Virtual power plant source network load storage cooperative control method and system based on three-layer architecture

Abstract

The invention discloses a virtual power plant source network load storage cooperative control method and system based on a three-layer architecture. The method comprises the steps that a data fusion layer comprises the steps of collecting multi-source heterogeneous data of a power system, constructing a multi-source heterogeneous original data set, fusing the multi-source heterogeneous data and generating a fused symbolized tensor flow; the modeling and reasoning layer comprises the steps of inputting symbolized tensor flow into a probability prediction model to calculate new energy output prediction uncertainty and consumption capability uncertainty, establishing joint uncertainty distribution of the two, modeling source-network-load-storage data into a quaternary interaction tensor, and combining the joint uncertainty distribution to obtain the new energy output prediction uncertainty and the consumption capability uncertainty. Generating a candidate regulation and control strategy set, and selecting an optimal regulation and control strategy to regulate and control source-network-load-storage; the execution and feedback layer comprises the steps that the power system is continuously monitored, and when the change rate of the key monitoring index exceeds a change rate threshold value, the edge node generates an emergency control strategy and executes the emergency control strategy.

Description

Technical Field

[0001] This invention belongs to the field of power system collaborative control technology, and more specifically, relates to a method and system for collaborative control of virtual power plant source-grid-load-storage based on a three-layer architecture. Background Technology

[0002] In existing research on power system operation and renewable energy consumption, mainstream technologies generally focus on three directions: renewable energy output prediction, load and energy storage regulation modeling, and system-level dispatch optimization. Firstly, regarding renewable energy output prediction, current methods commonly employ time series analysis, machine learning, and deep learning models, such as Autoregressive Models (ARIMA), Long Short-Term Memory Networks (LSTM), Convolutional Neural Networks (CNN), and hybrid models. These methods can capture the temporal characteristics of environmental factors such as wind speed and irradiance to a certain extent, enabling prediction of future renewable energy power. However, these methods are mostly point-based predictions, insufficiently characterizing the uncertainty of the prediction results and failing to effectively provide probability distributions or confidence intervals. This results in a lack of reliable risk quantification for dispatching when facing drastic fluctuations in wind and solar resources. Secondly, in load forecasting and energy storage modeling, traditional methods often rely on regression models, time series decomposition methods, or the introduction of deep networks to fit historical load curves and environmental factors to predict future load levels. Energy storage system modeling, on the other hand, often uses deterministic battery state equations, calculating future energy storage capacity based on known charge / discharge efficiencies and power constraints. However, this type of modeling usually assumes a stable environment and ignores uncertainties such as battery life decay and temperature fluctuations. It is difficult to truly reflect the dynamic behavior and available capacity of energy storage in long-term operation, resulting in a discrepancy between the assessment of absorption potential and the actual operating results.

[0003] Therefore, there is an urgent need for a technical solution that can integrate multiple elements of source-grid-load-storage, characterize the distribution characteristics of uncertainty, and generate robust optimization strategies using intelligent algorithms based on joint modeling. Summary of the Invention

[0004] To address the above technical problems, this invention proposes a three-layer architecture-based method for coordinated control of virtual power plant generation, grid, load, and storage. The three-layer architecture comprises a data fusion layer, a modeling and inference layer, and an execution and feedback layer. The data fusion layer includes: collecting multi-source heterogeneous data from the power system, constructing a multi-source heterogeneous raw data set, fusing the multi-source heterogeneous data, and generating a fused symbolic tensor flow; The modeling and reasoning layer includes: inputting symbolic tensor flow into the probabilistic prediction model to calculate the uncertainty of new energy output prediction and the uncertainty of absorption capacity, and establishing the joint uncertainty distribution of the two; modeling the source-grid-load-storage data as a quaternary interaction tensor, and combining the joint uncertainty distribution to generate a set of candidate regulation strategies, and selecting the optimal regulation strategy to regulate the source-grid-load-storage. The execution and feedback layer includes: continuous monitoring of the power system; when the rate of change of key monitoring indicators exceeds the rate of change threshold, the edge nodes generate and execute emergency control strategies.

[0005] Furthermore, the fusion of multi-source heterogeneous data includes: fusing the multi-source heterogeneous data through a cross-temporal adaptive mapping matrix.

