocessing on the data through the Pauli-Y rotating gate, and its quantum state representation is as follows: The processed feature data is sent to the federated computing layer via the communication connection.

[0030] III. Structure and Local Modeling Process of the Federated Computing Layer The federated computing layer is deployed on the edge server side to perform local spatiotemporal modeling on the feature data output by the edge perception layer and generate model update information for federated aggregation. The federated computing layer includes a spatial feature modeling unit, a temporal feature modeling unit, and a gradient generation and privacy protection unit.

[0031] (a) Spatial Feature Modeling Unit The spatial feature modeling unit is used to model the spatial relationships between energy nodes in a distributed energy system. The topology of the energy system is represented in the form of an adjacency matrix (A), which is constructed from the electrical connection relationships or equivalent impedance relationships between energy nodes, and the matrix elements are used to characterize the adjacency relationships between nodes.

[0032] The spatial feature modeling unit uses a graph convolutional network to aggregate features of the node neighborhood structure described by the adjacency matrix, thereby obtaining a spatial feature representation that reflects the spatial correlation characteristics of energy nodes.

[0033] (ii) Time Feature Modeling Unit The time-series modeling unit is used to model the time-series characteristics of energy system operation data. In this embodiment, the time-series modeling unit adopts a quantum recurrent neural network structure, and its state update relationship is as follows: in, This represents the quantum state at the previous time step. The quantum state representation of the current time step. For parameterized quantum gate circuits, It is a set of trainable parameters.

[0034] The parameterized quantum gate circuit includes RX gate, RY gate and CZ gate, wherein the RX gate and RY gate are used to rotate and modulate the quantum state to characterize the time series state changes, and the CZ gate is used to introduce entanglement between quantum states to characterize the correlation between states at different time steps.

[0035] (III) Gradient Generation and Privacy Protection Unit After completing the local spatiotemporal modeling, the gradient generation and privacy protection unit generate the model gradients. ,in Indicates the first Model parameters corresponding to each federated computing node.

[0036] Before uploading the gradient, a noise term that follows a Laplace distribution is introduced into the gradient: in, This represents the sensitivity of the gradient function. This is a privacy budget parameter used to control the intensity of noise injection and the degree of privacy protection.

[0037] Subsequently, the Paillier homomorphic encryption algorithm was used to encrypt the gradient after adding noise, resulting in: The encrypted gradient is sent to the quantum optimization layer via a communication connection.

[0038] IV. Structure of Quantum Optimization Layer and Federation Aggregation Process The quantum optimization layer is deployed on the federated server side and includes a gradient decryption unit, a quantum particle encoding unit, a federated aggregation and parameter optimization unit, and a model generation and distribution unit.

[0039] (I) Quantum Particle Encoding and Weight Representation The quantum particle encoding unit maps the aggregate weights of the federated clients to quantum state representations: in, and Let be the probability amplitude parameter, and satisfy... .

[0040] (II) ASL-QPSO Aggregation and Parameter Update The federated aggregation and parameter optimization unit uses the ASL-QPSO algorithm to update the aggregation weights, and the update relationship is as follows: in, This represents the weight value corresponding to the particle's historical best position. This represents the weight value corresponding to the global optimal position of the particle swarm. This represents a random function.

[0041] Inertia weight Nonlinear attenuation method is adopted: And the Levy flight disturbance mechanism is introduced: in, This indicates an operation performed according to the corresponding dimension.

[0042] The federated aggregation process is triggered when the following conditions are met: Upon triggering, the quantum optimization layer aggregates and updates the parameters of the received gradient information and generates updated global model parameters.

[0043] V. Spatiotemporal Prediction Layer and Result Output The spatiotemporal prediction layer includes a feature fusion unit, a prediction modeling unit, and a risk assessment unit. Multi-source features are concatenated to form the layer. The prediction model outputs the operational status prediction results, and the risk assessment unit calculates the risk probability: in, This represents the spatial feature representation of the output of a graph convolutional network.

[0044] The prediction results and risk assessment results are output through the results output interface.

[0045] VI. System Operation Process Description As attached Figure 2 As shown, during the operation of the energy prediction system provided in this embodiment, each functional layer executes collaboratively according to the predetermined data flow and model parameter flow. The overall process includes stages such as data acquisition, local modeling, federated updates, predictive analysis, and cyclic execution.

[0046] After the system starts up, the edge perception layer collects operational data from the distributed energy nodes. The collected data includes electrical parameter data, environmental parameter data, and energy output data of the energy equipment, and timestamps all types of data. After cleaning and processing, the collected data forms a continuous data sequence, and the feature encoding unit generates a feature representation of a unified dimension. The feature data is then uploaded to the federated computing layer.

[0047] After receiving the feature data, the federated computing layer models the spatial relationships between energy nodes based on the energy system topology, and further models the temporal evolution characteristics of energy operation data, thereby completing the training of the local spatiotemporal joint model and generating corresponding model update information.

