Transformer area operation state three-color regulation and control method and system based on multi-source data fusion

Abstract

The invention belongs to the technical field of power system automation and new energy, and relates to a transformer area operation state three-color regulation and control method and system based on multi-source data fusion, and the method comprises the following steps: S1, carrying out the collection and fusion of multi-source data; step S2, a step of collaborative prediction; s3, a dynamic regulation and control decision is generated; s4, an instruction issuing and executing step; according to the technical scheme of the invention, for the problems of low photovoltaic output prediction precision, lack of coordination of a transformer area load prediction model and stiffness of a regulation and control mechanism, dynamic balance of the source and the load is realized through multi-source data fusion and a three-color lamp hierarchical regulation and control mechanism, and the method is suitable for a power distribution network transformer area scene with high-permeability photovoltaic access.

Description

Technical Field

[0001] This invention belongs to the field of power system automation and new energy technology, specifically relating to a three-color control method and system for transformer substation operation status based on multi-source data fusion. Background Technology

[0002] Currently, high-penetration photovoltaic power stations face three core challenges in operation: (1) Low accuracy of photovoltaic power output prediction: Existing methods rely on historical data and basic weather forecasts, but fail to integrate multi-dimensional real-time parameters such as cloud movement and equipment temperature, resulting in prediction errors exceeding 15% and difficulty in depicting rapid fluctuations in power output; (2) Lack of coordination in the load forecasting model of the transformer area: Traditional methods do not distinguish between rigid loads and flexible loads, nor do they fully consider the impact of high-penetration photovoltaic access on load characteristics (especially peak-valley difference). Load forecasting and photovoltaic output forecasting are independent of each other and lack coordinated analysis, making it difficult to support refined regulation.

[0003] (3) Rigid control mechanism: The existing three-color lamp control in the transformer area relies on fixed manual experience thresholds, which cannot be dynamically and adaptively adjusted according to real-time photovoltaic output, load demand and forecast data. The control command delay is high (more than 200 milliseconds), the response is slow, and it is difficult to effectively cope with rapid changes such as sudden changes in photovoltaic output, and it cannot achieve dynamic balance between source and load.

[0004] In view of this, it is very necessary to provide a three-color control method and system for the operating status of a transformer substation based on multi-source data fusion to solve the above-mentioned defects in the prior art. Summary of the Invention

[0005] The purpose of this invention is to address the problems of low photovoltaic output prediction accuracy, lack of coordination in transformer area load prediction models, and rigid control mechanisms in the existing technologies mentioned above, by providing a three-color control method and system for transformer area operation status based on multi-source data fusion to solve the above-mentioned technical problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A three-color control method for the operating status of transformer substations based on multi-source data fusion includes the following steps: Step S1, the multi-source data acquisition and fusion step, in which: Data on photovoltaic output, energy storage status, load, environment, and photovoltaic equipment temperature are collected. A federated learning algorithm is used to fuse the collected photovoltaic output, energy storage status, and load data from multiple sources to generate a database of operating characteristics for the distribution area. Step S2, the collaborative prediction step, in which: Based on the data in the transformer area operation characteristic database, a hybrid model based on LSTM-Transformer is used to predict the photovoltaic power output trend and load trend; Step S3, the step of dynamically adjusting and generating decisions, in which: Based on the results of photovoltaic power output trend prediction and load trend prediction, the current operating status of the transformer substation is calculated and determined, and corresponding control strategies are generated; the boundary thresholds of each operating status of the transformer substation are dynamically adjusted using a reinforcement learning model. Step S4, the instruction issuance and execution step, in which: According to the control strategy, control commands are issued to the intelligent circuit breakers and flexible load controllers. The intelligent circuit breakers and flexible load controllers make corresponding adjustments according to the control commands to achieve power regulation and load control with millisecond-level response, so as to maintain the dynamic balance of source and load in the distribution area.

