A distributed energy aggregation method, device, equipment, medium and product

Abstract

The application discloses a distributed energy aggregation method and device, equipment, medium and product, and relates to the technical field of smart power grids. The method comprises the following steps: acquiring multi-source data of multiple types of distributed energy, and preprocessing the multi-source data to obtain standardized operation data; performing wind and light output prediction and load prediction on the standardized operation data by using a hybrid model of a long short-term memory network and a gradient boosting decision tree, so as to obtain new energy predicted output and load prediction values; solving a multi-objective optimization function by using an improved particle swarm optimization algorithm, so as to obtain optimal output values of each device at each time period; and generating power adjustment instructions for power balance adjustment and voltage and reactive power control based on the optimal output values of each device at each time period. The application significantly improves the new energy consumption level and operation economy, enables each distributed energy device to be collaboratively operated according to an optimal plan, realizes distributed energy aggregation, effectively suppresses power fluctuation and stabilizes voltage.

Description

Technical Field

[0001] This application relates to the field of smart grid technology, and in particular to a method, apparatus, equipment, medium and product for distributed energy aggregation. Background Technology

[0002] With the large-scale grid connection of various types of distributed energy sources such as distributed photovoltaics, wind power, energy storage, electric vehicles, and micro gas turbines, they exhibit significant characteristics of "large quantity, scattered distribution, different characteristics, and large fluctuations." There is an urgent need to integrate various dispersed distributed energy sources into a unified, adjustable, and flexible resource aggregate through effective aggregation methods to participate in grid coordinated dispatch and multi-energy complementary optimized operation.

[0003] Existing distributed energy aggregation methods typically handle data acquisition, output prediction, optimization scheduling, and command generation in a fragmented manner. Each stage operates independently, information flow is poor, and it is difficult to form an efficient closed-loop collaboration of "sensing → prediction → optimization → regulation". Specifically, the heterogeneous data formats and inconsistent dimensions of various types of distributed energy resources, coupled with noise interference and data gaps during data acquisition, make it difficult to standardize, integrate, and collaboratively utilize multi-source heterogeneous data, thus failing to provide a unified and reliable data foundation for subsequent prediction and optimization. In terms of output prediction, traditional methods often employ single prediction algorithms (such as simple time series analysis or single machine learning models), which struggle to fully capture the time-series dependence and nonlinear characteristics of renewable energy output, resulting in significant prediction errors and a lack of accurate data support for subsequent scheduling decisions. Regarding economic scheduling optimization, existing methods often use a single economic indicator as the optimization objective, failing to incorporate multiple objectives such as economic efficiency, renewable energy absorption rate, grid loss, and carbon emissions into a unified collaborative optimization framework. Furthermore, conventional particle swarm optimization algorithms are prone to getting trapped in local optima when solving complex high-dimensional optimization problems, making it difficult to generate globally optimal scheduling schemes for each time period. In addition, optimization results often remain at the planning level, failing to effectively translate into real-time power adjustment commands for each device, leading to a disconnect between "optimization plan" and "actual execution." This prevents distributed energy devices from operating collaboratively according to the optimal plan, limiting the effectiveness of power fluctuation mitigation and voltage stability control.

[0004] The problems mentioned above overlap, resulting in existing distributed energy aggregation methods having overall chain-wide collaborative defects such as "data fragmentation, inaccurate prediction, one-sided optimization, and disconnected execution." This makes it difficult to achieve closed-loop aggregation of distributed energy from data perception to real-time adjustment, and fails to meet the actual engineering needs of the power grid for distributed energy aggregation in scenarios with a high proportion of renewable energy access, in terms of coordinated scheduling, economic efficiency, and safe and stable operation. Summary of the Invention

[0005] The purpose of this application is to provide a distributed energy aggregation method, device, equipment, medium, and product that can achieve high-precision wind and solar power output and load forecasting, accurate solution of multi-objective collaborative optimization, and generation of global optimal scheduling scheme. This significantly improves the level of new energy consumption and operational economy, decomposes the optimal output value of multiple time periods into real-time power adjustment commands for each device, and establishes a complete closed-loop control link of "prediction → optimization → adjustment". This enables each distributed energy device to operate collaboratively according to the optimal plan, realizes distributed energy aggregation, effectively smooths power fluctuations, and stabilizes voltage.