[0006] Furthermore, the cross-spatial adaptive mapping matrix includes: Suppose that there are multi-source heterogeneous data. One data source, at any time No. The original sequences of the data sources are , No. The sampling frequency of each data source is , No. The data credibility of each data source is The unified target time window length is The target sampling frequency is ,but , in, For the first The data source corresponds to the first one on the unified time grid. The observations at each time point are mapped to the weights of the symbolic tensor flow. The time decay factor, For the first The data source corresponds to the first one on the unified time grid. At that moment The observed values, For the first The mean of the data source, For the first Standard deviation of the data source To prevent division by zero of small constants.

[0007] Furthermore, the modeling and inference layer includes: Based on the symbolic tensor flow, a probabilistic prediction model is used to model the future output of new energy sources, resulting in a prediction distribution that includes the mean and variance, which characterizes the uncertainty of new energy output prediction. Based on the symbolic tensor flow, the future load level is modeled according to the probabilistic prediction model, and the future available energy storage capacity is modeled in combination with the energy storage status, resulting in a probability distribution containing mean and variance, which characterizes the uncertainty of absorption capacity. Based on the predicted distribution of new energy output, the future load distribution, and the future available energy storage capacity distribution, the joint uncertainty distribution of the uncertainty of new energy output prediction and the uncertainty of absorption capacity is obtained through the dependency modeling function. Based on the joint uncertainty distribution, the probability that the predicted renewable energy output is less than or equal to the sum of the predicted load and the predicted energy storage capacity is obtained, so as to determine the power system's capacity to absorb renewable energy output and output absorption risk warnings.

[0008] Furthermore, modeling the source-grid-load-storage data as a quaternary interaction tensor includes: modeling the source-grid-load-storage data as a quaternary interaction tensor through a structured attention mechanism.

[0009] Furthermore, after the edge node generates and executes the emergency control strategy, it updates the symbolic tensor flow and re-executes the modeling and inference layer and the execution and feedback layer.

[0010] This invention also proposes a three-layer architecture-based virtual power plant source-grid-load-storage coordinated control system, comprising: a data fusion module, a modeling and inference module, and an execution and feedback module, wherein, The data fusion module includes: collecting multi-source heterogeneous data from the power system, constructing a multi-source heterogeneous raw data set, fusing the multi-source heterogeneous data, and generating a fused symbolic tensor flow; The modeling and reasoning module includes: inputting symbolic tensor flow into the probabilistic prediction model to calculate the uncertainty of new energy output prediction and the uncertainty of absorption capacity, and establishing the joint uncertainty distribution of the two; modeling the source-grid-load-storage data as a quaternary interaction tensor, and combining the joint uncertainty distribution to generate a set of candidate regulation strategies, and selecting the optimal regulation strategy to regulate the source-grid-load-storage. The execution and feedback module includes: continuous monitoring of the power system; when the rate of change of key monitoring indicators exceeds the rate of change threshold, the edge node generates and executes an emergency control strategy.

[0011] Furthermore, the fusion of multi-source heterogeneous data includes: fusing the multi-source heterogeneous data through a cross-temporal adaptive mapping matrix.

[0012] Furthermore, the cross-spatial adaptive mapping matrix includes: Suppose that there are multi-source heterogeneous data. One data source, at any time No. The original sequences of the data sources are , No. The sampling frequency of each data source is , No. The data credibility of each data source is The unified target time window length is The target sampling frequency is ,but , in, For the first The data source corresponds to the first one on the unified time grid. The observations at each time point are mapped to the weights of the symbolic tensor flow. The time decay factor, For the first The data source corresponds to the first one on the unified time grid. At that moment The observed values, For the first The mean of the data source, For the first Standard deviation of the data source To prevent division by zero of small constants.