[0007] Preferably, step S1 specifically includes: Deploy an integrated power resource aggregation and control cabinet to collect photovoltaic output data, energy storage status data including energy storage SOC (State of Charge) data, load data, environmental data including temperature and humidity data, photovoltaic equipment temperature data, etc., with a sampling frequency of not less than 10Hz; The millimeter-wave radar integrated into the power resource aggregation and control cabinet monitors cloud movement and, in conjunction with satellite remote sensing data, corrects the irradiance prediction to obtain meteorological data for photovoltaic power output prediction. Using a federated learning framework, while protecting data privacy, the system aggregates photovoltaic output data, energy storage SOC (state of charge) data, and load data collected by the distributed power resource aggregation and control cabinet, and constructs a distribution area operation feature library that includes the distribution transformer load rate calculated from the load data.

[0008] Preferably, step S2 specifically includes: A hybrid model based on LSTM-Transformer is deployed at the edge computing node of the distribution area (i.e., the integrated power resource aggregation and control cabinet) to achieve coordinated prediction of photovoltaic output and load: Meteorological data, environmental data, historical photovoltaic power output data, and photovoltaic equipment temperature data are input into a hybrid model based on LSTM-Transformer, which outputs a photovoltaic power output prediction value for the next 15 minutes. A hybrid model based on LSTM-Transformer is used to classify and predict loads, resulting in load classification predictions that distinguish between rigid and flexible loads. Based on the load classification prediction results and combined with real-time electricity price signals, a demand-response-oriented load regulation scheme is generated for the control of three-color lights.

[0009] Preferably, step S3 specifically includes: The operating status of the transformer area includes: red light status, yellow light status, and green light status; wherein, the red light status corresponds to the high penetration risk operating status, the yellow light status corresponds to the medium risk operating status, and the green light status corresponds to the low risk operating status. If the operating status of the transformer area is red, the following control strategy is generated: initiate the forced charging operation of the energy storage and reduce the flexible load. If the operating status of the transformer area is yellow, the following control strategy is generated: the model predictive control method (MPC) is used to perform rolling optimization of the optimal charging and discharging power of the energy storage equipment to minimize the operating cost. If the operating status of the transformer area is green, the control strategy is generated as follows: no active intervention control commands are issued, the photovoltaic system is kept generating electricity at its maximum capacity, and the photovoltaic system operation data is continuously monitored. A reinforcement learning model (such as the PPO algorithm) is used to dynamically optimize the boundary thresholds for red, yellow, and green light states based on historical control effects. The objective function of the reinforcement learning model is to maximize the long-term cumulative reward. :

[0010] in, In the first Prediction accuracy at the next iteration In the first Power fluctuation rate at the next iteration In the first The cost of adjustment during the next iteration and These are weighting coefficients used to balance volatility and control costs. To enhance the discount factor of the learning model, The total number of iterations performed for the reinforcement learning model.

[0011] Preferably, step S4 specifically includes: The control strategies corresponding to each operating state of the transformer area are converted into specific control commands for smart circuit breakers and flexible load controllers, including forced charging commands for energy storage, flexible load reduction commands, dynamic charging and discharging commands for energy storage, load optimization and adjustment commands, maximum output commands for photovoltaics, passive adjustment commands, and monitoring commands. After the adjustment is implemented, the integrated power resource aggregation and control cabinet collects and uploads feedback data in real time, forming a closed-loop control loop from decision-making, issuance, execution to feedback, in order to train the reinforcement learning model and maintain the dynamic balance of power source and load in the distribution area.

[0012] Furthermore, this invention also provides a three-color control system for the operating status of transformer substations based on multi-source data fusion, comprising: A module for multi-source data acquisition and fusion, in which: In the data acquisition layer, photovoltaic power output data, energy storage status data, load data, environmental data, and photovoltaic equipment temperature data are collected. The collected photovoltaic power output data, energy storage status data, and load data are fused from multiple sources using a federated learning algorithm to generate a distribution area operation feature library. The collaborative prediction module contains: In the edge computing layer, based on the data in the transformer area operation feature library, a hybrid model based on LSTM-Transformer is used to predict the photovoltaic output trend and load trend; The module for generating dynamic control decisions includes: In the cloud-based decision-making layer, based on the results of photovoltaic output trend forecasting and load trend forecasting, the current operating status of the transformer substation is calculated and judged, and corresponding control strategies are generated; the boundary thresholds of each operating status of the transformer substation are dynamically adjusted using a reinforcement learning model. The module that issues and executes instructions includes: In the control execution layer, control commands are issued to smart circuit breakers and flexible load controllers according to the control strategy. The smart circuit breakers and flexible load controllers make corresponding adjustments according to the control commands to achieve power regulation and load control with millisecond-level response, so as to maintain the dynamic balance of source and load in the distribution area.