[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a distributed energy aggregation method, including: Acquire multi-source data of various types of distributed energy, and perform outlier removal, missing value imputation and normalization on the multi-source data to obtain standardized operating data with unified dimensions. The multi-source data includes equipment operating data, meteorological data and power grid status data. The standardized operating data is used to predict wind and solar power output and load using a hybrid model of long short-term memory network and gradient boosting decision tree, respectively, to obtain the predicted power output and load values ​​of new energy sources. The multi-objective optimization function is solved by improving the particle swarm optimization algorithm to obtain the optimal output value of each device in each time period. The multi-objective optimization function takes the predicted output of the new energy source as the input variable, the predicted load value as the constraint condition, and the minimization of daily operating cost as the main objective. Based on the optimal output value of each device in each time period, power regulation commands for power balance regulation and voltage reactive power control are generated.

[0007] Optionally, the equipment operation data includes photovoltaic array output, wind turbine output, energy storage state of charge, energy storage charging and discharging power, electric vehicle charging power, and micro gas turbine output; the meteorological data includes light intensity, wind speed, temperature, and relative humidity; and the grid status data includes three-phase voltage, three-phase current, active power, reactive power, grid frequency, and power quality harmonic content.

[0008] Optionally, the step of using a hybrid model of long short-term memory network and gradient boosting decision tree to predict wind and solar power output and load from the standardized operating data specifically includes: The standardized operating data is input into a hybrid model of a pre-trained Long Short-Term Memory network and a gradient boosting decision tree, and rolling predictions are performed to obtain the predicted output and load values ​​of new energy sources in the future time domain, respectively. Specifically, the wind and solar power output prediction uses the light intensity, wind speed, temperature, and historical output data from the standardized operating data as input features, and the predicted output of photovoltaic power and wind power as prediction targets. The load prediction uses historical active power data from the standardized operating data as input features, combined with temperature factors, for prediction. The hybrid model of the Long Short-Term Memory Network and Gradient Boosting Decision Tree is incrementally trained using newly added historical running data, and the model parameters are adaptively updated.

[0009] Optionally, the multi-objective optimization function is: ; In the formula, Indicates daily operating costs. Indicates the renewable energy consumption rate. Indicates power grid transmission loss. Indicates carbon emissions, , , and These represent the weight coefficients of each objective, and .

[0010] Optionally, the step of solving the multi-objective optimization function by improving the particle swarm optimization algorithm specifically includes: The improved particle swarm optimization algorithm employs a linearly decreasing inertial weight strategy to balance global and local search capabilities, dynamically adjusts the learning factor to focus on individual search in the early stage of iteration and global convergence in the later stage of iteration, and randomly mutates the positions of some particles when the algorithm gets stuck in a local optimum to escape the local optimum.

[0011] Optionally, the generation of power regulation commands for power balance regulation and voltage reactive power control specifically includes: Obtain the actual operating status of each adjustable device at the current moment, including the current output value, state of charge, and voltage level; Using the actual operating state as the initial condition, the load prediction value as the power balance benchmark, and the upper and lower limits of the output of each device in the optimal output value of each device in each time period as the adjustment boundary constraints, model predictive control is used to solve the optimal adjustment sequence that minimizes the power deviation in the future time domain. The current adjustment command is extracted from the optimal adjustment sequence and used as the power adjustment command. The actual operating state is then reacquired at the next moment to achieve rolling optimization control.

[0012] Secondly, this application provides a distributed energy aggregation device, comprising: The data acquisition and preprocessing module is used to acquire multi-source data of various types of distributed energy, and to perform outlier removal, missing value imputation and normalization processing on the multi-source data to obtain standardized operating data with unified dimensions. The multi-source data includes equipment operation data, meteorological data and power grid status data. The hybrid forecasting module is used to perform wind and solar power output forecasting and load forecasting on the standardized operating data through a hybrid model of long short-term memory network and gradient boosting decision tree, so as to obtain the predicted power output and load values ​​of new energy respectively. The multi-objective optimization module is used to solve the multi-objective optimization function by improving the particle swarm optimization algorithm to obtain the optimal output value of each device in each time period. The multi-objective optimization function takes the predicted output of the new energy source as the input variable, the predicted load value as the constraint condition, and the minimization of daily operating cost as the main objective. The scheduling and control module is used to generate power regulation commands for power balance regulation and voltage reactive power control based on the optimal output value of each device in each time period.

[0013] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the distributed energy aggregation method described in any one of the above.

[0014] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the distributed energy aggregation method described above.