[0048] After completing local model training, the federated computing layer performs privacy protection processing on the model update information and encrypts it before uploading the encrypted model update information to the quantum optimization layer.

[0049] After receiving model update information from multiple federated computing nodes, the quantum optimization layer triggers the federated aggregation process according to preset conditions, and updates the federated aggregation weights based on the quantum particle swarm optimization mechanism to generate updated global model parameters.

[0050] The global model parameters are distributed to the federated computing layer to guide the next round of local model training, and simultaneously to the spatiotemporal prediction layer for predicting and analyzing the operating status of the energy system within future time windows. After obtaining the prediction results, the spatiotemporal prediction layer performs a risk assessment and outputs the corresponding prediction analysis results and risk assessment information.

[0051] After completing one predictive analysis and outputting the results, the system enters the next round of operation, repeating the above steps to achieve continuous updates of model parameters and prediction results.

[0052] For those skilled in the art, any changes, modifications, substitutions, and variations made to the embodiments without departing from the principles and spirit of the present invention, based on the teachings of the present invention, still fall within the protection scope of the present invention.

Claims

1. An energy prediction system based on quantum particle swarm optimization and spatiotemporal coupling, characterized in that, It includes an edge sensing layer, a federated computing layer, a quantum optimization layer, and a spatiotemporal prediction layer; The edge sensing layer is used to collect the operating data of distributed energy nodes, clean the operating data and generate feature representations. The federated computing layer is used to receive the feature representation, model the spatial correlation between energy nodes, and model the temporal evolution characteristics of energy operation data to form a local spatiotemporal joint model and generate model update information. The quantum optimization layer is used to receive model update information from multiple federated computing layers, trigger the federated aggregation process when preset conditions are met, and update the federated aggregation weights based on the quantum particle swarm optimization mechanism to generate global model parameters. The spatiotemporal prediction layer is used to predict and analyze the operating status of the energy system within a future time window based on the global model parameters, and output the prediction results and corresponding risk assessment information.

2. The energy forecasting system according to claim 1, characterized in that, The edge sensing layer includes a data acquisition unit, a data cleaning unit, and a feature encoding unit; The data acquisition unit is used to collect electrical parameter data, environmental parameter data, and energy output data of the energy equipment. The data cleaning unit is used to perform abnormal data removal, missing data filling or marking, and time alignment processing on the collected data. The feature encoding unit is used to map the cleaned data into a feature representation of a uniform dimension.

3. The energy forecasting system according to claim 1, characterized in that, The federated computing layer includes a spatial feature modeling unit, a temporal feature modeling unit, and a gradient generation and privacy protection unit. The spatial feature modeling unit models the spatial relationships between energy nodes based on the topology of the energy system. The time feature modeling unit models the time series characteristics of energy operation data based on the spatial feature modeling results. The gradient generation and privacy protection unit is used to generate model update information and encrypt the model update information after performing privacy protection processing.

4. The energy forecasting system according to claim 3, characterized in that, Before encrypting the model update information, the gradient generation and privacy protection unit performs noise perturbation processing on the model update information to reduce the risk of privacy leakage in the model update information.

5. The energy forecasting system according to claim 1, characterized in that, The quantum optimization layer includes a gradient decryption unit, a quantum particle encoding unit, a federated aggregation and parameter optimization unit, and a model generation and distribution unit. The gradient decryption unit is used to decrypt the encrypted model update information. The quantum particle encoding unit is used to map federated aggregation weights into quantum particle form; The federated aggregation and parameter optimization unit is used to update the federated aggregation weights based on the quantum particle swarm optimization mechanism; The model generation and distribution unit is used to generate global model parameters and distribute them.

6. The energy forecasting system according to claim 1, characterized in that, The quantum optimization layer triggers the federated aggregation process when the model update information meets the preset threshold conditions, and performs federated aggregation and parameter update operations after triggering.

7. The energy forecasting system according to claim 1, characterized in that, The spatiotemporal prediction layer includes a feature fusion unit, a prediction modeling unit, and a risk assessment unit; The feature fusion unit is used to fuse spatial features, temporal features, and environmental features. The predictive modeling unit is used to predict the operating status of the energy system based on fused features and global model parameters; The risk assessment unit is used to assess the risk of the prediction results and output risk assessment information.

8. The energy forecasting system according to claim 1, characterized in that, The edge perception layer, federated computing layer, quantum optimization layer, and spatiotemporal prediction layer interact with each other via communication connections to exchange data and model parameters.

9. The energy forecasting system according to claim 7, characterized in that, The feature fusion unit combines spatial, temporal, and environmental features to form a fused feature representation for prediction modeling.

10. The energy forecasting system according to claim 1, characterized in that, After completing one predictive analysis and outputting the results, the system enters the next round of operation, repeating the process of data collection, local modeling, federated aggregation, model updating, and predictive analysis to achieve continuous updates of model parameters and prediction results.

Images