[0013] Preferably, the multi-source data acquisition and fusion module specifically includes: Deploy an integrated power resource aggregation and control cabinet to collect photovoltaic output data, energy storage status data including energy storage SOC (State of Charge) data, load data, environmental data including temperature and humidity data, photovoltaic equipment temperature data, etc., with a sampling frequency of not less than 10Hz; The millimeter-wave radar integrated into the power resource aggregation and control cabinet monitors cloud movement and, in conjunction with satellite remote sensing data, corrects the irradiance prediction to obtain meteorological data for photovoltaic power output prediction. Using a federated learning framework, while protecting data privacy, the system aggregates photovoltaic output data, energy storage SOC (state of charge) data, and load data collected by the distributed power resource aggregation and control cabinet, and constructs a distribution area operation feature library that includes the distribution transformer load rate calculated from the load data.

[0014] Preferably, the collaborative prediction module specifically includes: A hybrid model based on LSTM-Transformer is deployed at the edge computing node of the distribution area (i.e., the integrated power resource aggregation and control cabinet) to achieve coordinated prediction of photovoltaic output and load: Meteorological data, environmental data, historical photovoltaic power output data, and photovoltaic equipment temperature data are input into a hybrid model based on LSTM-Transformer, which outputs a photovoltaic power output prediction value for the next 15 minutes. A hybrid model based on LSTM-Transformer is used to classify and predict loads, resulting in load classification predictions that distinguish between rigid and flexible loads. Based on the load classification prediction results and combined with real-time electricity price signals, a demand-response-oriented load regulation scheme is generated for the control of three-color lights.

[0015] Preferably, the module for generating dynamic control decisions specifically includes: The operating status of the transformer area includes: red light status, yellow light status, and green light status; wherein, the red light status corresponds to the high penetration risk operating status, the yellow light status corresponds to the medium risk operating status, and the green light status corresponds to the low risk operating status. If the operating status of the transformer area is red, the following control strategy is generated: initiate the forced charging operation of the energy storage and reduce the flexible load. If the operating status of the transformer area is yellow, the following control strategy is generated: the model predictive control method (MPC) is used to perform rolling optimization of the optimal charging and discharging power of the energy storage equipment to minimize the operating cost. If the operating status of the transformer area is green, the control strategy is generated as follows: no active intervention control commands are issued, the photovoltaic system is kept generating electricity at its maximum capacity, and the photovoltaic system operation data is continuously monitored. A reinforcement learning model (such as the PPO algorithm) is used to dynamically optimize the boundary thresholds for red, yellow, and green light states based on historical control effects. The objective function of the reinforcement learning model is to maximize the long-term cumulative reward. :

[0016] in, In the first Prediction accuracy at the next iteration In the first Power fluctuation rate at the next iteration In the first The cost of adjustment during the next iteration and These are weighting coefficients used to balance volatility and control costs. To enhance the discount factor of the learning model, The total number of iterations performed for the reinforcement learning model.

[0017] Preferably, the module for issuing and executing instructions specifically includes: The control strategies corresponding to each operating state of the transformer area are converted into specific control commands for smart circuit breakers and flexible load controllers, including forced charging commands for energy storage, flexible load reduction commands, dynamic charging and discharging commands for energy storage, load optimization and adjustment commands, maximum output commands for photovoltaics, passive adjustment commands, and monitoring commands. After the adjustment is implemented, the integrated power resource aggregation and control cabinet collects and uploads feedback data in real time, forming a closed-loop control loop from decision-making, issuance, execution to feedback, in order to train the reinforcement learning model and maintain the dynamic balance of power source and load in the distribution area.

