Method and system for determining reactive output power of wind farm with double-fed unit

Abstract

The present invention discloses a method and system for determining the reactive output power of a wind farm of a doubly-fed unit, and the present invention relates to the field of power system operation and control technology. The method comprises the following steps: collecting historical time series data of the wind farm, and recording the reactive output power value at the corresponding timestamp; dividing the wind farm scene into different wind speed load scenes according to the historical time series data, and calculating the estimated target value of the reactive output power of the wind farm under different scenes; establishing a deep learning model, taking the historical time series data as a sample, and training the model with the corresponding reactive output power value as a label; collecting real-time characteristic data of the wind farm, judging the current wind farm scene type, and calculating the current power estimated target value; obtaining a power prediction value according to a power prediction model, adjusting the power prediction value, and obtaining an accurate value of reactive output power; outputting an accurate value of reactive output power in accordance with the reactive power adjustment range of the doubly-fed wind turbine generator and the energy storage system in accordance with the adjustment range.

Description

Technical Field

[0001] The present invention relates to the technical field of power system operation and control, and in particular to a method and system for determining reactive output power of a wind farm of a doubly-fed unit. Background Art

[0002] With the global emphasis on renewable energy, wind energy, as an important component of clean energy, has made significant progress in its development and utilization. In particular, the proportion of wind power in the power system has continued to rise, becoming the mainstream form of green energy. However, the volatility and instability of wind energy have brought new challenges to the operation of the power system, especially in the management and regulation of reactive power.

[0003] Doubly-fed wind turbines (DFIGs) have become one of the core equipment in wind farms due to their advantages in operating efficiency and control flexibility. Through doubly-fed generator technology, DFIGs can maintain optimal power generation efficiency at different wind speeds while achieving flexible regulation of active and reactive power. However, although DFIGs have certain capabilities in regulating reactive power, existing control strategies still face some limitations.

[0004] Existing reactive power control strategies often rely on pre-set control parameters and lack a real-time feedback mechanism. This means that when the grid load changes greatly or the wind speed fluctuates significantly, the unit may not be able to respond quickly, resulting in insufficient reactive power, which in turn affects the voltage quality of the grid. This situation not only increases the instability of the grid, but may also cause equipment damage and system failures, bringing greater risks to power operations. In order to address the above problems, it is urgent to develop an efficient reactive power output determination method to flexibly adjust the reactive output of the doubly fed unit. This enables the wind turbine to maintain the best operating state under different load and wind speed conditions, and effectively stabilize the grid voltage and improve the power quality.

[0005] In the prior art, publication number CN105244894A discloses a method for determining the reactive output power of a doubly-fed wind farm, comprising the following steps: S1, establishing a set of typical scenarios for the active output of a doubly-fed wind farm based on the prediction results of the active power of the wind farm, and calculating the probability of occurrence of each typical scenario; S2, calculating the effective reactive regulation capacity of the doubly-fed wind farm under each typical scenario; S3, calculating the reactive regulation capacity of the synchronous generator under each typical scenario of the wind farm; S4, establishing a reactive optimization model that takes into account the reactive regulation capacity of the doubly-fed wind farm and the AVR regulation function of the synchronous generator; S5, determining the optimal reactive output power of the wind farm using a joint solution strategy of an interior point method-genetic algorithm based on relaxation iteration. This method needs to consider too many factors when establishing a reactive power optimization model, which may lead to an overly complex model, increased difficulty in solving, long calculation time, and even convergence problems. At the same time, a joint solution strategy of the interior point method-genetic algorithm based on relaxation iteration is adopted. Although the genetic algorithm has a certain global search capability, it has a long convergence time. Therefore, the use of complex algorithms and modeling not only wastes resources, but also reduces the real-time and effectiveness of the determination method.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The object of the present invention is to provide a method and system for determining reactive output power of a doubly-fed wind farm, so as to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] The method for determining the reactive output power of a doubly-fed wind farm comprises the following specific steps:

[0010] Collecting several groups of wind farm historical time series data, the historical time series data including the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, and recording the reactive output power value at the corresponding timestamp;

[0011] According to the historical time series data, the wind farm scenarios are classified to obtain different wind farm scenario types, wherein the scenario types include a level 1 wind speed load scenario, a level 2 wind speed load scenario, and a level 3 wind speed load scenario. Based on the different scenario types, the estimated target value of the reactive output power of the wind farm under different scenarios is calculated;

[0012] Establish a wind farm reactive output power prediction model, use the collected historical time series data as sample data, use the doubly fed unit speed, cut-in wind speed, cut-out wind speed, ambient temperature, and load demand as input data, and use the corresponding reactive output power value as a label to train the model;

[0013] Collecting real-time characteristic data of the wind farm, the real-time characteristic data including the rotation speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, judging the current wind farm scenario type according to the real-time characteristic data, and determining the estimated target value of the reactive output power of the wind farm under the scenario;

[0014] The real-time feature data is input into the trained wind farm reactive output power prediction model to obtain a reactive output power prediction value, and the reactive output power prediction value is adjusted by the reactive output power estimation target value to obtain an accurate reactive output power value;

[0015] According to the reactive power regulation range of the doubly-fed wind turbine generator and the energy storage system, it is determined whether the precise value of the reactive output power meets the regulation range. If it meets the regulation range, the precise value of the reactive output power is output.