[0013] Furthermore, the modeling and inference layer includes: Based on the symbolic tensor flow, a probabilistic prediction model is used to model the future output of new energy sources, resulting in a prediction distribution that includes the mean and variance, which characterizes the uncertainty of new energy output prediction. Based on the symbolic tensor flow, the future load level is modeled according to the probabilistic prediction model, and the future available energy storage capacity is modeled in combination with the energy storage status, resulting in a probability distribution containing mean and variance, which characterizes the uncertainty of absorption capacity. Based on the predicted distribution of new energy output, the future load distribution, and the future available energy storage capacity distribution, the joint uncertainty distribution of the uncertainty of new energy output prediction and the uncertainty of absorption capacity is obtained through the dependency modeling function. Based on the joint uncertainty distribution, the probability that the predicted renewable energy output is less than or equal to the sum of the predicted load and the predicted energy storage capacity is obtained, so as to determine the power system's capacity to absorb renewable energy output and output absorption risk warnings.

[0014] In summary, the technical solutions conceived by this invention have the following beneficial effects compared with the prior art: The technical solution of this invention constructs a four-element interactive tensor integrating source-grid-load-storage and combines it with the joint distribution of prediction and absorption capacity to achieve collaborative modeling of new energy output prediction and system absorption capacity. Then, it uses an AI large model with an attention mechanism to perform strategy reasoning and optimization, and selects the optimal solution with the minimum cost from the set of candidate control strategies. This enables the accurate characterization of the dynamic coupling relationship between new energy power fluctuations and load and energy storage under uncertain conditions, significantly improving the flexibility and stability of the power system and achieving safe, economical and sustainable operation under high proportion of new energy access. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention; Figure 2 This is a system structure diagram of Embodiment 2 of the present invention. Detailed Implementation

[0016] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0017] The method provided by this invention can be implemented in a terminal environment that may include one or more of the following components: a processor, a storage medium, and a display screen. The storage medium stores at least one instruction, which is loaded and executed by the processor to implement the method described in the following embodiments.

[0018] A processor may include one or more processing cores. The processor uses various interfaces and lines to connect various parts of the terminal, and performs various functions and processes data by running or executing instructions, programs, code sets or instruction sets stored in the storage medium, and by calling data stored in the storage medium.

[0019] Storage media can include random access memory (RAM) or read-only memory (ROM). Storage media can be used to store instructions, programs, code, code sets, or instructions.

[0020] The display screen is used to show the user interface of each application.

[0021] In addition, those skilled in the art will understand that the above-described structure of the terminal does not constitute a limitation on the terminal. The terminal may include more or fewer components, or combine certain components, or have different component arrangements. For example, the terminal may also include radio frequency circuits, input units, sensors, audio circuits, power supplies, and other components, which will not be described in detail here.

[0022] Example 1 like Figure 1 As shown, this embodiment proposes a three-layer architecture-based virtual power plant source-grid-load-storage coordinated control method, comprising: a data fusion layer, a modeling and inference layer, and an execution and feedback layer, wherein... The data fusion layer includes: collecting multi-source heterogeneous data from the power system, constructing a multi-source heterogeneous raw data set, fusing the multi-source heterogeneous data, and generating a fused symbolic tensor flow; Preferably, the symbolic tensor stream contains multi-dimensional temporal feature tensors, which retain both continuous quantities such as power, frequency, SOC, wind speed, light intensity, and temperature, as well as discrete symbolic information such as alarms and state transitions.

[0023] Specifically, fusing multi-source heterogeneous data includes fusing the multi-source heterogeneous data through a cross-temporal adaptive mapping matrix.

[0024] Specifically, the cross-spatial adaptive mapping matrix includes: Suppose that there are multi-source heterogeneous data. One data source, at any time No. The original sequences of the data sources are , No. The sampling frequency of each data source is , No. The data credibility of each data source is The unified target time window length is The target sampling frequency is ,but , in, For the first The data source corresponds to the first one on the unified time grid. The observations at each time point are mapped to the weights of the symbolic tensor flow. The time decay factor, For the first The data source corresponds to the first one on the unified time grid. At that moment The observed values, For the first The mean of the data source, For the first Standard deviation of the data source To prevent division by zero of small constants.

[0025] Specifically, generating the fused symbolic tensor flow includes: , , in, To symbolize tensor flow, To standardize the number of steps in the time grid, For symbolic transformation functions, It is a one-hot vector encoding.