[0015] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the distributed energy aggregation method described above.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method, apparatus, equipment, medium, and product for distributed energy aggregation. Through the steps of "acquiring multi-source data from various types of distributed energy and performing outlier removal, missing value imputation, and normalization to obtain standardized operational data with unified dimensions," it solves the technical problem of the inability to collaboratively utilize various types of distributed energy due to heterogeneous data formats, inconsistent dimensions, noise, and missing values. It achieves standardized fusion governance and high-quality storage of heterogeneous multi-source data, laying a unified and reliable data foundation for subsequent prediction and optimization. Through the steps of "using a hybrid model of Long Short-Term Memory Network and Gradient Boosting Decision Tree to predict wind and solar power output and load from the standardized operational data, obtaining predicted power output and load values ​​for new energy sources respectively" and "solving the multi-objective optimization function using an improved particle swarm optimization algorithm to obtain the optimal power output value of each device in each time period," it solves the problems of errors in traditional single prediction algorithms. The large discrepancy between the forecast and the target energy source leads to inaccurate scheduling, and single-objective optimization cannot simultaneously address the coordinated needs of economy, absorption rate, grid loss, and carbon emissions. Furthermore, conventional particle swarm optimization algorithms are prone to getting trapped in local optima. This approach achieves high-precision wind and solar power output and load forecasting, accurate solutions for multi-objective collaborative optimization, and the generation of globally optimal scheduling schemes, significantly improving the absorption rate of new energy and operational economy. By generating power regulation commands for power balance regulation and voltage reactive power control based on the optimal output values ​​of each device in each time period, the approach solves the technical problem that forecasting and optimization remain at the planning level and cannot be translated into actual executable adjustment actions by the devices. It decomposes the optimal output values ​​of multiple time periods into real-time power regulation commands for each device, establishing a complete closed-loop control link from "prediction → optimization → regulation." This enables distributed energy devices to operate collaboratively according to the optimal plan, achieving distributed energy aggregation, effectively smoothing power fluctuations, and stabilizing voltage. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a diagram illustrating the application environment of a distributed energy aggregation method according to an embodiment of this application. Figure 2 A schematic flowchart illustrating a distributed energy aggregation method provided in an embodiment of this application; Figure 3 A schematic diagram of the functional modules of a distributed energy aggregation device provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] The distributed energy aggregation method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, distributed energy device 102 communicates with edge computing gateway 104 via a network. The data storage system can store data that the edge computing gateway 104 needs to process, such as distributed energy device operation data, meteorological data, power grid status data, historical output data, prediction model parameters, and optimization results data. The data storage system can be set up separately, integrated into the edge computing gateway 104, or placed in the cloud or on other servers. Distributed energy device 102 can send its own operation data to edge computing gateway 104. After receiving the operation data from each distributed energy device, edge computing gateway 104, for connected distributed energy devices, preprocesses the multi-source data of various types of distributed energy, performs wind and solar power output prediction and load prediction using a hybrid model of long short-term memory network and gradient boosting decision tree, and solves the multi-objective optimization function using an improved particle swarm optimization algorithm to obtain the optimal output value of each device in each time period. Based on the optimal output value of each device in each time period, power regulation commands for power balance regulation and voltage reactive power control are generated. The edge computing gateway 104 can send the generated power adjustment commands to each distributed energy device 102 for execution. Furthermore, in some embodiments, the distributed energy aggregation method can also be implemented independently by the edge computing gateway 104 or a cloud server. For example, the edge computing gateway 104 can directly perform data collection, prediction, optimization, and scheduling control on the connected distributed energy devices. Alternatively, the cloud server can obtain multi-source data from each distributed energy device from the data storage system, perform prediction and optimization calculations, and then send the optimal output value to the edge computing gateway 104, which then generates and executes the power adjustment commands.

[0022] The distributed energy device 102 can be, but is not limited to, various photovoltaic power generation devices, wind power generation devices, energy storage devices, electric vehicle charging devices, and micro gas turbines. The edge computing gateway 104 can be a standalone industrial-grade edge computing gateway device, an integrated device combined with a power distribution automation terminal, or a virtual gateway deployed in the cloud, communicating with each distributed energy device via 5G or a wired network. The data storage system can be deployed independently, integrated into the edge computing gateway 104, or deployed in the cloud.

[0023] In one exemplary embodiment, such as Figure 2 As shown, a distributed energy aggregation method is provided. This method is executed by a computer device, specifically, it can be executed by a computer device such as an edge computing gateway or a cloud server alone, or it can be executed jointly by an edge computing gateway and a cloud server. In this embodiment, the method is applied to... Figure 1 Taking the edge computing gateway 104 as an example, the explanation includes the following steps 201 to 204. Wherein: Step 201: Obtain multi-source data of various types of distributed energy, and perform outlier removal, missing value imputation and normalization on the multi-source data to obtain standardized operation data with unified dimensions. The multi-source data includes equipment operation data, meteorological data and power grid status data. Specifically, after the edge computing gateway starts up, it first performs power-on initialization and device self-test to ensure that all hardware devices and software modules are working properly.