[0016] Furthermore, several groups of wind farm historical time series data are collected, and the historical time series data including the speed of the double-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand are organized into a time series data vector X i ={x i1 ,x i2 ,…,x i5}, where X i represents the time series data vector at the i-th moment, x i1 ,x i2 ,…,x i5 They correspond to five data, namely, the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand. i is the index of the time series data vector, where i=1,2,…,n, and n is the number of time series data vectors.

[0017] Furthermore, based on the historical time series data, the wind farm scenarios are classified and the logic for obtaining different wind farm scenario types is as follows: First, the time series data vector X i The data in is standardized, and the formula for standardization is:

[0018]

[0019] In the formula, x i1 ′ is the standardized value of the unit speed, μ1 is the mean of all unit speeds, and σ1 is the standard deviation of the unit speed;

[0020] The time series data vector X iAfter the remaining data in is standardized, the standardized time series data vector X is obtained. i ′={x i1 ′,x i2 ′,…,x i5 ′}, based on the standard time series data vector X i ′, use the K-means clustering algorithm to classify the wind farm scene. The specific steps include determining the number of clusters to be 3, selecting 3 points as the initial cluster centers, calculating the Euclidean distance between each standard time series data vector and all cluster centers, and assigning the standard time series data vector to the cluster center with the nearest distance. The formula for calculating the Euclidean distance is:

[0021]

[0022] In the formula, d(X i ′,α kj ) represents the standard time series data vector X i ′ is the distance from the jth initial cluster center, α kj The eigenvector representing the jth initial cluster center, j is the index of the cluster number, where j = 1, 2, 3;

[0023] After assigning the samples to the nearest initial cluster center, the cluster center is updated. For each cluster, the mean of all samples assigned to the cluster is calculated and the mean is used as the new cluster center. The formula for updating the cluster center is:

[0024]

[0025] In the formula, C j represents the set of standard time series data vectors assigned to the jth cluster, |C j | represents the number of standard time series data vectors of the jth cluster, α kj ′ is the updated cluster center feature vector; repeat the above steps of allocating standard time series data vectors and updating cluster centers until the cluster center no longer changes or reaches the preset maximum number of iterations, and define the three clusters as the first-level wind speed load scenario, the second-level wind speed load scenario, and the third-level wind speed load scenario, respectively. According to the clustering results, the time series data vector X i Mark and get each time series data vector X i The cluster to which it belongs.

[0026] Furthermore, based on different scenario types, the logic for calculating the estimated target value of reactive output power of the wind farm in different scenarios is as follows: The formula for calculating the estimated target value of reactive output power in the first-level wind speed load scenario is:

[0027]

[0028] In the formula, is the estimated target value of reactive output power under the first-level wind speed load scenario, P load is the load power of the power grid, pf represents the expected power factor, M is the safety margin, and ω1 is the weight coefficient of the safety margin under the first-level wind speed load scenario;

[0029] By setting different safety margin weight coefficients ω2 and ω3, the estimated reactive output power target value under the second and third wind speed load scenarios is calculated. and

[0030] Furthermore, based on the long short-term memory network model LSTM, a wind farm reactive output power prediction model is established, and the hyperparameters of the LSTM model are set. The hyperparameters of the LSTM model include: the number of network layers, the number of iterations, the learning rate, the batch size, the number of training times, the batch processing number, and the number of hidden layer neurons;

[0031] The number of network layers is set to a four-layer network structure, the number of iterations is set to 500, the learning rate is set to 0.001, the batch size is set to 32, the number of training times is set to 300, the batch size is set to 256, and the number of hidden layer neurons is 16.

[0032] Furthermore, the reactive output power prediction value is adjusted by the reactive output power estimation target value, and the formula for obtaining the precise value of reactive output power is as follows:

[0033]

[0034] In the formula, Q adj is the exact value of reactive output power, Q pr Represents the reactive output power prediction value, It represents the estimated target value of reactive output power under the real-time wind farm scenario type, and K represents the adjustment coefficient, where K∈(0,1).

