Distributed energy dynamic load adjusting method and system based on virtual power plant

Abstract

The invention discloses a distributed energy dynamic load adjusting method and system based on a virtual power plant, and relates to the technical field, and the method comprises the following steps: obtaining first data of the virtual power plant, and building a topological structure of the virtual power plant; constructing a distributed energy output prediction model based on an LSTM deep learning algorithm; combining the output of the distributed energy output prediction model and the topological structure to generate an energy scheduling scheme in real time; monitoring real-time data of the distributed energy, comparing the real-time data with the energy scheduling scheme, and performing feedback updating; according to the invention, a virtual power plant topological structure is constructed, a prediction model is constructed based on an LSTM algorithm, an energy scheduling scheme is generated in combination with model output and the topological structure, and a feedback updating technology is carried out. The problems that existing regulation and control based on a static model cannot adapt to dynamic changes of energy production and load requirements in real time, and due to the fact that an effective cooperation mechanism is lacked between distributed energy sources, energy waste is caused, and the regulation efficiency is low are solved.

Description

Technical Field

[0001] The present invention relates to the field of distributed energy regulation technology, and more specifically, to a distributed energy dynamic load regulation method and system based on a virtual power plant. Background Art

[0002] With the acceleration of global energy transformation, distributed energy, as an important energy supply method, is increasingly being used in modern energy systems. Distributed energy usually refers to energy production and storage facilities at or near the user end. Unlike traditional centralized power grid systems, distributed energy meets local load demands through widely distributed small power generation units. Along with this, the management and optimization of distributed energy loads has become one of the key technologies in the operation of energy systems.

[0003] For example, the invention patent with announcement number: CN114784790A announces a method and system for full-load regulation of a distributed energy station, which includes the following steps: establishing an equipment model and a system economic model of a distributed energy system; obtaining historical operation data of the distributed energy system according to the energy consumption characteristics of users, and constructing a basic load forecasting model; correcting the basic load forecasting model through weather parameters and production plans reported by users on their own; dividing the load of the distributed energy system into three load division intervals according to the configuration of the internal combustion engine module in the distributed energy system and the stable operation characteristics of the internal combustion engine, and setting different operation optimization strategies for the three load division intervals; determining the load division interval in which the user load is located according to the current load forecast result of the load forecasting basic model, taking the optimal economy as the optimization goal, executing the operation optimization strategy for the corresponding load division interval, and realizing stable regulation of the distributed energy system within the load range of 0-120%, thereby improving the stability and operation economy of the distributed energy system.

[0004] For example, the invention patent with announcement number CN117713212A announces a method and system for adjusting the power factor of the plant load of a distributed energy station, including the following steps: by monitoring the grid power factor and photovoltaic power factor of the plant load in each monitoring period, and automatically adjusting the photovoltaic power factor according to the relationship between the grid power factor and the power factor threshold and the current light intensity. When the light intensity is strong, the photovoltaic power factor is set to the minimum photovoltaic power factor threshold to prevent the plant load from absorbing too much active power from the photovoltaic power generation system, thereby causing instability in the power system. When the light intensity is not strong, the photovoltaic power factor is increased or decreased according to the comparison result between the grid power factor and the power factor threshold, so that the distributed energy station can fully utilize the power generated by the photovoltaic power generation system under the premise of ensuring the stable operation of the power system, while reducing energy waste, and further reducing the electricity cost of the distributed energy station.

[0005] The above disclosed technical solutions have at least the following technical problems: With the widespread application of distributed energy, its intermittency and volatility have brought challenges to the stable operation of the power grid. Traditional load regulation methods are difficult to effectively cope with the complex situation after large-scale access to distributed energy. In existing technologies, the regulation of distributed energy is mostly based on static models, which cannot adapt to the dynamic changes of energy production and load demand in real time. Moreover, there is a lack of effective coordination mechanism between distributed energy sources, resulting in energy waste and low regulation efficiency.

[0006] In view of the above problems, the present invention proposes a solution. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a distributed energy dynamic load regulation method and system based on a virtual power plant, which can adapt to the dynamic changes in energy production and load demand, achieve a dynamic balance between energy production and consumption, and improve system reliability, stability and energy utilization efficiency.