[0026] The modeling and reasoning layer includes: inputting symbolic tensor flow into the probabilistic prediction model to calculate the uncertainty of new energy output prediction and the uncertainty of absorption capacity, and establishing the joint uncertainty distribution of the two; modeling the source-grid-load-storage data as a quaternary interaction tensor, and combining the joint uncertainty distribution to generate a set of candidate regulation strategies, and selecting the optimal regulation strategy to regulate the source-grid-load-storage. Preferably, a set of candidate regulation strategies is generated by an AI large model based on the quaternary interaction tensor and the joint uncertainty distribution, and the optimal regulation strategy is selected (for example, the candidate regulation strategy with the lowest cost is selected as the optimal regulation strategy, and the cost may include indicators such as wind curtailment rate, load deviation or energy storage loss). The AI ​​large model can be a deep Q network (DQN) or a deep deterministic policy gradient (DDPG).

[0027] Specifically, the modeling and inference layer includes: Based on the symbolic tensor flow, a probabilistic prediction model is used to model the future output of new energy sources, resulting in a prediction distribution that includes the mean and variance, which characterizes the uncertainty of new energy output prediction. Preferably, the uncertainties in the forecasting of new energy output include: , in, Forecasting uncertainties in contributing to new energy sources For the future The output of new energy sources at that time Conditional probability distributions can be implemented using Bayesian neural networks (BNNs) or probabilistic transformers.

[0028] Preferably, input the historical output sequence of new energy sources. ( (For historical time intervals), input external environmental factors (wind speed, wind direction, irradiance, temperature, humidity, etc.), and generate symbolic tensor flows. Provides a predicted distribution using a probabilistic time series model (such as Transformer–VAE): , in, This represents the predicted average output of new energy sources. The variance of the predicted new energy output.

[0029] Based on the symbolic tensor flow, the future load level is modeled according to the probabilistic prediction model, and the future available energy storage capacity is modeled in combination with the energy storage status, resulting in a probability distribution containing mean and variance, which characterizes the uncertainty of absorption capacity. Preferably, the uncertainty of absorption capacity includes: , in, Due to the uncertainty of absorption capacity, For the future The load at that time, For the future Available energy storage capacity at that time.

[0030] Preferably, input historical load curves. Input weather, season, etc. (such as temperature, holidays, electricity type structure), and generate symbolic tensor flow. Provides the output distribution using a probabilistic time series model (such as Transformer–VAE): , in, The predicted average load, This represents the predicted load variance.

[0031] About the future Available energy storage capacity at that time The calculation is performed using the following formula: , in, For a moment Available energy storage capacity For charging efficiency, For charging power, To predict the step size, For discharge efficiency, This represents the discharge power.

[0032] By adding random perturbations (battery temperature, battery life degradation), the output distribution is obtained: , in, This represents the average predicted available energy storage capacity. This represents the variance of the predicted available energy storage capacity.

[0033] Based on the predicted distribution of new energy output, the future load distribution, and the future available energy storage capacity distribution, the joint uncertainty distribution of the uncertainty of new energy output prediction and the uncertainty of absorption capacity is obtained through the dependency modeling function (Copula function). Preferably, the predicted distribution of new energy output is as follows: Future load distribution is The future distribution of available energy storage capacity is as follows .

[0034] Based on the joint uncertainty distribution, the probability that the predicted renewable energy output is less than or equal to the sum of the predicted load and the predicted energy storage capacity is obtained, so as to determine the power system's capacity to absorb renewable energy output and output absorption risk warnings.

[0035] Preferably, if the probability that the predicted new energy output is less than or equal to the sum of the predicted load and the predicted energy storage capacity is less than or equal to 1 and greater than the preset absorption threshold, then the power system can absorb the new energy. If it is less than or equal to 1 and less than the preset absorption threshold, then an absorption risk warning is output (indicating that advance scheduling is needed to avoid wind and solar curtailment or load gap).

[0036] Specifically, modeling source-network-load-storage data as a quaternary interaction tensor includes: modeling source-network-load-storage data as a quaternary interaction tensor through a structured attention mechanism.

[0037] Preferably, regarding the structured attention mechanism, it is as follows: Let the input vector be: ,in, These are characteristics of the power grid (such as voltage amplitude and phase angle, line power flow, frequency deviation, etc.). The time step (prediction window length) is the time step. Each entity dimension is a feature dimension, and 4 is the entity dimension. ).

[0038] Time-series attention: Perform time-series attention on each entity dimension. , in, For the first Attention to time series data across entity dimensions.