[0024] In one embodiment, the edge computing gateway is deployed at a designated location within the power distribution room. This gateway includes two Ethernet interfaces, four RS485 interfaces, one LoRa wireless interface, and one 5G communication module, ensuring compatibility with the communication needs of different types of distributed energy devices. It supports concurrent access for at least 200 devices, meeting the device access scale requirements of substation-level and park-level aggregation scenarios. The edge computing gateway integrates an ARM Cortex-A9 architecture processor with a main frequency of no less than 1GHz and at least 8GB of storage space for storing protocol parsing libraries, device information models, and local cache data. It also integrates a security encryption chip. During deployment, the edge computing gateway needs to establish communication links with the perception layer acquisition terminal and the scheduling layer server, using the TCP / IP protocol for data interaction. After deployment, communication link testing is performed to ensure smooth communication between the edge computing gateway and each device and server, with no packet loss or latency, and communication latency controlled within 50ms. The edge computing accesses the NTP unified time service to ensure its time is synchronized with all devices in the station, with a time error of no more than 1 second, providing a time reference for subsequent time-series data acquisition and scheduling command execution.

[0025] It should be noted that the edge computing gateway has a built-in multi-protocol parsing and conversion module, integrating parsing libraries for various mainstream power communication protocols. Upon startup, the edge computing gateway automatically activates broadcast scanning mode and receives proactive reporting signals from newly connected devices. It automatically identifies the device type based on the identification information and communication characteristics reported by the device. The edge computing gateway then calls its built-in multi-protocol parsing library to automatically identify the communication protocol type of the newly connected device. Protocol parsing and data format unification are completed without manual parameter configuration, converting the different formats of data output by heterogeneous devices into a common standard data format.

[0026] Specifically, a unique unified resource identifier is assigned to each newly connected device to ensure device uniqueness. Then, the core parameters of the device are automatically read, including device capacity, output regulation limits, response time, and scheduling priority, completing device self-registration and standardized modeling. Standardized modeling organizes the structured information of each connected device according to a unified energy data model, clarifying basic attributes such as device type, rated capacity, operating constraints, regulation capability, scheduling priority, and response speed. After device registration, a status check is performed to confirm normal communication and complete parameters, enabling rapid operation and scheduling of the device.

[0027] After power-on, a device self-test is performed. Specifically, the perception layer acquisition terminal, edge computing gateway, scheduling layer server, optimization layer server, application layer monitoring device, and time-series database are activated to avoid equipment damage or malfunction due to improper power-on sequence. A communication link connectivity test is performed to check the communication links between the perception layer acquisition terminal and the edge computing gateway, between the edge computing gateway and the server, and between the server and the monitoring device. Links with abnormal communication are automatically marked and alarms are issued. Device address configuration verification is performed to confirm that the IP addresses and device IDs of each acquisition terminal and device are configured correctly, with no address conflicts, ensuring the accuracy of data acquisition and command issuance. Model parameter loading is performed, loading the initial parameters of the power prediction model and optimization algorithm to the server to complete model initialization. After the self-test is completed, the device enters real-time operation mode. If an anomaly is detected during the self-test, fault information is automatically recorded, the fault location is identified, and alarms are pushed through multiple channels. The self-test procedure is re-executed after the fault is resolved until the self-test passes.

[0028] After entering operational mode, the sensing layer data acquisition program is started, synchronously collecting multi-dimensional data through sensors and acquisition terminals deployed at various monitoring points. The acquisition frequency is set to 1Hz, and the raw data sequence can be represented as follows: ,in Let be the raw data of the i-th collection point, and n be the total number of collection points.

[0029] The data collected synchronously includes: (1) Equipment operation data: photovoltaic array output Wind turbine output Energy storage state of charge (SOC), energy storage charge / discharge power Electric vehicle charging power Micro gas turbine output ; (2) Meteorological data: light intensity G, wind speed V, temperature T, relative humidity RH; (3) Power grid status data: three-phase voltage , , Three-phase current , , Active power P, reactive power Q, grid frequency f, and power quality harmonic content.

