A virtual oscillator auxiliary wind farm reactive power coordination system and method

Abstract

The application discloses a kind of virtual oscillator auxiliary wind farm reactive power coordination system and method, for electric power system automatic voltage control technical field, this virtual oscillator auxiliary wind farm reactive power coordination system includes: comprehensive prediction module, multi-objective optimization module, high volatility response module and low volatility management module.The application obtains the active power output data of wind farm in real time, calculates voltage volatility and frequency volatility index, and combines the reactive power demand and voltage stability requirement of power grid, utilizes the reactive power adjustment amount required by multi-objective optimization model dynamic calculation, can determine response priority according to the adjustment capacity of equipment, and formulates the reactive power adjustment task and control strategy of equipment, to ensure that power grid meets reactive power demand while maintaining voltage and frequency stability, improve the operation efficiency and stability of wind farm and power grid, optimize the allocation and management of resources.

Description

Technical Field

[0001] This invention relates to the field of automatic voltage control technology for power systems, and more particularly to a virtual oscillator-assisted reactive power coordination system and method for wind farms. Background Technology

[0002] Wind energy, as one of the new energy sources with the greatest commercial development and application value, boasts extremely abundant global reserves. Compared to traditional fossil fuels such as coal, oil, and natural gas, wind energy's greatest advantage lies in its renewability. Wind energy is generated by the movement of large amounts of air across the Earth's surface, thus, to some extent, it is an inexhaustible energy source. Furthermore, as a representative of renewable energy, wind energy has the potential for sustainable utilization and is environmentally friendly.

[0003] Currently, with rapid economic and technological development, the construction speed of wind farms often outpaces the development of regional power grids. When the active power generated by wind turbines is excessive, the grid interconnections typically fail to meet safety standards. Due to the fluctuating and random nature of wind turbine power output, coupled with insufficient reactive power and voltage control measures in the power grid, grid voltage quality issues become particularly prominent.

[0004] Virtual oscillator control technology, as a novel power electronic control method, has attracted increasing attention. Virtual oscillators can mimic the dynamic behavior of traditional synchronous generators in the power grid. By introducing the oscillator's control mechanism, power output can be regulated, enabling intelligent coordination of reactive power in wind turbine units.

[0005] Chinese Patent Publication No. CN117639125A discloses a method, device, electronic equipment, and storage medium for reactive power coordination in wind farms. This method proposes a reactive power coordination control method for wind farm groups when offshore wind power is transmitted via flexible DC transmission, considering fluctuations in wind power output. It ensures the rationality of the reactive power flow distribution in the wind farm's power collection system, even when the active power output of offshore wind farms transmitted via flexible DC transmission is subject to significant fluctuations due to wind speed variations. This reduces active power losses and reactive power compensation capacity in the power collection system, lowers the construction cost of offshore wind power projects, and fully leverages the ability of offshore wind farms to participate in grid reactive power control. This method is applicable to solving the reactive power coordination control problem for wind farm groups when offshore wind power is transmitted via flexible DC transmission, considering fluctuations in wind power output. It has high computational efficiency and can provide a scientific and reasonable solution for the reactive power-voltage stability analysis and control of power systems during large-scale offshore wind power transmission.

[0006] However, the traditional method is difficult to achieve fast response and dynamic adjustment when processing the active power output data of the wind farm and the real-time demand of the power grid, cannot meet the real-time demand of the reactive power regulation in the operation process of the power grid, and often cannot effectively evaluate the reactive power regulation capacity of each device, resulting in that the regulation potential of different devices cannot be fully utilized in the reactive power distribution, affecting the regulation efficiency and the overall stability of the power grid. In addition, many existing methods rely on centralized control, and it is difficult to realize automatic synchronization and collaborative control among devices, and the response is slow when the stability of the power grid is challenged. At the same time, in the allocation of the reactive power regulation task, the traditional technology cannot effectively optimize the allocation of resources, resulting in that some devices are overloaded and some devices are not fully utilized, affecting the overall operation efficiency of the wind farm and the power grid.

[0007] At present, there is no effective solution to the problems in the related art. SUMMARY

[0008] In order to overcome the above problems, the present application aims to provide a virtual oscillator assisted wind farm reactive coordination system and method, which aims to solve the problem that the traditional method is difficult to achieve fast response and dynamic adjustment when processing the active power output data of the wind farm and the real-time demand of the power grid, cannot meet the real-time demand of the reactive power regulation in the operation process of the power grid, and often cannot effectively evaluate the reactive power regulation capacity of each device, resulting in that the regulation potential of different devices cannot be fully utilized in the reactive power distribution, affecting the regulation efficiency and the overall stability of the power grid.

[0009] To this end, the specific technical solutions adopted by the present application are as follows:

[0010] According to one aspect of the present application, a virtual oscillator assisted wind farm reactive coordination system is provided, which comprises: a comprehensive prediction module, a multi-objective optimization module, a high volatility response module and a low volatility management module.

[0011] The comprehensive prediction module is used to comprehensively predict the multi-dimensional operation parameters of the wind farm and the power grid by using a multi-level neural network, and to evaluate the parameter operation state and power fluctuation trend in a future preset time.

[0012] The multi-objective optimization module is used to dynamically calculate the voltage volatility index and the frequency volatility index based on the evaluation result and the real-time active power output of the wind farm, and to calculate the required reactive power regulation amount by combining the reactive power demand and the voltage stability requirement through multi-objective optimization analysis, so as to determine the response priority and control strategy of each device.