[0008] To achieve the above object, the present invention provides the following technical solutions: A distributed energy dynamic load regulation method based on a virtual power plant includes the following steps: obtaining the first data of the virtual power plant and establishing the topological structure of the virtual power plant; building a distributed energy output prediction model based on an LSTM deep learning algorithm; generating an energy scheduling plan in real time based on the output and topological structure of the distributed energy output prediction model; monitoring the real-time data of distributed energy, comparing it with the energy scheduling plan, and performing feedback updates.

[0009] In a preferred embodiment, the first data of the virtual power plant is obtained and the topological structure of the virtual power plant is established, specifically: the first data of the virtual power plant is obtained, and the first data of the virtual power plant includes site information of distributed energy, energy storage equipment information and user-end location information; the first data of the virtual power plant is used as a node in the topological structure of the virtual power plant; the electrical connection relationship between each node is obtained, and the edges of the topological structure are set according to the electrical connection relationship; and the topological structure of the virtual power plant is constructed by connecting each node with the corresponding edge.

[0010] In a preferred embodiment, a distributed energy output prediction model is constructed based on the LSTM deep learning algorithm, specifically: photovoltaic characteristic data, wind energy characteristic data and load data as well as historical first characteristic data are obtained, wherein the historical first characteristic data includes historical photovoltaic characteristic data, historical wind energy characteristic data and historical load data; the historical first characteristic data is cleaned and normalized, and the processed data is divided into a training set and a test set; the model is trained with the training set, and the model is optimized using L2 regularization at the LSTM layer to obtain a distributed energy output prediction model; the photovoltaic characteristic data, wind energy characteristic data and load data are used as inputs of the distributed energy output prediction model.

[0011] In a preferred embodiment, an energy scheduling plan is generated in real time in combination with the output of a distributed energy output prediction model and a topological structure, specifically: obtaining the output of several distributed energy output prediction models within a preset H time period and the total power demand of the area to be scheduled; obtaining the location information and line parameters of the area to be scheduled based on the topological structure; analyzing the location information of the area to be scheduled and the location of the distributed energy, sorting them in ascending order by distance, and obtaining the initial area to be scheduled; analyzing the line loss and voltage level of the initial area to be scheduled according to the line parameters, performing secondary screening, and obtaining the priority area to be scheduled; generating an energy scheduling plan based on the priority area to be scheduled in combination with the output of the distributed energy output prediction model and the total power demand of the area to be scheduled.

[0012] In a preferred embodiment, the line loss and voltage level of the initial area to be scheduled are analyzed according to the line parameters, and a secondary screening is performed to obtain the priority area to be scheduled, specifically: the voltage amplitude and phase angle of each node in the initial area to be scheduled are obtained by power flow calculation; the voltage offset rate is calculated according to the voltage amplitude and phase angle; if the voltage offset rate is within a preset voltage offset rate stable range, the initial area to be scheduled is marked; the line loss of the marked initial area to be scheduled is calculated according to the line parameters, and the line losses are compared and analyzed to obtain the priority area to be scheduled.

[0013] In a preferred embodiment, an energy scheduling plan is generated based on the priority area to be scheduled combined with the output of the distributed energy output prediction model and the total power demand of the area to be scheduled, specifically: obtaining the location information and storage device capacity of the storage device based on the topological structure; if the output of the distributed energy output prediction model is greater than the total power demand, the remaining power after the scheduling is completed is analyzed according to the storable capacity of the storage device and the distance to the distributed energy site, and stored in the storage device; if the output of the distributed energy output prediction model is less than the total power demand, it is analyzed according to the releasable capacity of the storage device and the distance to the priority area to be scheduled, and according to the real-time energy supply and demand situation and electricity price information, the scheduling of the power in the storage device is dynamically adjusted.