[0018] The beneficial effects of this invention are that it enables coordinated and refined forecasting of photovoltaic power and load: by using multi-source data fusion and LSTM-Transformer hybrid model, it achieves minute-level photovoltaic power output forecasting and hour-level load forecasting through real-time fusion and feature extraction of high-dimensional data (meteorology, photovoltaic, load, energy storage). This not only improves the forecasting accuracy, making the photovoltaic power output forecasting error ≤8% (50% improvement over traditional methods) and the load classification forecasting error ≤5% (30% improvement over independent forecasting models), but also enables coordinated forecasting of photovoltaic power output and load, and combines the influence of photovoltaic penetration rate and electricity price signals for coordinated modeling. Furthermore, this invention can significantly optimize control response speed and reduce latency: based on a decision architecture of edge computing and cloud collaboration, combined with reinforcement learning to dynamically optimize thresholds, the latency of issuing three-color light control commands is ≤50ms, which is more than 95% higher than existing technologies, achieving millisecond-level precise control; it can also enhance operational economy: through model predictive control and dynamic threshold adjustment, the energy storage charging and discharging strategy and flexible load scheduling are optimized, improving the energy storage charging and discharging efficiency by 20% and increasing the system's annual comprehensive revenue by 15%; the system has strong scalability: adopting a federated learning framework and standardized communication interface, it supports secure aggregation and collaborative control of data from multiple distribution areas, and can flexibly adapt to complex distribution network scenarios with high penetration and large-scale distributed photovoltaic access.

[0019] Furthermore, the design principle of this invention is reliable, the structure is simple, and it has a very wide range of application prospects.

[0020] Therefore, it is evident that the present invention has outstanding substantive features and significant progress compared with the prior art, and the beneficial effects of its implementation are also obvious. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0022] Figure 1This is a flowchart of a three-color control method for the operating status of a transformer substation based on multi-source data fusion, provided by the present invention.

[0023] Figure 2 This is a schematic diagram of a three-color control system for the operating status of a transformer substation based on multi-source data fusion, provided by the present invention.

[0024] Among them, 1-multi-source data acquisition and fusion module, 2-collaborative prediction module, 3-dynamic control decision generation module, and 4-instruction issuance and execution module. Detailed Implementation

[0025] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The following embodiments are explanations of the present invention, but the present invention is not limited to the following implementation methods.

[0026] Example 1: like Figure 1 As shown, this embodiment uses a digital twin platform for simulation to provide an operation and control process for a three-color control method for transformer area operation status based on multi-source data fusion when the transformer area operation status is in the red light state, including the following steps: Step S1, the multi-source data acquisition and fusion step, in which: Data on photovoltaic output, energy storage status, load, environment, and photovoltaic equipment temperature are collected. A federated learning algorithm is used to fuse the collected photovoltaic output, energy storage status, and load data from multiple sources to generate a database of operating characteristics for the distribution area. Step S1 specifically includes: Deploy an integrated power resource aggregation and control cabinet to collect data such as photovoltaic output value of 100kW, energy storage SOC value of 30%, load data, environmental data including temperature and humidity data, and photovoltaic equipment temperature data through the integrated power resource aggregation and control cabinet, with a sampling frequency of 10Hz. The millimeter-wave radar integrated into the power resource aggregation and control cabinet monitors cloud movement and, in conjunction with satellite remote sensing data, corrects the irradiance prediction to obtain meteorological data for photovoltaic power output prediction. Using a federated learning framework, while protecting data privacy, the system aggregates photovoltaic output data, energy storage SOC (State of Charge) data, and load data collected by the distributed power resource aggregation and control cabinet, and constructs a distribution area operation feature library including the distribution transformer load rate calculated from the load data, with a distribution transformer load rate value of 85%.

[0027] Step S2, the collaborative prediction step, in which: Based on the data in the transformer area operation characteristic database, a hybrid model based on LSTM-Transformer is used to predict the photovoltaic power output trend and load trend; Step S2 specifically includes: A hybrid model based on LSTM-Transformer is deployed at the edge computing node of the distribution area (i.e., the integrated power resource aggregation and control cabinet) to achieve coordinated prediction of photovoltaic output and load: Meteorological data, environmental data, historical photovoltaic power output data, and photovoltaic equipment temperature data are input into a hybrid model based on LSTM-Transformer, which outputs a photovoltaic power output prediction value for the next 15 minutes. A hybrid model based on LSTM-Transformer is used to classify and predict loads, resulting in load classification predictions that distinguish between rigid and flexible loads. Based on the load classification prediction results and combined with real-time electricity price signals, a demand-response-oriented load regulation scheme is generated for the control of three-color lights.