[0013] a high volatility response module, when the voltage volatility index and the frequency volatility index exceed the preset threshold, the virtual oscillator will adjust according to the response priority of each device, and take corresponding control measures based on the control strategy to stabilize the voltage and frequency of the power grid;

[0014] a low volatility management module, when the voltage volatility index and the frequency volatility index do not exceed the preset threshold, the virtual oscillator controls the reactive power output of each device according to the real-time demand of the power grid; and through a decentralized control mechanism, automatic synchronization between devices is realized, while dynamic allocation of reactive power regulation tasks is realized, and the power output of each device is adaptively adjusted.

[0015] Optionally, the comprehensive prediction module includes a data collection unit, a feature selection unit, a model construction and training unit, a prediction and verification unit, and a comparative analysis unit;

[0016] The data collection unit is configured to collect multi-dimensional operating parameters from the historical database of the wind farm and the power grid;

[0017] The feature selection unit is configured to extract feature data associated with the operating state and power fluctuation from the collected multi-dimensional operating parameters using a feature selection algorithm;

[0018] The model construction and training unit is configured to construct a multi-level neural network model, input the extracted feature data, perform multi-round iteration training through a training set, and adjust the weights and biases of the multi-level neural network model using an optimization algorithm;

[0019] The prediction and verification unit is configured to use the trained multi-level neural network model to predict the operating state of the wind farm and the power grid in a future preset time, output the prediction results of voltage fluctuation, frequency fluctuation and power fluctuation, and verify the prediction results through a test set;

[0020] The comparative analysis unit is configured to compare and analyze the predicted voltage fluctuation, frequency fluctuation and power fluctuation with the historical operating data of the wind farm and the power grid, and evaluate the parameter operating state and power fluctuation trend in the future preset time.

[0021] Optionally, the feature selection unit includes the following when extracting feature data associated with the operating state and power fluctuation from the collected multi-dimensional operating parameters using a feature selection algorithm:

[0022] standardizing the collected multi-dimensional operating parameters;

[0023] calculating the covariance matrix of the standardized multi-dimensional operating parameters;

[0024] perform eigenvalue decomposition on the covariance matrix to extract eigenvalues and corresponding eigenvectors, wherein the eigenvalues represent the size of each principal component explaining the data variance, and the eigenvectors represent the direction of each principal component;

[0025] Sort the eigenvalues from high to low according to their size, and select the principal components that meet the preset proportion requirement as the extracted feature data;

[0026] Project the collected multi-dimensional operating parameters into the space formed by the selected principal components to obtain the reduced dimension feature data.

[0027] Optionally, the model construction and training unit includes the following when constructing a multi-level neural network model, inputting the extracted feature data, performing multiple rounds of iterative training on the training set, and adjusting the weights and biases of the multi-level neural network model using an optimization algorithm:

[0028] Divide the extracted feature data into a training set, a validation set, and a test set, and prepare the time series data required by the long short-term memory network and the tensor format required by the convolutional neural network;

[0029] Construct a multi-level neural network model containing a 1D convolutional layer, a max pooling layer, a long short-term memory network layer, and a fully connected layer to extract spatial features and temporal dependencies respectively;

[0030] Construct a multi-level neural network model, select mean squared error as the loss function, Adam optimizer as the optimizer, and set the evaluation indicators;

[0031] Use the training data to perform multiple rounds of iterative training, and in each round of iteration, calculate the loss and evaluation indicators based on the data of the validation set, and dynamically adjust the weights and biases of the multi-level neural network model;

[0032] Monitor the loss change of the validation set during the training process, and select the best model parameters according to the performance of the validation set;

[0033] Use the test set to evaluate the prediction performance of the model and calculate the final evaluation indicators.

[0034] Optionally, the comparative analysis unit includes the following when comparing and analyzing the predicted voltage fluctuations, frequency fluctuations, and power fluctuations with historical operating data of the wind farm and the power grid to evaluate the parameter operating state and power fluctuation trend in the future preset time:

[0035] Collect the voltage fluctuation, frequency fluctuation, and power fluctuation data predicted by the model, and match them with the historical operating data of the corresponding time period;

[0036] Use error indicators to evaluate the error between the prediction results and the historical data, and draw comparative analysis charts to show the matching degree and error trend of the prediction results and the historical data;

[0037] According to the calculated error index and the chart analysis result, the prediction accuracy of the model for voltage, frequency and power fluctuation is evaluated, and the prediction effect of the model at different time points and conditions is judged;

[0038] Based on the error analysis result, the parameter operation state and power fluctuation trend in the future preset time are evaluated.

[0039] Optionally, the multi-objective optimization module includes the following when dynamically calculating the voltage fluctuation index and the frequency fluctuation index based on the evaluation result and the real-time active power output of the wind farm, combining the reactive power demand and the voltage stability requirement, calculating the required reactive power regulation amount through multi-objective optimization analysis, and determining the response priority and control strategy of each device:

[0040] Obtaining real-time active power output data at the current time from the wind farm;

[0041] According to the real-time active power output data and the operation state of the wind farm, the current voltage fluctuation index and the frequency fluctuation index are calculated;

[0042] Obtaining the current reactive power demand and voltage stability requirement of the power grid, and minimizing the voltage fluctuation, minimizing the frequency fluctuation, and satisfying the power grid reactive power demand as the target, establishing a multi-objective optimization model;

[0043] Based on the multi-objective optimization model, the reactive power regulation amount that satisfies the voltage stability and frequency stability requirements is dynamically calculated; and according to the reactive power regulation amount and the regulation capacity of each type of device, the response priority of different devices is determined;

[0044] Based on the response priority of different devices, the control strategy of the device is formulated, the reactive power output of each device is adjusted, the real-time reactive power demand of the power grid is ensured, the voltage and frequency stability is maintained, and the wind farm operation efficiency is optimized.