[0014] In a preferred embodiment, the real-time data of distributed energy is monitored and compared with the energy scheduling plan, and feedback is updated, specifically: the real-time data of distributed energy is monitored, the real-time data includes the actual output value of distributed energy and the actual storable and releasable capacity of storage equipment; the real-time data of distributed energy is compared with the energy scheduling plan, if the output deviation of the actual output value of distributed energy and the distributed energy output prediction model exceeds a preset deviation threshold, the distributed energy output prediction model is optimized; if the deviation of the actual storable and releasable capacity of the storage device and the energy scheduling plan exceeds a preset deviation threshold, the strategy of the storage device is adjusted to regenerate the energy scheduling plan.

[0015] In a preferred embodiment, the photovoltaic characteristic data is specifically acquired in the following steps: acquiring photovoltaic module data, the photovoltaic module data including the effective area of ​​the photovoltaic module, the photoelectric conversion efficiency of the photovoltaic module and the temperature of the photovoltaic module; obtaining the solar radiation intensity and the illumination time by analyzing the current meteorological data; obtaining the output power of the photovoltaic module based on the Hollands model according to the photovoltaic module data and the solar radiation intensity; obtaining the efficiency of the inverter, and obtaining the photovoltaic characteristic data in combination with the output power of the photovoltaic module.

[0016] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: Through the construction of the topological structure, the appropriate type is selected according to the scale, distribution and functional requirements of the virtual power plant. By obtaining the information of distributed energy, energy storage equipment and user end as nodes, and setting edges in combination with the electrical connection relationship, the topological structure can be constructed to achieve a dynamic balance between energy production and consumption, monitor and dispatch distributed energy in real time, improve system reliability, stability and load management efficiency, optimize resource allocation, and improve energy utilization efficiency. In terms of distributed energy output prediction, a model is constructed based on the LSTM deep learning algorithm. The model can capture the long-term and short-term dependencies of time series data, handle nonlinear relationships, have strong adaptability, improve prediction accuracy, have strong scalability and is suitable for multi-dimensional input. By cleaning and normalizing the relevant data, dividing the training set and the test set, and constructing and optimizing the model, it can help operators better dispatch energy and optimize resource allocation. 3. In the generation of energy dispatching schemes, by obtaining the output of the distributed energy output prediction model, the total power demand of the area to be dispatched, the location information and the line parameters, the priority area to be dispatched is selected through distance sorting, line loss and voltage level analysis, and the scheme is generated in combination with multiple aspects of data to achieve reasonable dispatch according to energy supply and demand and line conditions.

[0017] 4. In terms of feedback updates, monitor the real-time data of distributed energy, compare it with the energy dispatch plan, optimize and adjust the model and energy storage strategy according to the output forecast deviation and energy storage equipment capacity deviation, ensure the accuracy and adaptability of energy dispatch, and thus continuously improve energy utilization efficiency and system stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A schematic flow chart of a distributed energy dynamic load regulation method based on a virtual power plant provided in an embodiment of the present application.

[0019] Figure 2 A schematic diagram of the structure of a distributed energy dynamic load regulation system based on a virtual power plant provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Embodiment 1, Figure 1 The schematic diagram of the process flow of the distributed energy dynamic load adjustment method based on the virtual power plant provided in the embodiment of the present application includes the following steps: S1, obtaining the first data of the virtual power plant and establishing the topological structure of the virtual power plant.

[0022] In this example, the appropriate topology type is selected according to the scale, distribution characteristics and functional requirements of the virtual power plant. Common topologies include star, tree, mesh, etc. Generally speaking, a virtual power plant with a smaller scale and relatively concentrated distribution can adopt a star structure to facilitate centralized control and management; a virtual power plant with a larger scale and wider distribution can consider a tree or mesh structure to improve the reliability and flexibility of the system.