[0030] The collected raw data is processed by the preprocessing unit. Specifically: (1) Outlier removal: using The principle for identifying abnormal data is as follows: In the formula This is the mean of the data set. The standard deviation of this set of data is used to mark and remove outliers to prevent them from affecting subsequent analysis results.

[0031] (2) Missing value imputation: For missing data caused by communication interruption or equipment failure during data acquisition, linear interpolation is used to imput the missing data. The interpolation formula is as follows: In the formula For data missing at time t, , These are the valid data adjacent to the missing time points.

[0032] (3) Normalization: The min-max normalization method was used to map all data to the [0,1] interval to eliminate the influence of different dimensions on subsequent analysis. The normalization formula is as follows: In the formula The original data, The minimum value of this set of data. x is the maximum value of this set of data, and x_norm is the normalized data.

[0033] The preprocessed data is encrypted using the AES-256 encryption algorithm to prevent data leakage during transmission and storage. The encrypted data is then written to a time-series database for rapid querying and historical data backtracking. Simultaneously, the stored data undergoes periodic quality checks to verify its integrity. Acquisition terminals with a data missing rate exceeding 5% or data deviations exceeding allowable limits are automatically flagged and alerts are reported, notifying maintenance personnel to promptly investigate and resolve the faults, ensuring that data quality meets the needs of subsequent prediction, scheduling, and optimization.

[0034] Step 202: The standardized operating data is used to predict wind and solar power output and load using a hybrid model of long short-term memory network and gradient boosting decision tree, so as to obtain the predicted output and load values ​​of new energy sources, respectively. Specifically, in one embodiment, the prediction model employs a hybrid architecture of Long Short-Term Memory (LSTM) and Gradient Boosting Decision Tree (GBDT). The LSTM algorithm is used to capture long-term dependencies in time-series data, while the GBDT algorithm is used to capture nonlinear features. The fusion of the two can fully leverage their respective advantages to construct a high-precision wind and solar power output prediction model and load prediction model.

[0035] The wind and solar power output prediction model is constructed as follows: solar irradiance, wind speed, temperature, and historical power output data are used as input features, and predicted photovoltaic power output and predicted wind power output are used as output targets. A hybrid algorithm combining LSTM and GBDT is employed to build the prediction model. Cross-validation is used during model training to optimize model parameters and ensure the model's generalization ability.

[0036] The load forecasting model is constructed by using historical active power data as input features, combined with temperature factors, and employing a hybrid algorithm of LSTM and GBDT to build the forecasting model.

[0037] After the prediction model is built, standardized operational data is input into a pre-trained hybrid prediction model combining a Long Short-Term Memory (LSTM) network and a gradient boosting decision tree, and rolling predictions are performed. Rolling predictions are executed every preset time interval (e.g., 15 minutes) to obtain the predicted power output and load values ​​for future time domains. Power output prediction refers to forecasting the power generation of photovoltaic and wind power generation equipment in future periods; load prediction refers to forecasting the power demand of all electricity loads within the power supply area covered by the distributed energy aggregation system in future periods.

[0038] Prediction accuracy is quantified and evaluated using the Mean Absolute Percentage Error (MAPE), which is calculated using the following formula: ; In the formula, N is the number of predicted data points. To predict the output value, This represents the actual power output. The prediction accuracy must meet preset standards to ensure that the prediction results accurately reflect the fluctuation patterns and trends in new energy power output.

[0039] To address the impact of changes in weather conditions and equipment operating status on prediction accuracy, a hybrid model combining a Long Short-Term Memory network and a gradient boosting decision tree is incrementally trained using newly added historical operating data, with adaptive updates to the model parameters. Specifically, the prediction model is incrementally trained daily using newly added historical operating data, with adaptive updates to the model's weight parameters.

[0040] Step 203: Solve the multi-objective optimization function by improving the particle swarm optimization algorithm to obtain the optimal output value of each device in each time period. The multi-objective optimization function takes the predicted output of new energy as the input variable, the load prediction value as the constraint condition, and the minimization of daily operating cost as the main objective. Specifically, in one embodiment, the multi-objective optimization function takes the predicted output of new energy sources as the input variable, the predicted load value as the constraint condition, and the minimization of daily operating costs as the main objective.

[0041] First, construct a full lifecycle cost database. Comprehensively analyze and aggregate all costs incurred during the operation process, and establish a cost parameter matrix. Full lifecycle costs include: photovoltaic operation and maintenance costs. Energy storage cycle loss cost Fuel cost of micro gas turbines Grid interaction costs Environmental costs The formula for calculating total cost is as follows: This enables comprehensive quantification and detailed accounting of costs.