[0028] Step S3, the step of dynamically adjusting and generating decisions, in which: Based on the results of photovoltaic power output trend prediction and load trend prediction, the distribution transformer load rate is >80% and the photovoltaic power output mutation rate is >10% / min, the current operating status of the transformer area is determined to be red, and the corresponding control strategy is generated; the boundary thresholds of each operating status of the transformer area are dynamically adjusted using a reinforcement learning model. Step S3 specifically includes: When the front-end area is in a red light state, a control strategy is generated: initiate forced charging of energy storage and reduce flexible loads. A reinforcement learning model (such as the PPO algorithm) is used to dynamically optimize the boundary thresholds for red, yellow, and green light states based on historical control effects. The objective function of the reinforcement learning model is to maximize the long-term cumulative reward. :

[0029] in, In the first Prediction accuracy at the next iteration In the first Power fluctuation rate at the next iteration In the first The cost of adjustment during the next iteration and These are weighting coefficients used to balance volatility and control costs. To enhance the discount factor of the learning model, The total number of iterations performed for the reinforcement learning model.

[0030] Step S4, the instruction issuance and execution step, in which: According to the control strategy, control commands are issued to the intelligent circuit breakers and flexible load controllers. The intelligent circuit breakers and flexible load controllers make corresponding adjustments according to the control commands to achieve power regulation and load control with millisecond-level response, so as to maintain the dynamic balance of source and load in the distribution area.

[0031] Step S4 specifically includes: The control strategies corresponding to each operating state of the transformer area are converted into specific control commands for smart circuit breakers and flexible load controllers, including forced charging commands for energy storage and flexible load reduction commands. After the adjustment is implemented, the integrated power resource aggregation and control cabinet collects and uploads feedback data in real time, forming a closed-loop control loop from decision-making, issuance, execution to feedback, in order to train the reinforcement learning model and maintain the dynamic balance of power source and load in the distribution area.

[0032] Example 2: like Figure 2 As shown, this embodiment uses a digital twin platform for simulation to provide an operation and control system for a transformer substation operating status based on multi-source data fusion when the substation operating status is in the yellow light state, including: Module 1 for multi-source data acquisition and fusion, in which: In the data acquisition layer, photovoltaic power output data, energy storage status data, load data, environmental data, and photovoltaic equipment temperature data are collected. The collected photovoltaic power output data, energy storage status data, and load data are fused from multiple sources using a federated learning algorithm to generate a distribution area operation feature library. The multi-source data acquisition and fusion module 1 specifically includes: Deploy an integrated power resource aggregation and control cabinet to collect photovoltaic output data, energy storage status data including energy storage SOC (State of Charge) data, load data, environmental data including temperature and humidity data, photovoltaic equipment temperature data, etc., with a sampling frequency of not less than 10Hz; The millimeter-wave radar integrated into the power resource aggregation and control cabinet monitors cloud movement and, in conjunction with satellite remote sensing data, corrects the irradiance prediction to obtain meteorological data for photovoltaic power output prediction. Using a federated learning framework, while protecting data privacy, the system aggregates photovoltaic output data, energy storage SOC (state of charge) data, and load data collected by the distributed power resource aggregation and control cabinet, and constructs a distribution area operation feature library that includes the distribution transformer load rate calculated from the load data.

[0033] Module 2 of the collaborative prediction module, in which: In the edge computing layer, based on the data in the transformer area operation feature library, a hybrid model based on LSTM-Transformer is used to predict the photovoltaic output trend and load trend; The collaborative prediction module 2 specifically includes: A hybrid model based on LSTM-Transformer is deployed at the edge computing node of the distribution area (i.e., the integrated power resource aggregation and control cabinet) to achieve coordinated prediction of photovoltaic output and load: Meteorological data, environmental data, historical photovoltaic power output data, and photovoltaic equipment temperature data are input into a hybrid model based on LSTM-Transformer, which outputs a photovoltaic power output prediction value for the next 15 minutes, predicting a peak photovoltaic power output of 80kW. The load is classified and predicted using a hybrid model based on LSTM-Transformer, resulting in load classification predictions that distinguish between rigid and flexible loads. Based on the load classification prediction results and combined with real-time electricity price signals, a load regulation scheme oriented towards demand response is generated, predicting a peak load of 60kW for use in the regulation of three-color lights.