[0023] The first data of the virtual power plant is obtained, and the topological structure of the virtual power plant is established, specifically: Acquire first data of a virtual power plant, wherein the first data of the virtual power plant includes site information of distributed energy, energy storage device information, and user terminal location information; Among them, the site information of distributed energy includes the type and location information of distributed energy equipment, the capacity and power of each type of distributed energy equipment, the energy storage equipment information includes the capacity, charging and discharging efficiency and installation location information of the energy storage equipment, and the user terminal location information includes the user terminal location information and user power consumption characteristics; The site information of distributed energy, energy storage equipment information and user location information are used as nodes in the topology of the virtual power plant; Obtaining the electrical connection relationship between each node, and setting the edge of the topological structure according to the electrical connection relationship; The electrical connection relationship here refers to the connection between the site information of distributed energy, the energy storage equipment information and the geographical location information of the user end, such as connection through high-voltage transmission lines, medium-voltage distribution networks or low-voltage microgrids. At the same time, the transmission capacity, loss and reliability factors of different connection methods are considered; The topological structure of the virtual power plant is constructed by connecting each node with the corresponding edge.

[0024] Virtual power plants achieve a dynamic balance between energy production and consumption by integrating multiple distributed energy sources (such as solar energy, wind energy, energy storage equipment, etc.) and load regulation equipment. Its topological structure enables each distributed energy system to work in coordination, improving the reliability and stability of the overall system. Through flexible scheduling, virtual power plants can respond quickly to load fluctuations and reduce instability in system operation.

[0025] The load regulation of traditional power systems is often carried out through large centralized power plants, while the topology of virtual power plants can monitor and dispatch distributed energy in real time. By combining with demand response management, energy storage systems, etc., virtual power plants can dynamically adjust loads when demand surges or decreases to help balance supply and demand. This instant regulation capability not only improves load management efficiency, but also provides a guarantee for the stable operation of the power market.

[0026] In the topology of a virtual power plant, distributed energy can be dynamically dispatched based on real-time load demand and energy prices. For example, when some distributed energy has excess power generation capacity, the virtual power plant can store or sell excess power back to the grid. Conversely, when power generation is insufficient, other energy sources can be called upon. This optimized resource allocation method not only improves energy utilization efficiency, but also reduces energy waste.

[0027] S2, builds a distributed energy output prediction model based on the LSTM deep learning algorithm.

[0028] In this example, LSTM is a neural network model specially designed to process time series data, which can effectively capture the long-term and short-term dependencies of data. Using LSTM deep learning algorithm to build a distributed energy output prediction model has the following advantages: Capturing long-term and short-term dependencies of time series data: In distributed energy systems, energy output is usually a continuous time series data that may be affected by multiple factors. LSTM can automatically learn these complex time dependencies, thereby making more accurate predictions.

[0029] Handling nonlinear relationships: Traditional time series forecasting methods have certain limitations when dealing with data with nonlinear characteristics, while LSTM can handle complex nonlinear relationships well. The output of distributed energy systems is affected by many nonlinear factors, and LSTM can automatically model these nonlinear relationships through its multi-layer network structure, thereby improving forecasting accuracy.

[0030] Strong adaptability: LSTM has strong adaptability and can dynamically adjust its internal state and parameters according to new input data. This enables LSTM to handle changes in energy output under various environmental conditions. For distributed energy systems, especially in the face of uncertain factors such as weather changes, equipment aging, and load fluctuations, LSTM can adapt and adjust the prediction results more flexibly.

[0031] Improve prediction accuracy: Since LSTM can process and model complex time series data features, it can significantly improve the accuracy of the model when applied to distributed energy output forecasting. Accurate forecasting can help operators better dispatch energy, optimize the allocation of energy resources, and improve energy utilization efficiency.

[0032] Strong scalability: One advantage of the LSTM model is its strong scalability, which is particularly suitable for processing large-scale distributed energy data. In a distributed energy system where multiple energy forms operate in parallel, LSTM can model and accurately predict the energy output from multiple sources.

[0033] Applicable to multi-dimensional input: The output prediction of distributed energy systems does not rely solely on a single factor, but may involve multi-dimensional data such as meteorological data, equipment status data, and grid load data. LSTM can effectively process these multi-dimensional time series data and integrate information from various factors, thereby improving the comprehensive prediction ability of the model.