[0042] Secondly, a multi-objective optimization function is established. The primary objective is to minimize daily operating costs, while also considering the renewable energy integration rate. Power grid transmission loss Carbon emissions Construct a multi-objective optimization function: ; In the formula Indicates daily operating costs. Indicates the renewable energy consumption rate. E represents power grid transmission losses, and E represents carbon emissions. , , and These represent the weight coefficients of each objective, and Each weighting coefficient can be flexibly adjusted according to the needs of the actual operating scenario.

[0043] Then, an improved particle swarm optimization algorithm is used to solve the optimization problem. To improve the algorithm's solution accuracy and convergence speed, the standard particle swarm optimization algorithm is improved as follows: (1) Adopting a linearly decreasing inertia weight strategy: ; In the formula, This is the initial inertia weight, which can be 0.9. To terminate the inertia weight, a value of 0.4 can be used, where t is the current iteration number. This represents the maximum number of iterations. The algorithm's global search capability is balanced with its local search capability by linearly decreasing the inertia weight.

[0044] (2) Dynamically adjust learning factors: individual learning factors The global learning factor decreases from 2.5 to 1.5. The value was increased from 1.5 to 2.5, which made the algorithm focus on individual search in the early stage of iteration and on global convergence in the later stage of iteration.

[0045] (3) Introduce a mutation mechanism: When the algorithm gets stuck in a local optimum, the position of some particles is randomly mutated to help the algorithm get out of the local optimum and improve the solution accuracy.

[0046] In this embodiment, after solving the multi-objective optimization function using the improved Particle Swarm Optimization (PSO) algorithm, the output result is the optimal output value of each distributed energy device in each scheduling period. Taking a scheduling period of 15 minutes and a prediction period of 24 hours as an example, a single device outputs a total of 96 output values. The output values ​​of all devices in each period constitute a complete output allocation scheme, that is, the optimal output plan of each device in each period. Taking the energy storage system as an example, the PSO output result is the charging and discharging power value of each of the 96 periods (positive value is discharging, negative value is charging); taking the gas turbine as an example, the PSO output result is the generating power value of each of the 96 periods. After summing the optimal output values ​​of each device in each period, a complete scheduling timetable is formed.

[0047] The optimization layer also supports adaptive switching between three typical operating scenarios: (1) Grid-connected operation mode: With the core optimization goal of maximizing the absorption rate of new energy and minimizing the grid interaction cost, new energy output is given priority to be connected to the grid to reduce the grid purchase cost.

[0048] (2) Islanded microgrid mode: Prioritize the highest power supply reliability and the most stable power balance, and rationally allocate the output of each energy source to ensure continuous power supply to the load.

[0049] (3) Industrial park model: With optimal power quality and power supply continuity as the core objectives, the focus is on controlling voltage fluctuations and power surges to meet the power demand of precision equipment in industrial parks.

[0050] Step 204: Based on the optimal output value of each device in each time period, generate power regulation commands for power balance regulation and voltage reactive power control.

[0051] Specifically, in one embodiment, the actual operating status of each adjustable device at the current moment is first obtained, including the current output value, state of charge, and voltage level.

[0052] Using the current operating status of the equipment as the initial condition, the load forecast value as the power balance benchmark, and the upper and lower limits of the output of each equipment in each time period as the adjustment boundary constraints, Model Predictive Control (MPC) is used to solve for the optimal adjustment sequence that minimizes the power deviation in the future time domain.

[0053] The dynamic adjustment objective function of model predictive control is: ; In the formula, The output value or power balance target value required by the power grid dispatching command. To contribute practically, To adjust the amount, The smoothness weight is used to balance the output tracking accuracy and the smoothness adjustment.

[0054] After obtaining the optimal adjustment sequence, the adjustment command at the current moment is extracted from the optimal adjustment sequence as the power adjustment command and issued for execution. The power adjustment commands include the charging and discharging power commands of the energy storage system, the output adjustment commands of the micro gas turbine, the reactive power adjustment commands of the photovoltaic inverter, and the reactive power adjustment commands of the energy storage converter.

[0055] In the next moment, the actual operating status is re-acquired based on the measured data, and prediction and optimization are re-executed to achieve rolling optimization control.

[0056] The specific process of power balance regulation is as follows: real-time comparison of the predicted output of new energy sources with load demand. When the predicted output of photovoltaic power... Greater than the current load When the photovoltaic power output is predicted, priority is given to charging the energy storage system, and the charging power is rationally allocated according to the excess output to ensure that the excess power is effectively absorbed and to avoid power fluctuations affecting grid stability. With wind power forecast output The sum is less than the current load When necessary, the energy storage system will automatically discharge or the micro gas turbine will replenish energy to ensure that power balance requirements are met. The power balance equation is: In the formula The sum of all energy contributions This refers to power grid transmission losses.