[0034] Module 3, which generates dynamic control decisions, includes: In the cloud-based decision-making layer, based on the results of photovoltaic output trend forecasting and load trend forecasting, the current operating status of the transformer area is calculated and determined to be in the yellow light state, and corresponding control strategies are generated; the boundary thresholds of each operating status of the transformer area are dynamically adjusted using a reinforcement learning model. The module 3 for generating dynamic control decisions specifically includes: When the front-end area is in yellow light status, the following control strategy is generated: the model predictive control method (MPC) is used to optimize the energy storage charging and discharging plan every 5 minutes to ensure that the difference between photovoltaic output and load is ≤10kW and minimize operating costs. A reinforcement learning model (such as the PPO algorithm) is used to dynamically optimize the boundary thresholds for red, yellow, and green light states based on historical control effects. The objective function of the reinforcement learning model is to maximize the long-term cumulative reward. :

[0035] in, In the first Prediction accuracy at the next iteration In the first Power fluctuation rate at the next iteration In the first The cost of adjustment during the next iteration and These are weighting coefficients used to balance volatility and control costs. To enhance the discount factor of the learning model, The total number of iterations performed for the reinforcement learning model.

[0036] Module 4, which handles instruction issuance and execution, contains: In the control execution layer, control commands are issued to smart circuit breakers and flexible load controllers according to the control strategy. The smart circuit breakers and flexible load controllers make corresponding adjustments according to the control commands to achieve power regulation and load control with millisecond-level response, so as to maintain the dynamic balance of source and load in the distribution area.

[0037] The module 4 for issuing and executing instructions specifically includes: The control strategies corresponding to each operating state of the transformer area are converted into specific control instructions for smart circuit breakers and flexible load controllers, including dynamic charging and discharging instructions for energy storage and load optimization adjustment instructions. After the adjustment is executed according to the control instructions, the integrated power resource aggregation and control cabinet collects and uploads feedback data in real time, forming a closed-loop control loop from decision-making, issuance, execution to feedback, in order to train the reinforcement learning model and maintain the dynamic balance of power source and load in the distribution area.

[0038] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The methods disclosed in the embodiments are described simply because they correspond to the systems disclosed in the embodiments; relevant details can be found in the method section.

[0039] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0040] In the embodiments provided by this invention, it should be understood that the disclosed systems, methods, and approaches can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0041] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0042] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit.

[0043] Similarly, in the various embodiments of the present invention, each processing unit can be integrated into a functional module, or each processing unit can exist physically, or two or more processing units can be integrated into a functional module.

[0044] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0045] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0046] The above-disclosed embodiments are merely preferred embodiments of the present invention, but the present invention is not limited thereto. Any non-creative variations that can be conceived by those skilled in the art, as well as any improvements and modifications made without departing from the principles of the present invention, should fall within the protection scope of the present invention.

Claims

1. A three-color regulation method for the operation state of a transformer area based on multi-source data fusion, characterized in that, Comprise the following steps: Step S1, the step of multi-source data acquisition and fusion, in which: Collect photovoltaic output data, energy storage state data, load data, environmental data, photovoltaic equipment temperature data, and use a federated learning algorithm to perform multi-source data fusion on the collected photovoltaic output data, energy storage state data, and load data to generate a distribution area operation feature library; Step S2, the step of collaborative prediction, in which: According to the data in the distribution area operation feature library, a hybrid model based on LSTM-Transformer is used to predict the photovoltaic output trend and the load trend; Step S3, the step of generating dynamic regulation and control decisions, in which: Based on the results of photovoltaic output trend prediction and load trend prediction, the current distribution area operation state is calculated and judged, and the corresponding regulation and control strategy is generated; The boundary threshold of each operation state of the distribution area is dynamically adjusted using a reinforcement learning model; Step S4, the step of issuing and executing instructions, in which: According to the regulation and control strategy, regulation and control instructions are issued to the intelligent circuit breaker and the flexible load controller, and the intelligent circuit breaker and the flexible load controller adjust accordingly according to the regulation and control instructions.