[0045] Optionally, the expression with the target of minimizing the voltage fluctuation, minimizing the frequency fluctuation, and satisfying the power grid reactive power demand is:

[0046] ;

[0047] In the formula, The weighted sum of voltage fluctuation, frequency fluctuation and reactive power deviation in the multi-objective optimization model is represented;

[0048] The current reactive power demand of the power grid is represented;

[0049] The reactive power output provided by the wind farm is represented;

[0050] Represents the absolute value of the frequency deviation;

[0051] Represents the absolute value of the voltage deviation;

[0052] w 1 indicates the weighting factor for voltage fluctuations;

[0053] w 2 represents the weighting factor for frequency fluctuations;

[0054] w 3 represents the weighting factor for reactive power deviation.

[0055] Optionally, based on a multi-objective optimization model, the reactive power regulation amount that meets the requirements of voltage stability and frequency stability is dynamically calculated; and based on the reactive power regulation amount and the regulation capabilities of various equipment, the response priorities of different equipment are determined, including:

[0056] The optimal reactive power regulation that satisfies the requirements of voltage and frequency stability is calculated based on a multi-objective optimization model.

[0057] The reactive power regulation capacity of each device is calculated by the ratio of its actual reactive power output capacity to its rated output.

[0058] Assess the regulation capacity of each device based on its reactive power regulation capability.

[0059] Based on the reactive power regulation amount and the regulation capacity of each device, determine the response priority of each device, and formulate the reactive power regulation task for each device according to the order of response priority.

[0060] The expression for determining the response priority of each device is as follows:

[0061] ;

[0062] ;

[0063] In the formula, Indicates device i The actual reactive power output;

[0064] Indicates device i Rated reactive power output;

[0065] α i Indicates device i Reactive power regulation capability;

[0066] Indicates device i Response priority.

[0067] Optionally, the low volatility management module controls the reactive power output of each device according to the real-time demand of the power grid when the voltage volatility index and the frequency volatility index do not exceed the preset threshold; and automatically synchronizes between devices through a decentralized control mechanism, while dynamically allocating reactive power regulation tasks and adaptively adjusting the power output of each device, including:

[0068] Real-time monitoring of the voltage volatility index and the frequency volatility index of the power grid to determine whether they exceed the preset threshold;

[0069] When the voltage and frequency volatility indexes do not exceed the preset threshold, the virtual oscillator controls the reactive power output of each device according to the real-time demand of the power grid;

[0070] Using a decentralized control mechanism, each device is allowed to automatically adjust its output according to the control instructions of the virtual oscillator, achieving automatic synchronization between devices;

[0071] According to the real-time demand of the power grid, the reactive power regulation tasks are dynamically allocated to different devices;

[0072] Each device adaptively adjusts its reactive power output according to the real-time grid demand and its own operating state, ensuring that the voltage and frequency of the power grid remain stable when the demand changes;

[0073] Continuously monitor the output of the device and the response effect of the power grid, and dynamically adjust the reactive power output of each device according to the feedback.

[0074] According to another aspect of the present application, a virtual oscillator assisted wind farm reactive power coordination method is also provided, which comprises the following steps:

[0075] S1, using a multi-level neural network to comprehensively predict the multi-dimensional operating parameters of the wind farm and the power grid, to evaluate the parameter operating state and power fluctuation trend in the future preset time;

[0076] S2, based on the evaluation results and the real-time active power output of the wind farm, dynamically calculating the voltage volatility index and the frequency volatility index, and combining the reactive power demand and the voltage stability requirement, calculating the required reactive power regulation amount through multi-objective optimization analysis, determining the response priority and control strategy of each device;

[0077] S3, when the voltage volatility index and the frequency volatility index exceed the preset threshold, the virtual oscillator will adjust according to the response priority of each device, and take corresponding control measures based on the control strategy to stabilize the voltage and frequency of the power grid;

[0078] S4, when the voltage fluctuation index and the frequency fluctuation index do not exceed the preset threshold, the virtual oscillator controls the reactive power output of each device according to the real-time demand of the power grid, and realizes automatic synchronization between devices through a decentralized control mechanism, and dynamically allocates the reactive power regulation task, and adaptively adjusts the power output of each device.

[0079] Compared with the prior art, the present application has the following beneficial effects:

[0080] 1, the present application trains through multi-level neural network model, combines feature selection algorithm and historical data, effectively predicts future voltage fluctuation, frequency fluctuation and power fluctuation of wind farm and power grid, and evaluates the prediction accuracy of the model by comparing and analyzing historical data and prediction results, which can predict the future power grid operation state and power fluctuation trend in advance, thereby improving the operation stability and regulation efficiency of wind farm and power grid, and helping to optimize the decision and management of power system.