[0057] The specific process of voltage reactive power control is as follows: real-time monitoring of the voltage U at key nodes of the distribution network, and precise voltage control achieved by adjusting the reactive power output Q of the photovoltaic inverter and energy storage converter to ensure that the node voltage meets the requirements. And voltage deviation , It sets the rated voltage for the power grid, improves power quality, and prevents grid faults caused by voltage exceeding limits.

[0058] Implementing steps 201 to 204 above solves the technical problems of the inability to coordinate the use of various types of distributed energy due to heterogeneous data formats, inconsistent dimensions, noise, and missing values. It achieves standardized fusion governance and high-quality storage of multi-source heterogeneous data, laying a unified and reliable data foundation for subsequent prediction and optimization. It also solves the technical problems of large errors in traditional single prediction algorithms leading to inaccurate scheduling, the inability of single-objective optimization to simultaneously consider the coordinated needs of economy, absorption rate, grid loss, and carbon emissions, and the tendency of conventional particle swarm optimization algorithms to get trapped in local optima. It achieves high-precision wind and solar power output and load prediction, accurate solutions for multi-objective collaborative optimization, and the generation of globally optimal scheduling schemes, significantly improving the level of new energy absorption and operational economy. Furthermore, it solves the technical problem that prediction and optimization remain at the planning level and cannot be translated into actual equipment-executable adjustment actions. It decomposes the optimal output values ​​of multiple time periods into real-time power adjustment commands for each device, opening up a complete closed-loop control link of "prediction → optimization → adjustment", enabling each distributed energy device to operate collaboratively according to the optimal plan, realizing distributed energy aggregation, effectively smoothing power fluctuations, and stabilizing voltage.

[0059] This application also provides an application scenario in which the above-described distributed energy aggregation method is applied. Specifically, the distributed energy aggregation method provided in this embodiment can be applied in the distributed photovoltaic aggregation dispatch scenario of a distribution network. Distributed energy enters the aggregation control link from decentralized access, and forms an adjustable and flexible resource aggregate through a predictive-optimized-adjusted collaborative approach, participating in the downstream grid collaborative dispatch process. The distributed energy aggregation method provided in this embodiment belongs to the collaborative dispatch and optimization stage in distributed energy aggregation control. Specifically, in the aggregation control process for distributed energy, the aggregate can be dispatched based on a collaborative approach of automatic collaborative optimization and manual supervision and intervention, that is, generating corresponding power adjustment commands for various distributed energy devices.

[0060] Based on the same inventive concept, this application also provides a distributed energy aggregation device for implementing the distributed energy aggregation method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the distributed energy aggregation device provided below can be found in the limitations of the distributed energy aggregation method described above, and will not be repeated here.

[0061] In one exemplary embodiment, such as Figure 3 As shown, a distributed energy aggregation device is provided, comprising: The data acquisition and preprocessing module is used to acquire multi-source data of various types of distributed energy, and to perform outlier removal, missing value imputation and normalization on the multi-source data to obtain standardized operating data with unified dimensions. The multi-source data includes equipment operation data, meteorological data and power grid status data. The hybrid forecasting module is used to forecast wind and solar power output and load using a hybrid model of long short-term memory network and gradient boosting decision tree, and obtain the forecasted power output and load values ​​of new energy sources, respectively. The multi-objective optimization module is used to solve the multi-objective optimization function by improving the particle swarm optimization algorithm, so as to obtain the optimal output value of each device in each time period. The multi-objective optimization function takes the predicted output of new energy as the input variable, the load prediction value as the constraint condition, and the minimization of daily operating cost as the main objective. The scheduling and control module is used to generate power regulation commands for power balance regulation and voltage reactive power control based on the optimal output value of each device in each time period.

[0062] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores distributed energy device operation data, meteorological data, grid status data, historical output data, predictive model parameters, and optimization results. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a distributed energy aggregation method.