[0039] Regarding obtaining , and Specifically: Such as a certain entity dimension (e.g.) The input vector is The three sets of learnable weight matrices (which are linear transformation parameters in the neural network layers, automatically updated during training via backpropagation to minimize the loss function) are mapped as follows: , in, This is the first weight matrix, used to weight the input vector. Mapped to the first Query vectors for each entity dimension , This is the second weight matrix, used to weight the input vector. Mapped to the first Key vectors of each entity dimension , This is the third weight matrix, used to weight the input vector. Mapped to the first Value vector of each entity dimension .

[0040] Entity-Dimensional Attention: Concatenate the attention sequences of each entity dimension's time series to generate... , in, For entity dimension Attention to time series, For entity dimension Attention to time series, For entity dimension Attention to time series, For entity dimension Attention to time series data.

[0041] right Perform attention between entities: , in, Attention is focused on the entity dimension.

[0042] Regarding obtaining , and Methods and acquisition , and The method is the same, so it will not be repeated here.

[0043] Combining time series attention and entity-level attention: , in, For a quaternion interaction tensor, This is the weight matrix. This is a bias term.

[0044] The execution and feedback layer includes: continuous monitoring of the power system; when the rate of change of key monitoring indicators (such as power, frequency, SOC, etc.) exceeds the rate of change threshold, the edge nodes generate and execute emergency control strategies.

[0045] Specifically, after the edge node generates and executes the emergency control strategy, it updates the symbolic tensor flow and re-executes the modeling and inference layer and the execution and feedback layer.

[0046] Example 2 like Figure 2 As shown, this embodiment proposes a three-layer architecture-based virtual power plant source-grid-load-storage coordinated control system, comprising: a data fusion module, a modeling and inference module, and an execution and feedback module, wherein... The data fusion module includes: collecting multi-source heterogeneous data from the power system, constructing a multi-source heterogeneous raw data set, fusing the multi-source heterogeneous data, and generating a fused symbolic tensor flow; Preferably, the symbolic tensor stream contains multi-dimensional temporal feature tensors, which retain both continuous quantities such as power, frequency, SOC, wind speed, light intensity, and temperature, as well as discrete symbolic information such as alarms and state transitions.

[0047] Specifically, fusing multi-source heterogeneous data includes fusing the multi-source heterogeneous data through a cross-temporal adaptive mapping matrix.

[0048] Specifically, the cross-spatial adaptive mapping matrix includes: Suppose that there are multi-source heterogeneous data. One data source, at any time No. The original sequences of the data sources are , No. The sampling frequency of each data source is , No. The data credibility of each data source is The unified target time window length is The target sampling frequency is ,but , in, For the first The data source corresponds to the first one on the unified time grid. The observations at each time point are mapped to the weights of the symbolic tensor flow. The time decay factor, For the first The data source corresponds to the first one on the unified time grid. At that moment The observed values, For the first The mean of the data source, For the first Standard deviation of the data source To prevent division by zero of small constants.

[0049] Specifically, generating the fused symbolic tensor flow includes: , , in, To symbolize tensor flow, To standardize the number of steps in the time grid, For symbolic transformation functions, It is a one-hot vector encoding.

[0050] The modeling and reasoning module includes: inputting symbolic tensor flow into the probabilistic prediction model to calculate the uncertainty of new energy output prediction and the uncertainty of absorption capacity, and establishing the joint uncertainty distribution of the two; modeling the source-grid-load-storage data as a quaternary interaction tensor, and combining the joint uncertainty distribution to generate a set of candidate regulation strategies, and selecting the optimal regulation strategy to regulate the source-grid-load-storage. Preferably, a set of candidate control strategies is generated by an AI large model based on the quaternary interaction tensor and the joint uncertainty distribution, and the optimal control strategy is selected. The AI ​​large model can be a deep Q-network (DQN) or a deep deterministic policy gradient (DDPG), etc.

[0051] Specifically, the modeling and inference layer includes: Based on the symbolic tensor flow, a probabilistic prediction model is used to model the future output of new energy sources, resulting in a prediction distribution that includes the mean and variance, which characterizes the uncertainty of new energy output prediction. Preferably, the uncertainties in the forecasting of new energy output include: , in, Forecasting uncertainties in contributing to new energy sources For the future The output of new energy sources at that time Conditional probability distributions can be implemented using Bayesian neural networks (BNNs) or probabilistic transformers.