[0034] A distributed energy output prediction model is built based on the LSTM deep learning algorithm, specifically: Acquire photovoltaic characteristic data, wind energy characteristic data, load data and historical first characteristic data, wherein the historical first characteristic data includes historical photovoltaic characteristic data, historical wind energy characteristic data and historical load data; The first threshold is obtained by using the median algorithm for the historical photovoltaic characteristic data, the historical wind energy characteristic data and the historical load data; The first threshold is taken as a separate class, the historical first feature data is screened by the K-Medoids algorithm, and the screened data is divided into a training set and a test set; The photovoltaic characteristic data, wind energy characteristic data and load data are used as the input layer, and the input layer dimension is set to 3. Then, the LSTM layer is added, and the number of neurons in the LSTM layer is set to the preset N. The output prediction value of distributed energy is used as the output layer, and the output layer dimension is 1. Connect the input layer, LSTM layer, and output layer to build a complete LSTM model; The model is trained with the training set, and L2 regularization is used in the LSTM layer to optimize the model to obtain the distributed energy output prediction model; Use the test set to evaluate the trained model; Photovoltaic characteristic data, wind energy characteristic data and load data are used as inputs of the distributed energy output prediction model.

[0035] It should be noted that adding regularization terms, such as L1 or L2 regularization, to the model can help prevent overfitting, improve the generalization ability of the model, and improve the prediction accuracy of the model. The data cleaning and normalization processing are existing technologies and will not be elaborated here. The load data is the actual load data of the power system over a period of time.

[0036] Among them, photovoltaic characteristic data refers to the power output data generated by the photovoltaic power generation system in distributed energy within a certain period of time.

[0037] The photovoltaic characteristic data is specifically obtained in the following steps: Acquiring photovoltaic module data, wherein the photovoltaic module data includes an effective area of ​​the photovoltaic module, a photoelectric conversion efficiency of the photovoltaic module, and a temperature of the photovoltaic module; Analyze the current meteorological data to obtain the solar radiation intensity and illumination time; The output power of the photovoltaic module is obtained based on the Hollands model according to the photovoltaic module data and solar radiation intensity; The efficiency of the inverter is obtained, and the photovoltaic characteristic data is obtained in combination with the output power of the photovoltaic module.

[0038] The specific calculation formula for the output power of the photovoltaic module is as follows:

[0039] The specific calculation formula of the photovoltaic characteristic data is as follows:

[0040] In the formula, is the photovoltaic characteristic data, is the output power of the photovoltaic module, is the efficiency of the inverter, is the solar radiation intensity, is the effective area of ​​the photovoltaic module, is the photoelectric conversion efficiency of the photovoltaic module, is the temperature of the photovoltaic module, is the reference temperature, is the temperature coefficient of the photovoltaic module, For the light exposure time.

[0041] It should be noted that light intensity is one of the most critical factors affecting the output power of photovoltaic cells. It directly affects the photoelectric conversion process on the solar panel and is usually obtained based on meteorological data or remote sensing data. The reference temperature is usually set to 25°C, which is the reference temperature under standard test conditions. The output power of the photovoltaic module is affected by the intensity of solar radiation and the duration of illumination, and the photovoltaic characteristic data is affected by the efficiency of the photovoltaic inverter, which refers to the conversion of photovoltaic converted direct current into usable alternating current. The Hollands model is a dynamic model used to simulate and predict complex systems, especially in the study of energy systems and power markets. It is often used to describe the relationship between different types of energy, load demand, price changes, and energy transmission.

[0042] Wind energy characteristic data refers to the power output data generated by the wind power generation system in distributed energy within a certain period of time.

[0043] The specific steps for obtaining the wind energy characteristic data are as follows: The wind speed is obtained through meteorological data, and the distribution data of the wind speed is obtained based on the probability density function of the Weibull distribution; Wind energy characteristic data is obtained according to the distribution of wind speed in combination with wind energy generation system data, wherein the wind energy generation system data includes the area of ​​the wind energy generation system and the wind power conversion efficiency.

[0044] The specific calculation formula of the wind speed distribution data is as follows: ,

[0045] The specific calculation formula of the wind energy characteristic data is as follows:

[0046] In the formula, is the wind energy characteristic data, is the probability density function of the Weibull distribution, is the wind power conversion efficiency of the wind power generation system, is the area of ​​the wind power generation system, Working time of wind power generation system, is the shape parameter, is the scale parameter, is the wind speed, The minimum wind speed for the wind power generation system to work. The maximum wind speed at which a wind power generation system operates.