[0081] 2, the present application acquires the active power output data of wind farm in real time, calculates the voltage fluctuation index and the frequency fluctuation index, and combines the reactive power demand and voltage stability requirement of the power grid, and uses multi-objective optimization model to dynamically calculate the required reactive power regulation amount, which can determine the response priority according to the regulation capacity of the device, and formulate the reactive power regulation task and control strategy of the device, so as to ensure that the power grid meets the reactive power demand while maintaining the stability of voltage and frequency, improve the operation efficiency and stability of wind farm and power grid, and optimize the allocation and management of resources. BRIEF DESCRIPTION OF DRAWINGS

[0082] The above characteristics, features and advantages of the present application and its implementation and method become more apparent and understandable in combination with the following description of embodiments, which are described in detail in combination with the drawings. Herein is shown schematically:

[0083] Figure 1 is a principle block diagram of a virtual oscillator auxiliary wind farm reactive coordination system according to an embodiment of the present application;

[0084] Figure 2 is a flow chart of a virtual oscillator auxiliary wind farm reactive coordination method according to an embodiment of the present application.

[0085] In the drawings:

[0086] 1, comprehensive prediction module; 2, multi-objective optimization module; 3, high fluctuation response module; 4, low fluctuation management module. DETAILED DESCRIPTION

[0087] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0088] According to an embodiment of the present application, a virtual oscillator auxiliary wind farm reactive power coordination system and method are provided.

[0089] The present application will be further described in conjunction with the drawings and specific embodiments. As shown in the drawings, according to an embodiment of the present application, a virtual oscillator auxiliary wind farm reactive power coordination system is provided, which comprises a comprehensive prediction module 1, a multi-objective optimization module 2, a high volatility response module 3 and a low volatility management module 4. Figure 1

[0090] The comprehensive prediction module 1 is configured to use a multi-level neural network to comprehensively predict multi-dimensional operating parameters of the wind farm and the power grid, and to evaluate the parameter operating state and power fluctuation trend in a future preset time.

[0091] Preferably, the comprehensive prediction module comprises a data collection unit, a feature selection unit, a model construction and training unit, a prediction and verification unit and a comparative analysis unit.

[0092] The data collection unit is configured to collect multi-dimensional operating parameters from a historical database of the wind farm and the power grid.

[0093] The feature selection unit is configured to use a feature selection algorithm to extract feature data associated with the operating state and power fluctuation from the collected multi-dimensional operating parameters.

[0094] The model construction and training unit is configured to construct a multi-level neural network model, input the extracted feature data, perform multi-round iteration training through a training set, and adjust the weights and biases of the multi-level neural network model using an optimization algorithm.

[0095] The prediction and verification unit is configured to use the trained multi-level neural network model to predict the operating state of the wind farm and the power grid in a future preset time, output the prediction results of voltage fluctuation, frequency fluctuation and power fluctuation, and verify the prediction results through a test set.

[0096] The comparative analysis unit is configured to compare and analyze the predicted voltage fluctuation, frequency fluctuation and power fluctuation with the historical operating data of the wind farm and the power grid, and evaluate the parameter operating state and power fluctuation trend in a future preset time. ​

[0097] Preferably, the feature selection unit comprises the following when extracting feature data associated with the operating state and power fluctuation from the collected multi-dimensional operating parameters using a feature selection algorithm:

[0098] standardizing the collected multi-dimensional operating parameters;

[0099] calculating the covariance matrix of the standardized multi-dimensional operating parameters;

[0100] performing eigenvalue decomposition on the covariance matrix to extract eigenvalues and corresponding eigenvectors, wherein the eigenvalues represent the size of each principal component explaining the data variance, and the eigenvectors represent the direction of each principal component;

[0101] sorting the eigenvalues from high to low according to their size, and selecting principal components that meet the preset proportion requirement as extracted feature data;

[0102] projecting the collected multi-dimensional operating parameters into the space formed by the selected principal components to obtain the reduced dimension feature data.

[0103] Preferably, the model construction and training unit comprises the following when constructing a multi-level neural network model, inputting the extracted feature data, performing multi-round iteration training through the training set, and adjusting the weights and biases of the multi-level neural network model using an optimization algorithm:

[0104] dividing the extracted feature data into a training set, a validation set, and a test set, and preparing time series data required by a long short-term memory network and tensor format required by a convolutional neural network;

[0105] constructing a multi-level neural network model containing a 1D convolutional layer, a max-pooling layer, a long short-term memory network layer, and a fully connected layer to extract spatial features and temporal dependencies respectively;

[0106] constructing a multi-level neural network model, selecting mean squared error as the loss function, Adam optimizer as the optimizer, and setting the evaluation index;

[0107] performing multi-round iteration training using the training data, in each round of iteration, calculating the loss and evaluation index based on the data of the validation set, and dynamically adjusting the weights and biases of the multi-level neural network model;

[0108] monitoring the loss change of the validation set during the training process, and selecting the best model parameters according to the performance of the validation set;

[0109] evaluating the prediction performance of the model using the test set, and calculating the final evaluation index.

[0110] ​​Preferably, the comparative analysis unit includes the following when comparing the predicted voltage fluctuations, frequency fluctuations and power fluctuations with historical operation data of the wind farm and the power grid to assess the parameter operation state and power fluctuation trend in the future preset time:

[0111] Collecting the model-predicted voltage fluctuations, frequency fluctuations and power fluctuations data, and matching them with the historical operation data of the corresponding time period;

[0112] Using error indicators to evaluate the error between the prediction results and the historical data, and drawing comparative analysis charts to show the matching degree and error trend of the prediction results and the historical data;

[0113] According to the calculated error indicators and chart analysis results, evaluate the prediction accuracy of the model for voltage, frequency and power fluctuations, and judge the prediction effect of the model under different time points and conditions;

[0114] Based on the error analysis results, assess the parameter operation state and power fluctuation trend in the future preset time.