2. The three-color regulation method for the operation state of a transformer area based on multi-source data fusion according to claim 1, characterized in that, The step S1 specifically comprises: Deploying an electric power resource aggregation and regulation integrated cabinet, collecting photovoltaic output data, energy storage state data including energy storage state of charge data, load data, environmental data including temperature data and humidity data, and photovoltaic equipment temperature data through the electric power resource aggregation and regulation integrated cabinet, and the sampling frequency is not less than 10Hz; And through the millimeter wave radar integrated by the electric power resource aggregation and regulation integrated cabinet, the cloud layer movement is monitored, and the irradiation prediction is corrected combined with satellite remote sensing data to obtain meteorological data for photovoltaic output prediction; Using a federated learning framework, the photovoltaic output data, energy storage state data, and load data collected by the distributed electric power resource aggregation and regulation integrated cabinet are aggregated to construct a distribution area operation feature library containing the distribution transformer load rate calculated from the load data.

3. The three-color regulation method for the operation state of a transformer area based on multi-source data fusion according to claim 1, characterized in that, The step S2 specifically comprises: Deploying a hybrid model based on LSTM-Transformer in the electric power resource aggregation and regulation integrated cabinet for collaborative prediction of photovoltaic output and load: Inputting meteorological data, environmental data, historical photovoltaic output data, and photovoltaic equipment temperature data into the hybrid model based on LSTM-Transformer to output photovoltaic output prediction values for the next 15 minutes; The hybrid model based on LSTM-Transformer classifies and predicts the load to obtain load classification prediction results that distinguish between rigid load and flexible load; Based on the load classification prediction results, combined with real-time electricity price signals, a load adjustment scheme for demand response is generated.

4. The three-color regulation method for the operation state of a transformer area based on multi-source data fusion according to claim 1, characterized in that, The step S3 specifically comprises: The distribution area operation state includes: red light state, yellow light state, and green light state; wherein the red light state corresponds to a high penetration risk operation state, the yellow light state corresponds to a medium risk operation state, and the green light state corresponds to a low risk operation state; If the distribution area operation state is the red light state, generate the regulation and control strategy: start the forced charging operation of the energy storage, and reduce the flexible load; If the operating state of the transformer area is yellow, a regulation strategy is generated: using model predictive control method, the optimal charging and discharging power of the energy storage device is optimized, and the operating cost is minimized; If the operating state of the transformer area is green, a regulation strategy is generated: no active intervention control command is issued, the photovoltaic system generates power according to the maximum capacity, and the photovoltaic system operating data is continuously monitored; The reinforcement learning model is used to dynamically optimize the boundary threshold of the red light state, the yellow light state and the green light state according to historical control effects, and a target function of the reinforcement learning model is to maximize long-term cumulative rewards : in, In the first Prediction accuracy at the next iteration In the first Power fluctuation rate at the next iteration In the first The cost of adjustment during the next iteration and These are weighting coefficients used to balance volatility and control costs. To enhance the discount factor of the learning model, The total number of iterations performed for the reinforcement learning model.

5. The three-color regulation method for the operation state of a transformer area based on multi-source data fusion according to claim 1, characterized in that, The step S4 specifically comprises: The regulation strategy corresponding to each operating state of the transformer area is converted into specific control instructions for the intelligent circuit breaker and the flexible load controller, including energy storage forced charging instruction, flexible load reduction instruction, energy storage dynamic charging and discharging instruction, load optimization adjustment instruction, photovoltaic maximum output instruction, passive adjustment instruction and monitoring instruction; After the adjustment is performed, the power resource aggregation regulation integrated cabinet collects feedback data in real time and uploads the feedback data.

6. A three-color regulation system for the operation state of a transformer area based on multi-source data fusion, characterized in that, It comprises: A multi-source data acquisition and fusion module, in which: In the data acquisition layer, photovoltaic output data, energy storage state data, load data, environmental data and photovoltaic equipment temperature data are collected, and a federal learning algorithm is used to fuse the collected photovoltaic output data, energy storage state data and load data to generate a transformer area operating feature library; A collaborative prediction module, in which: In the edge computing layer, a hybrid model based on LSTM-Transformer is used to predict the photovoltaic output trend and the load trend according to the data in the transformer area operating feature library; A dynamic regulation decision generation module, in which: In the cloud decision layer, based on the results of the photovoltaic output trend prediction and the load trend prediction, the current operating state of the transformer area is calculated and judged, and the corresponding regulation strategy is generated; The boundary threshold values of each operating state of the transformer area are dynamically adjusted by using a reinforcement learning model; An instruction issuing and executing module, in which: In the regulation execution layer, regulation instructions are issued to the intelligent circuit breaker and the flexible load controller according to the regulation strategy, and the intelligent circuit breaker and the flexible load controller adjust accordingly according to the regulation instructions.