[0047] It should be noted that the minimum and maximum wind speeds for the wind power generation system are obtained from the manual of the wind power generation system. Weibull distribution is a probability distribution function that describes the change of wind speed. Through this model, the distribution of wind speed within a certain period of time can be predicted, thereby estimating the wind energy output.

[0048] The specific calculation formula of the distributed energy output prediction model is as follows:

[0049] In the formula, is the output forecast value of distributed energy, is the wind energy characteristic data, is the photovoltaic characteristic data, is the load data, is the weight of photovoltaic characteristic data, is the weight of wind energy characteristic data, is the weight of the load data.

[0050] S3, combines the output of the distributed energy output prediction model and the topological structure to generate an energy dispatch plan in real time, specifically: Obtain the output of several distributed energy output prediction models within a preset H time period and the total power demand of the area to be dispatched; Acquire location information and line parameters of the area to be scheduled based on the topological structure, wherein the line parameters include conductance and susceptance; Analyze the location information of the area to be dispatched and the location of the distributed energy, sort them in order from small to large distance, and obtain the initial area to be dispatched; According to the line parameters, the line loss and voltage level of the initial area to be dispatched are analyzed, and a secondary screening is performed to obtain the priority area to be dispatched; And generate an energy dispatch plan based on the priority dispatch area combined with the output of the distributed energy output prediction model and the total power demand of the dispatch area.

[0051] The line loss and voltage level of the initial area to be scheduled are analyzed according to the line parameters, and a secondary screening is performed to obtain the priority area to be scheduled, which is specifically: The voltage amplitude and phase angle of each node in the initial dispatching area are obtained through power flow calculation; By calculating the voltage deviation rate according to the voltage amplitude and phase angle; If the voltage deviation rate is within the preset voltage deviation rate stable range, the initial area to be scheduled is marked; The line loss of the marked initial area to be scheduled is calculated according to the line parameters, and the line loss is compared and analyzed to obtain the priority area to be scheduled. The comparison and analysis here refers to sorting from small to large according to the size of the line loss.

[0052] The specific calculation formula of the voltage offset rate is as follows:

[0053] The specific calculation formula of the line loss is as follows:

[0054] In the formula, is the line loss, For Node and nodes The conductivity, For Node and nodes The phase difference, is the voltage offset rate, For Node Voltage amplitude, For Node Rated voltage, For Node The voltage amplitude.

[0055] The energy dispatching plan is generated according to the output of the distributed energy output prediction model combined with the priority dispatching area and the total power demand of the dispatching area, specifically: If the output of the distributed energy output prediction model is greater than the total power demand, the distributed energy is dispatched according to the priority area to be dispatched. The remaining power after the dispatch is completed is analyzed according to the storage capacity of the storage device and the distance from the distributed energy site, and stored in the storage device; first, the storage device with a smaller distance to the distributed energy site is given priority, and then the capacity of the storage device is considered. If the storable power is full, the next storage device is replaced; If the output of the distributed energy output prediction model is less than the total power demand, after the priority area to be dispatched, the available capacity of the storage device and the distance to the priority area to be dispatched are analyzed, and the dispatch of the power in the storage device is dynamically adjusted according to the real-time energy supply and demand situation and electricity price information; first look at the capacity of the storage device to see whether the available power can meet the remaining power demand scheduling. If it does, consider the distance between the storage device and the priority area to be dispatched, and the smaller the distance, the higher the priority.

[0056] S4 monitors the real-time data of distributed energy, compares it with the energy dispatch plan, and provides feedback updates, specifically: Monitor the real-time data of distributed energy, including the actual output value of distributed energy and the actual storable and releasable capacity of storage equipment; Compare the real-time data of distributed energy with the energy dispatch plan. If the deviation between the actual output value of distributed energy and the output of the distributed energy output prediction model exceeds the preset deviation threshold, the distributed energy output prediction model is optimized. If the deviation is small, the model performance can be optimized by adjusting some parameters of the model (such as learning rate, number of hidden layer neurons, etc.) to better adapt to the changes in real-time data. If the deviation is large, it may be necessary to retrain the model, use the latest collected historical data and real-time data, rebuild the training set, validation set and test set, and conduct comprehensive training and optimization of the model to improve the accuracy of the prediction; If the actual storable and releasable capacity of the storage device deviates from that in the energy scheduling plan by more than a preset deviation threshold, the strategy of the storage device is adjusted to regenerate the energy scheduling plan.