[0115] It needs to be explained that first, the data collection unit obtains multi-dimensional operation data from the historical database, and the feature selection unit extracts key features related to power fluctuations through feature selection algorithms. Then, the model construction and training unit constructs and trains a multi-level neural network model, adjusts the weights and biases using optimization algorithms, and performs multiple rounds of iterative optimization through the validation set. The trained model is used to predict future voltage, frequency and power fluctuations, and the prediction results are compared with historical data to evaluate the accuracy and future trend of the model.

[0116] The multi-objective optimization module 2 is used to dynamically calculate the voltage fluctuation index and frequency fluctuation index based on the evaluation results and the real-time active power output of the wind farm, and to calculate the required reactive power regulation amount through multi-objective optimization analysis combined with the reactive power demand and voltage stability requirements, to determine the response priority and control strategy of each device.

[0117] Among them, the control strategy specifically refers to the specific operation scheme for managing and regulating the reactive power output of each device, aiming to ensure the voltage and frequency stability of the power grid and meet the reactive power demand. It mainly includes the following aspects:

[0118] 1. Priority allocation of reactive power regulation: according to the response priority of the device, determine which devices should be prioritized for reactive power regulation. The response priority is usually determined based on the adjustment capacity, real-time state and grid demand of the device, and the devices with fast response and strong adjustment capacity are prioritized.

[0119] 2. Reactive power output control: For each selected device, the control strategy defines its specific reactive power output. This involves adjusting the reactive power output of the device, ensuring that the total reactive power meets the grid's requirements, while optimizing voltage and frequency stability.

[0120] 3. Decentralized control and automatic synchronization: Through a decentralized control mechanism, each device independently adjusts its reactive power output, avoiding delays and single-point failures associated with centralized control. Additionally, the control strategy ensures that devices automatically synchronize their power outputs, maintaining coordination and avoiding unnecessary power conflicts or interference.

[0121] 4. Dynamic adjustment and feedback mechanism: The control strategy continuously monitors the real-time operating state of the grid and dynamically adjusts the output of each device. When the grid conditions or device status change, the control strategy updates the output instructions for each device in real-time, ensuring the stability of the grid is not affected.

[0122] 5. Emergency response strategy: When voltage or frequency fluctuations exceed the pre-set threshold, the control strategy triggers emergency response measures, such as rapidly increasing or decreasing reactive power output, ensuring that the grid can quickly recover stability in extreme situations.

[0123] Preferably, the multi-objective optimization module includes the following when calculating the required reactive power regulation amount, determining the response priority of each device, and formulating the control strategy based on the evaluation results and real-time active power output of the wind farm, voltage fluctuation index, and frequency fluctuation index, as well as reactive power demand and voltage stability requirements:

[0124] Obtain real-time active power output data at the current time from the wind farm;

[0125] Calculate the current voltage fluctuation index and frequency fluctuation index based on the real-time active power output data and the operating state of the wind farm;

[0126] Obtain the current reactive power demand and voltage stability requirements of the grid, and establish a multi-objective optimization model with the objectives of minimizing voltage fluctuation, minimizing frequency fluctuation, and meeting the grid's reactive power demand;

[0127] Based on the multi-objective optimization model, dynamically calculate the reactive power regulation amount that meets the voltage stability and frequency stability requirements; and based on the reactive power regulation amount and the regulation capacity of each type of device, determine the response priority of different devices;

[0128] Based on the response priority of different devices, formulate the control strategy for each device to adjust the reactive power output of each device, ensuring that the real-time reactive power demand of the grid is met, the voltage and frequency are stable, and the wind farm operating efficiency is optimized.

[0129] Preferably, the expression aiming at minimizing voltage fluctuation, minimizing frequency fluctuation and meeting the reactive power demand of the power grid is:

[0130] ;

[0131] wherein, represents the weighted sum of voltage fluctuation, frequency fluctuation and reactive power deviation in the multi-objective optimization model;

[0132] represents the current reactive power demand of the power grid;

[0133] represents the reactive power output provided by the wind farm;

[0134] represents the absolute value of frequency deviation;

[0135] represents the absolute value of voltage deviation;

[0136] w 1 represents the weight factor of voltage fluctuation;

[0137] w 2 represents the weight factor of frequency fluctuation;

[0138] w 3 represents the weight factor of reactive power deviation.

[0139] Preferably, based on the multi-objective optimization model, the reactive power regulation amount meeting the voltage stability and frequency stability requirements is dynamically calculated; and according to the reactive power regulation amount and the regulation capacity of each type of device, the response priority of different devices is determined, including:

[0140] The optimal reactive power regulation amount meeting the voltage and frequency stability requirements is calculated based on the multi-objective optimization model;

[0141] The reactive power regulation capacity of each device is calculated by the ratio of the actual reactive power output capacity of the device to its rated output;

[0142] The regulation capacity of each device is evaluated according to the reactive power regulation capacity of each device;

[0143] According to the reactive power regulation amount and the regulation capacity of each device, the response priority of each device is determined, and the reactive power regulation task of each device is formulated in the order of the response priority;

[0144] wherein, the expression for determining the response priority of each device is:

[0145] ;

[0146] ;

[0147] wherein, represents the actual reactive power output of the device i ;

[0148] represents the rated reactive power output of the device i ;

[0149] α i represents the reactive power regulation capability of the device i ;

[0150] represents the response priority of the device i .