[0052] Preferably, input the historical output sequence of new energy sources. ( (For historical time intervals), input external environmental factors (wind speed, wind direction, irradiance, temperature, humidity, etc.), and generate symbolic tensor flows. Provides a predicted distribution using a probabilistic time series model (such as Transformer–VAE): , in, This represents the predicted average output of new energy sources. The variance of the predicted new energy output.

[0053] Based on the symbolic tensor flow, the future load level is modeled according to the probabilistic prediction model, and the future available energy storage capacity is modeled in combination with the energy storage status, resulting in a probability distribution containing mean and variance, which characterizes the uncertainty of absorption capacity. Preferably, the uncertainty of absorption capacity includes: , in, Due to the uncertainty of absorption capacity, For the future The load at that time, For the future Available energy storage capacity at that time.

[0054] Preferably, input historical load curves. Input weather, season, etc. (such as temperature, holidays, electricity type structure), and generate symbolic tensor flow. Provides the output distribution using a probabilistic time series model (such as Transformer–VAE): , in, The predicted average load, This represents the predicted load variance.

[0055] About the future Available energy storage capacity at that time The calculation is performed using the following formula: , in, For a moment Available energy storage capacity For charging efficiency, For charging power, To predict the step size, For discharge efficiency, This represents the discharge power.

[0056] By adding random perturbations (battery temperature, battery life degradation), the output distribution is obtained: , in, This represents the average predicted available energy storage capacity. This represents the variance of the predicted available energy storage capacity.

[0057] Based on the predicted distribution of new energy output, the future load distribution, and the future available energy storage capacity distribution, the joint uncertainty distribution of the uncertainty of new energy output prediction and the uncertainty of absorption capacity is obtained through the dependency modeling function (Copula function). Preferably, the predicted distribution of new energy output is as follows: Future load distribution is The future distribution of available energy storage capacity is as follows .

[0058] Based on the joint uncertainty distribution, the probability that the predicted renewable energy output is less than or equal to the sum of the predicted load and the predicted energy storage capacity is obtained, so as to determine the power system's capacity to absorb renewable energy output and output absorption risk warnings.

[0059] Preferably, if the probability that the predicted new energy output is less than or equal to the sum of the predicted load and the predicted energy storage capacity is less than or equal to 1 and greater than the preset absorption threshold, then the power system can absorb the new energy. If it is less than or equal to 1 and less than the preset absorption threshold, then an absorption risk warning is output (indicating that advance scheduling is needed to avoid wind and solar curtailment or load gap).

[0060] Specifically, modeling source-network-load-storage data as a quaternary interaction tensor includes: modeling source-network-load-storage data as a quaternary interaction tensor through a structured attention mechanism.

[0061] Preferably, regarding the structured attention mechanism, it is as follows: Let the input vector be: ,in, These are characteristics of the power grid (such as voltage amplitude and phase angle, line power flow, frequency deviation, etc.). The time step (prediction window length) is the time step. Each entity dimension is a feature dimension, and 4 is the entity dimension. ).

[0062] Time-series attention: Perform time-series attention on each entity dimension. , in, For the first Attention to time series data across entity dimensions.

[0063] Regarding obtaining , and Specifically: Such as a certain entity dimension (e.g.) The input vector is The three sets of learnable weight matrices (which are linear transformation parameters in the neural network layers, automatically updated during training via backpropagation to minimize the loss function) are mapped as follows: , in, This is the first weight matrix, used to weight the input vector. Mapped to the first Query vectors for each entity dimension , This is the second weight matrix, used to weight the input vector. Mapped to the first Key vectors of each entity dimension , This is the third weight matrix, used to weight the input vector. Mapped to the first Value vector of each entity dimension .

[0064] Entity-Dimensional Attention: Concatenate the attention sequences of each entity dimension's time series to generate... , in, For entity dimension Attention to time series, For entity dimension Attention to time series, For entity dimension Attention to time series, For entity dimension Attention to time series data.

[0065] right Perform attention between entities: , in, Attention is focused on the entity dimension.