[0057] Embodiment 2, Figure 2 A schematic diagram of the structure of a distributed energy dynamic load regulation system based on a virtual power plant provided in an embodiment of the present application includes a topology structure building module, a prediction model building module, an energy scheduling scheme generating module, and a feedback updating module, and there are connections between the modules: A topology structure building module, used to obtain first data of the virtual power plant and establish a topology structure of the virtual power plant; Prediction model building module, used to build a distributed energy output prediction model based on the LSTM deep learning algorithm; Energy dispatch plan generation module, used to generate energy dispatch plan in real time by combining the output of distributed energy output prediction model and topology structure; The feedback update module is used to monitor the real-time data of distributed energy, compare it with the energy scheduling plan, and perform feedback updates.

[0058] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0059] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0060] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0061] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0062] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0063] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A distributed energy dynamic load regulation method based on a virtual power plant, characterized in that: The steps include: Acquire first data of the virtual power plant and establish a topological structure of the virtual power plant; Build a distributed energy output prediction model based on LSTM deep learning algorithm; Combine distributed energy output prediction models and topological structures to generate energy scheduling plans in real time; Monitor real-time data of distributed energy, compare and analyze it with energy dispatch plans, and provide real-time feedback.

2. The distributed energy dynamic load regulation method based on virtual power plant according to claim 1 is characterized in that: The step of obtaining the first data of the virtual power plant and establishing the topological structure of the virtual power plant is specifically as follows: Acquire first data of a virtual power plant, wherein the first data of the virtual power plant includes site information of distributed energy, energy storage device information, and user terminal location information; Using the first data of the virtual power plant as a node in the topological structure of the virtual power plant; Obtaining the electrical connection relationship between each node, and setting the edge of the topological structure according to the electrical connection relationship; The topological structure of the virtual power plant is constructed by connecting each node with the corresponding edge.

3. The distributed energy dynamic load regulation method based on virtual power plant according to claim 1 is characterized in that: The distributed energy output prediction model is constructed based on the LSTM deep learning algorithm, specifically: Acquire photovoltaic characteristic data, wind energy characteristic data, load data and historical first characteristic data, wherein the historical first characteristic data includes historical photovoltaic characteristic data, historical wind energy characteristic data and historical load data; The first threshold is obtained by using the median algorithm for the historical photovoltaic characteristic data, the historical wind energy characteristic data and the historical load data; The first threshold is taken as a separate class, the historical first feature data is screened by the K-Medoids algorithm, and the screened data is divided into a training set and a test set; The model is trained with the training set, and L2 regularization is used in the LSTM layer to optimize the model to obtain the distributed energy output prediction model; Photovoltaic characteristic data, wind energy characteristic data and load data are used as inputs of the distributed energy output prediction model.

4. The distributed energy dynamic load regulation method based on virtual power plant according to claim 1 is characterized in that: The energy dispatching scheme is generated in real time by combining the distributed energy output prediction model and the topological structure, specifically: Obtain the output of several distributed energy output prediction models within a preset H time period and the total power demand of the area to be dispatched; Acquire location information and line parameters of the area to be scheduled based on the topological structure; Analyze the location information of the area to be dispatched and the location of the distributed energy, sort them in order from small to large distance, and obtain the initial area to be dispatched; According to the line parameters, the line loss and voltage level of the initial area to be dispatched are analyzed, and a secondary screening is performed to obtain the priority area to be dispatched; An energy dispatch plan is generated based on the priority dispatch area combined with the output of the distributed energy output prediction model and the total power demand of the dispatch area.