[0151] It needs to be explained that the above describes the workflow of a multi-objective optimization module for dynamically calculating voltage fluctuation and frequency fluctuation indicators, and determining the reactive power regulation amount and the response priority of each device through optimization analysis. First, based on the real-time active power output of the wind farm and the evaluation results, the current voltage and frequency fluctuation indicators are calculated, then, combined with the reactive power demand of the power grid and the voltage stability requirement, a multi-objective optimization model is established to minimize voltage, frequency fluctuation and reactive power deviation. Through the optimization model, the required reactive power regulation amount is calculated, and the response priority of each device is determined according to its reactive power regulation capability, and the device control strategy is formulated to ensure the stability of the power grid and optimize the operation efficiency of the wind farm.

[0152] High fluctuation response module 3, for when the voltage fluctuation indicator and the frequency fluctuation indicator exceed the preset threshold, the virtual oscillator will adjust according to the response priority of each device, and take corresponding control measures based on the control strategy to stabilize the voltage and frequency of the power grid.

[0153] It needs to be explained that when the voltage and frequency fluctuation indicators exceed the preset safety threshold, the virtual oscillator will adjust according to the response priority of each device, and take corresponding control measures according to the pre-set control strategy to quickly stabilize the voltage and frequency of the power grid, and ensure the stable operation of the power grid under high fluctuation conditions.

[0154] Low fluctuation management module 4, for when the voltage fluctuation indicator and the frequency fluctuation indicator do not exceed the preset threshold, the virtual oscillator controls the reactive power output of each device according to the real-time demand of the power grid; and through the decentralized control mechanism, automatic synchronization between devices is realized, and at the same time, the reactive power regulation task is dynamically allocated, and the power output of each device is adaptively adjusted.

[0155] Preferably, the low volatility management module controls the reactive power output of each device according to the real-time demand of the power grid when the voltage fluctuation index and the frequency fluctuation index do not exceed the preset threshold; and realizes automatic synchronization between devices through a decentralized control mechanism, while dynamically allocating reactive power regulation tasks and adaptively adjusting the power output of each device, including:

[0156] Real-time monitoring of the voltage fluctuation index and the frequency fluctuation index of the power grid to determine whether they exceed the preset threshold;

[0157] When the voltage and frequency fluctuation indexes do not exceed the preset threshold, the virtual oscillator controls the reactive power output of each device according to the real-time demand of the power grid;

[0158] Using a decentralized control mechanism, each device is allowed to automatically adjust its output according to the control instructions of the virtual oscillator, realizing automatic synchronization between devices;

[0159] According to the real-time demand of the power grid, the reactive power regulation tasks are dynamically allocated to different devices;

[0160] Each device adaptively adjusts its reactive power output according to the real-time grid demand and its own operating state, ensuring that the voltage and frequency of the power grid remain stable when the demand changes;

[0161] Continuously monitor the output of the device and the response of the power grid, and dynamically adjust the reactive power output of each device according to the feedback.

[0162] It needs to be explained that when the voltage and frequency fluctuation indexes do not exceed the preset threshold, the virtual oscillator controls the reactive power output of each device according to the real-time demand of the power grid. Through a decentralized control mechanism, each device automatically adjusts its output and remains synchronized, dynamically allocates reactive power regulation tasks according to real-time demand, and each device adaptively adjusts its power output to ensure the stability of the power grid. At the same time, the output of the device and the response of the power grid are continuously monitored, and the reactive power output is dynamically adjusted according to the feedback to ensure the stability of the voltage and frequency.

[0163] According to another embodiment of the present application, as Figure 2 shown, a virtual oscillator assisted wind farm reactive power coordination method is also provided, which includes the following steps:

[0164] S1, using a multi-level neural network to comprehensively predict the multi-dimensional operating parameters of the wind farm and the power grid, to evaluate the parameter operating state and power fluctuation trend in the future preset time;

[0165] S2, based on the evaluation result and the real-time active power output of the wind farm, dynamically calculating the voltage fluctuation index and the frequency fluctuation index, combining the reactive power demand and the voltage stability requirement, calculating the required reactive power regulation amount through multi-objective optimization analysis, determining the response priority and control strategy of each device;

[0166] S3, when the voltage fluctuation index and the frequency fluctuation index exceed the preset threshold, the virtual oscillator will adjust according to the response priority of each device, and take corresponding control measures based on the control strategy to stabilize the voltage and frequency of the power grid;

[0167] S4, when the voltage fluctuation index and the frequency fluctuation index do not exceed the preset threshold, the virtual oscillator controls the reactive power output of each device according to the real-time demand of the power grid; and through the decentralized control mechanism, automatic synchronization between devices is realized, and the reactive power regulation task is dynamically allocated, and the power output of each device is adaptively adjusted.

[0168] To sum up, by means of the above technical scheme of the present application, the present application effectively predicts the future voltage fluctuation, frequency fluctuation and power fluctuation of the wind farm and the power grid by training through a multi-level neural network model combined with a feature selection algorithm and historical data, and by comparing and analyzing the historical data and the prediction results, the prediction accuracy of the model is evaluated, the future power grid operation state and power fluctuation trend can be predicted in advance, thereby improving the operation stability and regulation efficiency of the wind farm and the power grid, and helping to optimize the decision and management of the power system; the present application can calculate the voltage fluctuation index and the frequency fluctuation index by real-time acquisition of the active power output data of the wind farm, combine the reactive power demand and the voltage stability requirement of the power grid, and dynamically calculate the required reactive power regulation amount by using a multi-objective optimization model, so as to determine the response priority according to the regulation capacity of the device, and formulate the reactive power regulation task and control strategy of the device, thereby ensuring that the power grid meets the reactive power demand while maintaining the stability of voltage and frequency, improving the operation efficiency and stability of the wind farm and the power grid, and optimizing the allocation and management of resources.