[0066] Regarding obtaining , and Methods and acquisition , and The method is the same, so it will not be repeated here.

[0067] Combining time series attention and entity-level attention: , in, For a quaternion interaction tensor, This is the weight matrix. This is a bias term.

[0068] The execution and feedback module includes: continuous monitoring of the power system; when the rate of change of key monitoring indicators (such as power, frequency, SOC, etc.) exceeds the rate of change threshold, the edge node generates and executes an emergency control strategy.

[0069] Specifically, after the edge node generates and executes the emergency control strategy, it updates the symbolic tensor flow and re-executes the modeling and inference module and the execution and feedback module.

[0070] Example 3 This invention also proposes a storage medium storing multiple instructions, which are used to implement the aforementioned three-layer architecture-based virtual power plant source-grid-load-storage coordinated control method.

[0071] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0072] Optionally, in this embodiment, the storage medium is configured to store program code for performing the method steps of Embodiment 1.

[0073] Example 4 This invention also proposes an electronic device, including a processor and a storage medium connected to the processor. The storage medium stores multiple instructions, which can be loaded and executed by the processor to enable the processor to execute the aforementioned three-layer architecture-based virtual power plant source-grid-load-storage coordinated control method.

[0074] Specifically, the electronic device in this embodiment can be a computer terminal, which may include one or more processors and a storage medium.

[0075] The storage medium can be used to store software programs and modules, such as the three-layer architecture-based virtual power plant source-grid-load-storage coordinated control method in this embodiment of the invention. The processor executes the software programs and modules stored in the storage medium to perform various functional applications and data processing, thus realizing the aforementioned three-layer architecture-based virtual power plant source-grid-load-storage coordinated control method. The storage medium may include high-speed random access storage media, and may also include non-volatile storage media, such as one or more magnetic storage systems, flash memory, or other non-volatile solid-state storage media. In some instances, the storage medium may further include storage media remotely configured relative to the processor, which can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0076] The processor can execute the method steps of Embodiment 1 by calling the information and application stored in the storage medium through the transmission system.

[0077] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0078] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0079] In the several embodiments provided by this invention, it should be understood that the disclosed technical content can be implemented in other ways. The system embodiments described above are merely illustrative; for example, 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, indirect coupling or communication connection between units or modules, and may be electrical or other forms.

[0080] 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.

[0081] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0082] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, optical disks, and other media capable of storing program code.

[0083] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for coordinated control of power plant generation, grid, load, and storage based on a three-layer architecture, characterized in that, include: The three-tier architecture consists of a data fusion layer, a modeling and inference layer, and an execution and feedback layer. The data fusion layer includes: collecting multi-source heterogeneous data from the power system, constructing a multi-source heterogeneous raw data set, fusing the multi-source heterogeneous data, and generating a fused symbolic tensor flow; The modeling and reasoning layer includes: inputting symbolic tensor flow into the probabilistic prediction model to calculate the uncertainty of new energy output prediction and the uncertainty of absorption capacity, and establishing the joint uncertainty distribution of the two; modeling the source-grid-load-storage data as a quaternary interaction tensor, and combining the joint uncertainty distribution to generate a set of candidate regulation strategies, and selecting the optimal regulation strategy to regulate the source-grid-load-storage. The execution and feedback layer includes: continuous monitoring of the power system; when the rate of change of key monitoring indicators exceeds the rate of change threshold, the edge nodes generate and execute emergency control strategies.

2. The method for coordinated control of virtual power plant source-grid-load-storage based on a three-layer architecture as described in claim 1, characterized in that, The fusion of multi-source heterogeneous data includes fusing the multi-source heterogeneous data through a cross-temporal adaptive mapping matrix.

3. The method for coordinated control of virtual power plant source-grid-load-storage based on a three-layer architecture as described in claim 2, characterized in that, The spatiotemporal adaptive mapping matrix includes: Suppose that there are multi-source heterogeneous data. One data source, at any time No. The original sequences of the data sources are , No. The sampling frequency of each data source is , No. The data credibility of each data source is The unified target time window length is The target sampling frequency is ,but ， in, For the first The data source corresponds to the first one on the unified time grid. The observations at each time point are mapped to the weights of the symbolic tensor flow. The time decay factor, For the first The data source corresponds to the first one on the unified time grid. At that moment The observed values, For the first The mean of the data source, For the first Standard deviation of the data source To prevent division by zero of small constants.