5. The distributed energy dynamic load regulation method based on virtual power plant according to claim 4 is characterized in that: The line loss and voltage level of the initial area to be scheduled are analyzed according to the line parameters, and a secondary screening is performed to obtain the priority area to be scheduled, which is specifically: The voltage amplitude and phase angle of each node in the initial dispatching area are obtained through power flow calculation; By calculating the voltage deviation rate according to the voltage amplitude and phase angle; If the voltage deviation rate is within the preset voltage deviation rate stable range, the initial area to be scheduled is marked; The line loss of the marked initial area to be scheduled is calculated according to the line parameters, and the line loss is compared and analyzed to obtain the priority area to be scheduled.

6. The distributed energy dynamic load regulation method based on virtual power plant according to claim 4 is characterized in that: The energy dispatching scheme is generated according to the priority area to be dispatched combined with the output of the distributed energy output prediction model and the total power demand of the area to be dispatched, specifically: Acquire location information and storage device capacity of storage devices based on the topological structure; If the output of the distributed energy output prediction model is greater than the total power demand, the remaining power after the dispatch is completed will be analyzed based on the storage capacity of the storage device and the distance to the distributed energy site, and stored in the storage device; If the output of the distributed energy output prediction model is less than the total power demand, an analysis is performed based on the available capacity of the storage device and the distance to the priority dispatch area, and the dispatch of electricity in the storage device is dynamically adjusted based on the real-time energy supply and demand situation and electricity price information.

7. The distributed energy dynamic load regulation method based on virtual power plant according to claim 1 is characterized in that: The real-time data of distributed energy is monitored, and compared and analyzed with the energy dispatch plan, and real-time feedback is provided, specifically: Monitor the real-time data of distributed energy, including the actual output value of distributed energy and the actual storable and releasable capacity of storage equipment; Compare the real-time data of distributed energy with the energy dispatch plan. If the deviation between the actual output value of distributed energy and the output of the distributed energy output prediction model exceeds a preset deviation threshold, optimize the distributed energy output prediction model. If the actual storable and releasable capacity of the storage device deviates from that in the energy scheduling plan by more than a preset deviation threshold, the strategy of the storage device is adjusted to regenerate the energy scheduling plan.

8. The distributed energy dynamic load regulation method based on virtual power plant according to claim 3 is characterized in that: The photovoltaic characteristic data is specifically obtained in the following steps: Acquiring photovoltaic module data, wherein the photovoltaic module data includes an effective area of ​​the photovoltaic module, a photoelectric conversion efficiency of the photovoltaic module, and a temperature of the photovoltaic module; Analyze the current meteorological data to obtain the solar radiation intensity and illumination time; The output power of the photovoltaic module is obtained based on the Hollands model according to the photovoltaic module data and solar radiation intensity; The efficiency of the inverter is obtained, and the photovoltaic characteristic data is obtained in combination with the output power of the photovoltaic module.

9. The distributed energy dynamic load regulation method based on virtual power plant according to claim 8 is characterized in that: The specific calculation formula for the output power of the photovoltaic module is as follows: The specific calculation formula of the photovoltaic characteristic data is as follows: In the formula, is the photovoltaic characteristic data, is the output power of the photovoltaic module, is the efficiency of the inverter, is the solar radiation intensity, is the effective area of ​​the photovoltaic module, is the photoelectric conversion efficiency of the photovoltaic module, is the temperature of the photovoltaic module, is the reference temperature, is the temperature coefficient of the photovoltaic module, For the light exposure time.

10. A system using the distributed energy dynamic load regulation method based on a virtual power plant as described in any one of claims 1 to 9, characterized in that: It includes topology building module, prediction model building module, energy scheduling plan generation module and feedback update module. There are connections between modules: A topology structure building module, used to obtain first data of the virtual power plant and establish a topology structure of the virtual power plant; Prediction model building module, used to build a distributed energy output prediction model based on the LSTM deep learning algorithm; Energy dispatch plan generation module, used to generate energy dispatch plan in real time by combining the output of distributed energy output prediction model and topology structure; The feedback update module is used to monitor the real-time data of distributed energy, compare it with the energy scheduling plan, and perform feedback updates.

Images