[0035] Furthermore, the reactive power range of the doubly-fed wind turbine generator in the wind farm is calibrated as [Q min ,Q max ], where the maximum reactive power of the doubly-fed wind turbine generator is Q max and minimum reactive power Q min The calculations are based on the following formulas:

[0036]

[0037] In the formula, S r is the rated power of the doubly-fed wind turbine generator, and PF is the set power factor;

[0038] The reactive power regulation capability of the energy storage system is calibrated as The maximum reactive power that the energy storage system can output and minimum reactive power The calculations are based on the following formulas:

[0039]

[0040] In the formula, gl cha is the charging power of the energy storage system, PF sto is the power factor of the energy storage system;

[0041] The comprehensive reactive power output capacity of the wind farm is determined based on the reactive power output range of the doubly-fed wind turbine generator and the energy storage system. The overall reactive power regulation capacity range can be expressed as:

[0042]

[0043] In the formula, is the minimum value of the wind farm’s comprehensive reactive power output capacity, It is the maximum value of the comprehensive reactive power output capacity of the wind farm.

[0044] Further, it is determined whether the precise value of reactive output power meets the regulation range. The logic for outputting the precise value of reactive output power within the regulation range is as follows:

[0045] when When determining that the calculated precise value of reactive output power meets the requirements, the precise value is output.

[0046] The present invention also provides a system for determining reactive output power of a doubly-fed wind farm, wherein the system for determining reactive output power of a doubly-fed wind farm is used to execute a method for determining reactive output power of a doubly-fed wind farm, comprising:

[0047] A historical data collection module is used to collect several groups of wind farm historical time series data, including the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, and record the reactive output power value at the corresponding timestamp;

[0048] A wind farm scenario classification module is used to classify wind farm scenarios according to historical time series data to obtain different wind farm scenario types, wherein the scenario types include level 1 wind speed load scenario, level 2 wind speed load scenario and level 3 wind speed load scenario, and based on different scenario types, calculate the estimated target value of reactive output power of the wind farm under different scenarios;

[0049] The prediction model training module is used to establish a wind farm reactive output power prediction model, using the collected historical time series data as sample data, the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature, and the load demand as input data, and the corresponding reactive output power value as a label to train the model;

[0050] A real-time scenario determination module is used to collect real-time characteristic data of the wind farm, the real-time characteristic data including the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, and to determine the current wind farm scenario type according to the real-time characteristic data, and determine the estimated target value of the reactive output power of the wind farm under the scenario;

[0051] The power prediction value correction module is used to input the real-time feature data into the trained wind farm reactive output power prediction model to obtain the reactive output power prediction value, and adjust the reactive output power prediction value by the reactive output power estimation target value to obtain the precise reactive output power value;

[0052] The power precise value output module is used to determine whether the precise value of reactive output power meets the adjustment range according to the reactive power adjustment range of the doubly-fed wind turbine generator and the energy storage system, and output the precise value of reactive output power if it meets the adjustment range.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] The historical time series data of the wind farm is collected, and the K-means clustering algorithm is used to classify the wind farm scenarios based on the historical time series data. The estimated target value of reactive output power is calculated for different wind farm scenario types (such as level 1, level 2, and level 3 wind speed load scenarios); the strategy based on scenario classification is more targeted than the prediction of a single model, thereby improving the accuracy of reactive output power prediction and reducing grid voltage fluctuations caused by prediction errors; at the same time, the reactive output power prediction value corresponding to the real-time feature data is calculated through a deep learning model, and the deep learning model is used to improve the system convergence speed, so that the reactive output power of the wind farm can be quickly adjusted, thereby improving the stability and response speed of the system; finally, the reactive output power prediction value is corrected by calculating the estimated target value of reactive output power for different wind farm scenario types to obtain the precise value of reactive output power, which better reflects the actual situation of the current wind farm, reduces prediction errors, improves the regulation accuracy of reactive power, and ensures the voltage stability of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0056] Figure 2 It is a schematic diagram of the overall system flow of the present invention. DETAILED DESCRIPTION

[0057] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0058] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0059] Example:

[0060] See also Figure 1 , the present invention provides a technical solution:

[0061] The method for determining the reactive output power of a doubly-fed wind farm comprises the following specific steps:

[0062] Step 1: Collect several groups of historical time series data of the wind farm, the historical time series data including the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, and record the reactive output power value at the corresponding timestamp;

[0063] Collect several groups of wind farm historical time series data, and organize a group of the historical time series data including the speed of the double-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand into a time series data vector X i ={x i1 ,x i2 ,…,x i5}, where X i represents the time series data vector at the i-th moment, x i1 ,x i2 ,…,x i5 They correspond to five data, namely, the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand. i is the index of the time series data vector, where i=1,2,…,n, and n is the number of time series data vectors.

[0064] Step 2: Classify the wind farm scenarios according to the historical time series data to obtain different wind farm scenario types, where the scenario types include level 1 wind speed load scenario, level 2 wind speed load scenario and level 3 wind speed load scenario. Based on the different scenario types, calculate the estimated target value of reactive output power of the wind farm under different scenarios.