4. The method for coordinated control of virtual power plant source-grid-load-storage based on a three-layer architecture as described in claim 1, characterized in that, The modeling and inference layer includes: Based on the symbolic tensor flow, a probabilistic prediction model is used to model the future output of new energy sources, resulting in a prediction distribution that includes the mean and variance, which characterizes the uncertainty of new energy output prediction. Based on the symbolic tensor flow, the future load level is modeled according to the probabilistic prediction model, and the future available energy storage capacity is modeled in combination with the energy storage status, resulting in a probability distribution containing mean and variance, which characterizes the uncertainty of absorption capacity. Based on the predicted distribution of new energy output, the future load distribution, and the future available energy storage capacity distribution, the joint uncertainty distribution of the uncertainty of new energy output prediction and the uncertainty of absorption capacity is obtained through the dependency modeling function. Based on the joint uncertainty distribution, the probability that the predicted renewable energy output is less than or equal to the sum of the predicted load and the predicted energy storage capacity is obtained, so as to determine the power system's capacity to absorb renewable energy output and output absorption risk warnings.

5. The method for coordinated control of virtual power plant source-grid-load-storage based on a three-layer architecture as described in claim 4, characterized in that, If the probability that the predicted renewable energy output is less than or equal to the sum of the predicted load and the predicted energy storage capacity is less than or equal to 1 and greater than the preset absorption threshold, then the power system can absorb renewable energy. If it is less than or equal to 1 and less than the preset absorption threshold, then an absorption risk warning will be output.

6. The method for coordinated control of virtual power plant source-grid-load-storage based on a three-layer architecture as described in claim 1, characterized in that, Modeling source-grid-load-storage data as a quaternary interaction tensor involves using a structured attention mechanism to model the source-grid-load-storage data as a quaternary interaction tensor.

7. The method for coordinated control of virtual power plant source-grid-load-storage based on a three-layer architecture as described in claim 1, characterized in that, After the edge node generates and executes the emergency control strategy, it updates the symbolic tensor flow and re-executes the modeling and inference layer and the execution and feedback layer.

8. A virtual power plant source-grid-load-storage coordinated control system based on a three-layer architecture, characterized in that, include: The three-tier architecture consists of a data fusion module, a modeling and inference module, and an execution and feedback module. Composition, in which, The data fusion module includes: collecting multi-source heterogeneous data from the power system, constructing a multi-source heterogeneous raw data set, fusing the multi-source heterogeneous data, and generating a fused symbolic tensor flow; The modeling and reasoning module includes: inputting symbolic tensor flow into the probabilistic prediction model to calculate the uncertainty of new energy output prediction and the uncertainty of absorption capacity, and establishing the joint uncertainty distribution of the two; modeling the source-grid-load-storage data as a quaternary interaction tensor, and combining the joint uncertainty distribution to generate a set of candidate regulation strategies, and selecting the optimal regulation strategy to regulate the source-grid-load-storage. The execution and feedback module includes: continuous monitoring of the power system; when the rate of change of key monitoring indicators exceeds the rate of change threshold, the edge node generates and executes an emergency control strategy.

9. A virtual power plant source-grid-load-storage coordinated control system based on a three-layer architecture as described in claim 8, characterized in that, The fusion of multi-source heterogeneous data includes fusing the multi-source heterogeneous data through a cross-temporal adaptive mapping matrix.

10. The method for coordinated control of virtual power plant source-grid-load-storage based on a three-layer architecture as described in claim 9, characterized in that, The spatiotemporal adaptive mapping matrix includes: Suppose that there are multi-source heterogeneous data. One data source, at any time No. The original sequences of the data sources are , No. The sampling frequency of each data source is , No. The data credibility of each data source is The unified target time window length is The target sampling frequency is ,but ， in, For the first The data source corresponds to the first one on the unified time grid. The observations at each time point are mapped to the weights of the symbolic tensor flow. The time decay factor, For the first The data source corresponds to the first one on the unified time grid. At that moment The observed values, For the first The mean of the data source, For the first Standard deviation of the data source To prevent division by zero of small constants.

Images