[0169] Although the present application has been disclosed as above with preferred embodiments, the embodiments are only for illustration and do not limit the present application, and those skilled in the art can make some changes and modifications without departing from the spirit and scope of the present application, and the protection scope claimed by the present application should be subject to the description of the claims.

Claims

1. A virtual oscillator assisted wind farm reactive power coordination system, characterized by, The virtual oscillator auxiliary wind farm reactive power coordination system comprises: a comprehensive prediction module, a multi-objective optimization module, a high volatility response module and a low volatility management module; The comprehensive prediction module is configured to utilize a multi-level neural network to comprehensively predict multi-dimensional operation parameters of the wind farm and the power grid, and to evaluate parameter operation states and power fluctuation trends in a future preset time; The multi-objective optimization module is configured to dynamically calculate voltage volatility indexes and frequency volatility indexes based on the evaluation results and real-time active power output of the wind farm, and to calculate required reactive power regulation amounts by multi-objective optimization analysis in combination with reactive power demands and voltage stability requirements, to determine response priorities and control strategies of the devices, and specifically comprises: acquiring real-time active power output data at a current time from the wind farm; calculating current voltage volatility indexes and frequency volatility indexes according to the real-time active power output data and the operation state of the wind farm; acquiring current reactive power demands and voltage stability requirements of the power grid, and establishing a multi-objective optimization model with the objectives of minimizing voltage volatility, minimizing frequency volatility and meeting the reactive power demands of the power grid; the expression of the objectives of minimizing voltage volatility, minimizing frequency volatility and meeting the reactive power demands of the power grid is: ； In the formula, denotes the weighted sum of voltage fluctuations, frequency fluctuations and reactive power deviations in the multi-objective optimization model; represents the current reactive power demand of the power grid; Q represents the reactive power output provided by the wind farm; denotes the absolute value of the frequency deviation; represents the absolute value of the voltage deviation; w 1 represents a weight factor for voltage fluctuations; w 2 represents a weight factor for the frequency fluctuation; w 3 represents a weight factor for the reactive power deviation; based on the multi-objective optimization model, dynamically calculating reactive power regulation amounts that meet voltage stability and frequency stability requirements; and determining response priorities of different devices according to the reactive power regulation amounts and regulation capacities of the devices; based on the response priorities of different devices, formulating control strategies of the devices, regulating reactive power outputs of the devices, and ensuring that the reactive power demands of the power grid are met in real time, voltage and frequency are stabilized, and wind farm operation efficiency is optimized; the high volatility response module is configured to, when the voltage volatility indexes and the frequency volatility indexes exceed preset threshold values, adjust the virtual oscillator according to the response priorities of the devices, and take corresponding control measures based on the control strategies to stabilize voltage and frequency of the power grid; the low volatility management module is configured to, when the voltage volatility indexes and the frequency volatility indexes do not exceed the preset threshold values, control reactive power outputs of the devices according to real-time demands of the power grid; and realize automatic synchronization among the devices through a decentralized control mechanism, dynamically allocate reactive power regulation tasks, and adaptively adjust power outputs of the devices.

2. A virtual oscillator assisted wind farm reactive power coordination system according to claim 1, characterized in that, The comprehensive prediction module comprises a data collection unit, a feature selection unit, a model construction and training unit, a prediction and verification unit, and a comparative analysis unit; The data collection unit is configured to collect multi-dimensional operation parameters from historical databases of the wind farm and the power grid; The feature selection unit is configured to utilize a feature selection algorithm to extract feature data associated with operation states and power fluctuations from the collected multi-dimensional operation parameters; The model construction and training unit is configured to construct a multi-level neural network model, input the extracted feature data, perform multi-round iteration training through a training set, and adjust weights and biases of the multi-level neural network model by utilizing an optimization algorithm; The prediction and verification unit is configured to utilize the trained multi-level neural network model to predict the wind farm and power grid operation states in a future preset time, output prediction results of voltage fluctuation, frequency fluctuation and power fluctuation, and verify the prediction results by using a test set; The comparative analysis unit is configured to compare and analyze the predicted voltage fluctuation, frequency fluctuation and power fluctuation with historical operation data of the wind farm and power grid, and evaluate the parameter operation states and power fluctuation trend in the future preset time.

3. A virtual oscillator auxiliary wind farm reactive power coordination system according to claim 2, characterized in that, The feature selection unit includes the following steps when extracting feature data associated with the operation states and power fluctuation from the collected multi-dimensional operation parameters by using a feature selection algorithm: standardizing the collected multi-dimensional operation parameters; calculating the covariance matrix of the standardized multi-dimensional operation parameters; performing eigenvalue decomposition on the covariance matrix to extract eigenvalues and corresponding eigenvectors, wherein the eigenvalues represent the size of the data variance explained by each principal component, and the eigenvectors represent the direction of each principal component; sorting the eigenvalues from high to low according to their size, and selecting principal components that meet the preset proportion requirement as the extracted feature data; projecting the collected multi-dimensional operation parameters into the space formed by the selected principal components to obtain the reduced feature data.