[0063] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0064] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0065] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0066] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0067] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0068] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0069] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0070] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0071] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A distributed energy aggregation method, characterized in that, The distributed energy aggregation method includes: Acquire multi-source data of various types of distributed energy, and perform outlier removal, missing value imputation and normalization on the multi-source data to obtain standardized operating data with unified dimensions. The multi-source data includes equipment operating data, meteorological data and power grid status data. The standardized operating data is used to predict wind and solar power output and load using a hybrid model of long short-term memory network and gradient boosting decision tree, respectively, to obtain the predicted power output and load values ​​of new energy sources. The multi-objective optimization function is solved by improving the particle swarm optimization algorithm to obtain the optimal output value of each device in each time period. The multi-objective optimization function takes the predicted output of the new energy source as the input variable, the predicted load value as the constraint condition, and the minimization of daily operating cost as the main objective. Based on the optimal output value of each device in each time period, power regulation commands for power balance regulation and voltage reactive power control are generated.

2. The distributed energy aggregation method according to claim 1, characterized in that, The equipment operation data includes photovoltaic array output, wind turbine output, energy storage state of charge, energy storage charging and discharging power, electric vehicle charging power, and micro gas turbine output; the meteorological data includes light intensity, wind speed, temperature, and relative humidity; the power grid status data includes three-phase voltage, three-phase current, active power, reactive power, power grid frequency, and power quality harmonic content.

3. The distributed energy aggregation method according to claim 1, characterized in that, The method of using a hybrid model combining a long short-term memory network and a gradient boosting decision tree to predict wind and solar power output and load from the standardized operating data specifically includes: The standardized operating data is input into a hybrid model of a pre-trained Long Short-Term Memory network and a gradient boosting decision tree, and rolling predictions are performed to obtain the predicted output and load values ​​of new energy sources in the future time domain, respectively. Specifically, the wind and solar power output prediction uses the light intensity, wind speed, temperature, and historical output data from the standardized operating data as input features, and the predicted output of photovoltaic power and wind power as prediction targets. The load prediction uses historical active power data from the standardized operating data as input features, combined with temperature factors, for prediction. The hybrid model of the Long Short-Term Memory Network and Gradient Boosting Decision Tree is incrementally trained using newly added historical running data, and the model parameters are adaptively updated.

4. The distributed energy aggregation method according to claim 1, characterized in that, The multi-objective optimization function is: ； In the formula, Indicates daily operating costs. Indicates the renewable energy consumption rate. Indicates power grid transmission loss. Indicates carbon emissions, , , and These represent the weight coefficients of each objective, and .

5. The distributed energy aggregation method according to claim 1, characterized in that, The method of solving the multi-objective optimization function by improving the particle swarm optimization algorithm specifically includes: The improved particle swarm optimization algorithm employs a linearly decreasing inertial weight strategy to balance global and local search capabilities, dynamically adjusts the learning factor to focus on individual search in the early stage of iteration and global convergence in the later stage of iteration, and randomly mutates the positions of some particles when the algorithm gets stuck in a local optimum to escape the local optimum.

6. The distributed energy aggregation method according to claim 1, characterized in that, The generation of power regulation commands for power balance regulation and voltage reactive power control specifically includes: Obtain the actual operating status of each adjustable device at the current moment, including the current output value, state of charge, and voltage level; Using the actual operating state as the initial condition, the load prediction value as the power balance benchmark, and the upper and lower limits of the output of each device in the optimal output value of each device in each time period as the adjustment boundary constraints, model predictive control is used to solve the optimal adjustment sequence that minimizes the power deviation in the future time domain. The current adjustment command is extracted from the optimal adjustment sequence and used as the power adjustment command. The actual operating state is then reacquired at the next moment to achieve rolling optimization control.

7. A distributed energy aggregation device, characterized in that, The distributed energy aggregation device includes: The data acquisition and preprocessing module is used to acquire multi-source data of various types of distributed energy, and to perform outlier removal, missing value imputation and normalization processing on the multi-source data to obtain standardized operating data with unified dimensions. The multi-source data includes equipment operation data, meteorological data and power grid status data. The hybrid forecasting module is used to perform wind and solar power output forecasting and load forecasting on the standardized operating data through a hybrid model of long short-term memory network and gradient boosting decision tree, so as to obtain the predicted power output and load values ​​of new energy respectively. The multi-objective optimization module is used to solve the multi-objective optimization function by improving the particle swarm optimization algorithm to obtain the optimal output value of each device in each time period. The multi-objective optimization function takes the predicted output of the new energy source as the input variable, the predicted load value as the constraint condition, and the minimization of daily operating cost as the main objective. The scheduling and control module is used to generate power regulation commands for power balance regulation and voltage reactive power control based on the optimal output value of each device in each time period.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the distributed energy aggregation method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the distributed energy aggregation method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the distributed energy aggregation method according to any one of claims 1-6.

Images