7. The three-color regulation system for the operation state of a transformer area based on multi-source data fusion according to claim 6, characterized in that, The multi-source data acquisition and fusion module specifically comprises: A power resource aggregation regulation integrated cabinet is deployed, and the photovoltaic output, the energy storage state data including the energy storage state of charge data, the load data, the environmental data including the temperature data and the humidity data, and the photovoltaic equipment temperature data are collected by the power resource aggregation regulation integrated cabinet at a sampling frequency of not less than 10Hz; The cloud layer movement is monitored by the millimeter wave radar integrated in the power resource aggregation regulation integrated cabinet, and the irradiation prediction is corrected in combination with the satellite remote sensing data to obtain the meteorological data used for photovoltaic output prediction; Using a federal learning framework, the photovoltaic output data, the energy storage state of charge data and the load data collected by the distributed power resource aggregation regulation integrated cabinet are aggregated to construct a transformer area operating feature library including the transformer load rate calculated from the load data.

8. The three-color regulation system for the operation state of a transformer area based on multi-source data fusion according to claim 6, characterized in that, The collaborative prediction module specifically comprises: The hybrid model based on LSTM-Transformer is deployed on the power resource aggregation regulation integrated cabinet to realize collaborative prediction of photovoltaic output and load: The meteorological data, environmental data, historical photovoltaic output data and photovoltaic equipment temperature data are input into the hybrid model based on LSTM-Transformer, and the photovoltaic output prediction value of the next 15 minutes is output; The hybrid model based on LSTM-Transformer classifies and predicts the load to obtain the load classification prediction result distinguishing rigid load and flexible load; Based on the load classification prediction result, combined with the real-time electricity price signal, a load adjustment scheme for demand response is generated.

9. The three-color regulation system for the operation state of a transformer area based on multi-source data fusion according to claim 6, characterized in that, The module for generating dynamic regulation and control decisions specifically includes: The operating state of the transformer area includes: red light state, yellow light state, green light state; wherein the red light state corresponds to the operating state of high penetration risk, the yellow light state corresponds to the operating state of medium risk, and the green light state corresponds to the operating state of low risk; If the operating state of the transformer area is the red light state, the regulation and control strategy is generated: start the forced charging operation of the energy storage, and reduce the flexible load; If the operating state of the transformer area is the yellow light state, the regulation and control strategy is generated: use the model predictive control method to rollingly optimize the optimal charging and discharging power of the energy storage device, and minimize the operation cost; If the operating state of the transformer area is the green light state, the regulation and control strategy is generated: no active intervention control instruction is issued, the photovoltaic system generates power according to the maximum capacity, and the photovoltaic system operation data is continuously monitored; The reinforcement learning model is used to dynamically optimize the boundary threshold of the red light state, the yellow light state and the green light state according to historical regulation effects, and a target function of the reinforcement learning model is to maximize long-term cumulative rewards : in, In the first Prediction accuracy at the next iteration In the first Power fluctuation rate at the next iteration In the first The cost of adjustment during the next iteration and These are weighting coefficients used to balance volatility and control costs. To enhance the discount factor of the learning model, The total number of iterations performed for the reinforcement learning model.

10. The three-color regulation system for the operation state of a transformer area based on multi-source data fusion according to claim 6, characterized in that, The module for issuing and executing instructions specifically includes: The regulation and control strategies corresponding to each operating state of the transformer area are converted into specific control instructions for the intelligent circuit breaker and the flexible load controller, including the energy storage forced charging instruction, the flexible load reduction instruction, the energy storage dynamic charging and discharging instruction, the load optimization adjustment instruction, the photovoltaic maximum output instruction, the passive adjustment instruction and the monitoring instruction; After the adjustment is executed, the power resource aggregation regulation and control integrated cabinet collects feedback data in real time and uploads them.

Images