4. A virtual oscillator auxiliary wind farm reactive power coordination system according to claim 3, characterized in that, The model construction and training unit includes the following steps when constructing a multi-level neural network model, inputting the extracted feature data, performing multi-round iterative training by using a training set, and adjusting the weights and biases of the multi-level neural network model by using an optimization algorithm: dividing the extracted feature data into a training set, a verification set and a test set, and preparing time series data required by a long short-term memory network and tensor format required by a convolutional neural network; constructing a multi-level neural network model including a 1D convolutional layer, a max-pooling layer, a long short-term memory network layer and a fully connected layer, and extracting spatial features and time-dependent relationships respectively; constructing a multi-level neural network model, selecting mean square error as a loss function, Adam optimizer as an optimizer, and setting evaluation indicators; performing multi-round iterative training by using training data, calculating loss and evaluation indicators based on the data of the verification set in each round of iteration, and dynamically adjusting the weights and biases of the multi-level neural network model; monitoring the loss change of the verification set during the training process, and selecting the best model parameters according to the performance of the verification set; evaluating the prediction performance of the model by using the test set, and calculating the final evaluation indicators.

5. A virtual oscillator auxiliary wind farm reactive power coordination system according to claim 4, characterized in that, The comparative analysis unit includes the following steps when comparing and analyzing the predicted voltage fluctuation, frequency fluctuation and power fluctuation with historical operation data of the wind farm and power grid, and evaluating the parameter operation states and power fluctuation trend in the future preset time: collecting voltage fluctuation, frequency fluctuation and power fluctuation data predicted by the model, and matching them with historical operation data of the corresponding time period; evaluating the error between the prediction results and the historical data by using error indicators, and drawing comparative analysis charts to show the matching degree and error trend of the prediction results and the historical data; evaluating the prediction accuracy of the model for voltage, frequency and power fluctuation according to the calculated error indicators and chart analysis results, and judging the prediction effect of the model under different time points and conditions; Based on the error analysis results, the parameter running state and power fluctuation trend in the future preset time are evaluated.

6. A virtual oscillator auxiliary wind farm reactive power coordination system according to claim 1, characterized in that, The reactive power regulation amount meeting the requirements of voltage stability and frequency stability is dynamically calculated based on the multi-objective optimization model; And according to the reactive power regulation amount and the regulation capacity of various devices, the response priority of different devices is determined, including: The optimal reactive power regulation amount meeting the requirements of voltage and frequency stability is calculated based on the multi-objective optimization model; The reactive power regulation capacity of each device is calculated through the ratio of the actual reactive power output capacity of the device to its rated output; The regulation capacity of each device is evaluated according to its reactive power regulation capacity; According to the reactive power regulation amount and the regulation capacity of various devices, the response priority of each device is determined, and the reactive power regulation task of each device is formulated in the order of response priority; The expression for determining the response priority of each device is: ； ； In the formula, representing the actual reactive output of the device i ; representing device i of the rated reactive output; α i representing device i reactive power regulation capability; representing the response priority of the device i of the device.

7. A virtual oscillator assisted wind farm reactive power coordination system according to claim 1, wherein, When the voltage fluctuation index and the frequency fluctuation index do not exceed the preset threshold, the virtual oscillator controls the reactive power output of each device according to the real-time demand of the power grid, and the automatic synchronization between devices is realized through a decentralized control mechanism, while dynamically allocating reactive power regulation tasks and adaptively adjusting the power output of each device, including: Real-time monitoring of the voltage fluctuation index and the frequency fluctuation index of the power grid to determine whether they exceed the preset threshold; When the voltage and frequency fluctuation indexes do not exceed the preset threshold, the virtual oscillator controls the reactive power output of each device according to the real-time demand of the power grid; Using a decentralized control mechanism, each device is allowed to automatically adjust its output according to the control instructions of the virtual oscillator, realizing automatic synchronization between devices; According to the real-time demand of the power grid, reactive power regulation tasks are dynamically allocated to different devices; Each device adaptively adjusts its reactive power output according to the real-time power grid demand and its own running state, ensuring the voltage and frequency stability of the power grid when the demand changes; Continuously monitor the output of the device and the response effect of the power grid, and dynamically adjust the reactive power output of each device according to the feedback. The virtual oscillator auxiliary wind farm reactive power coordination method includes the following steps:

8. A virtual oscillator auxiliary wind farm reactive power coordination method for implementing the virtual oscillator auxiliary wind farm reactive power coordination system of any one of claims 1-7, characterized in that, S1, using a multi-level neural network to comprehensively predict the multi-dimensional running parameters of the wind farm and the power grid, and evaluate the parameter running state and power fluctuation trend in the future preset time; S2, based on the evaluation results and the real-time active power output of the wind farm, dynamically calculate the voltage fluctuation index and the frequency fluctuation index, and combine the reactive power demand and the voltage stability requirement to calculate the required reactive power regulation amount through multi-objective optimization analysis, determine the response priority and control strategy of each device; S3, when the voltage fluctuation index and the frequency fluctuation index exceed the preset threshold, the virtual oscillator will adjust according to the response priority of each device, and take corresponding control measures based on the control strategy to stabilize the voltage and frequency of the power grid; S4, when the voltage fluctuation index and the frequency fluctuation index do not exceed the preset threshold, the virtual oscillator controls the reactive power output of each device according to the real-time demand of the power grid; and through a decentralized control mechanism, the automatic synchronization between devices is realized, while dynamically allocating reactive power regulation tasks and adaptively adjusting the power output of each device. ​

Images