[0065] According to the historical time series data, the wind farm scenarios are classified and the logic of obtaining different wind farm scenario types is as follows: First, the time series data vector X i The data in is standardized, and the formula for standardization is:

[0066]

[0067] In the formula, x i1 ′ is the standardized value of the unit speed, μ1 is the mean of all unit speeds, and σ1 is the standard deviation of the unit speed;

[0068] The time series data vector X i After the remaining data in is standardized, the standardized time series data vector X is obtained. i ′={x i1 ′,x i2 ′,…,x i5 ′}, based on the standard time series data vector X i ′, use the K-means clustering algorithm to classify the wind farm scene. The specific steps include determining the number of clusters to be 3, selecting 3 points as the initial cluster centers, calculating the Euclidean distance between each standard time series data vector and all cluster centers, and assigning the standard time series data vector to the cluster center with the nearest distance. The formula for calculating the Euclidean distance is:

[0069]

[0070] In the formula, d(X i ′,α kj ) represents the standard time series data vector X i ′ is the distance from the jth initial cluster center, α kj The eigenvector representing the jth initial cluster center, j is the index of the cluster number, where j = 1, 2, 3;

[0071] After assigning the samples to the nearest initial cluster center, the cluster center is updated. For each cluster, the mean of all samples assigned to the cluster is calculated and the mean is used as the new cluster center. The formula for updating the cluster center is:

[0072]

[0073] In the formula, C jrepresents the set of standard time series data vectors assigned to the jth cluster, |C j | represents the number of standard time series data vectors of the jth cluster, α kj ′ is the updated cluster center feature vector; the allocation step and the update step are repeated until the cluster center no longer changes significantly or the set maximum number of iterations is reached.

[0074] The three clusters are defined as the first-level wind speed load scenario, the second-level wind speed load scenario, and the third-level wind speed load scenario. According to the clustering results, the time series data vector X i Mark and get each time series data vector X i The cluster to which it belongs, the first-level wind speed load scenario described in this embodiment represents a scenario with a large wind speed and a high load demand; the second-level wind speed load scenario represents a scenario with a moderate wind speed and a low load demand; the third-level wind speed load scenario represents a scenario with a low wind speed and a stable load demand; the method for selecting 3 points as the initial clustering center is: setting a classification threshold, and dividing the set classification threshold into different size ranges, for example (0.7*yz, 1.0*yz), (0.4*yz, 0.7*yz) and (0, 0.4*yz); randomly selecting a point in the range of (0.7*yz, 1.0*yz) as the initial clustering center of the third-level wind speed load scenario, and randomly selecting a point in the range of (0.4*yz, 0.7*yz) as the initial clustering center of the second-level wind speed load scenario, and the same is true for the first-level wind speed load scenario, where the value range of the classification threshold yz is 0 to 1; and the definition of the scene can be dynamically adjusted according to actual conditions.

[0075] Based on different scenario types, the logic for calculating the estimated target value of reactive output power of wind farms in different scenarios is as follows: The formula for calculating the estimated target value of reactive output power in the first-level wind speed load scenario is:

[0076]

[0077] In the formula, is the estimated target value of reactive output power under the first-level wind speed load scenario, P load is the load power of the power grid, pf is the expected power factor, usually with a value of 0.95, M is the safety margin, and ω1 is the weight coefficient of the safety margin under the first-level wind speed load scenario;

[0078] By setting different safety margin weight coefficients ω2 and ω3, the estimated reactive output power target value under the second and third wind speed load scenarios is calculated. and

[0079] The weight coefficients are different in different wind speed load scenarios, which are divided into ω1, ω2 and ω3, where ω1, ω2 and ω3 are all greater than 0. Because the demand for reactive power of the power grid will increase significantly under high wind speed (for example, greater than 10m / s) and high load (for example, the grid load is close to the maximum carrying capacity), and the demand for reactive power of the power grid is low under medium wind speed (for example, 5-10m / s) and low load (for example, the grid load is lower than the rated value), ω1<ω2<ω3 is set.

[0080] Step 3: Establish a wind farm reactive output power prediction model, use the collected historical time series data as sample data, use the speed of the doubly fed unit, cut-in wind speed, cut-out wind speed, ambient temperature, and load demand as input data, and use the corresponding reactive output power value as a label to train the model;

[0081] Based on the long short-term memory network model LSTM, a wind farm reactive output power prediction model is established, and the hyperparameters of the LSTM model are set. The hyperparameters of the LSTM model include: the number of network layers, the number of iterations, the learning rate, the batch size, the number of training times, the batch processing number, and the number of hidden layer neurons;

[0082] The number of network layers is set to a four-layer network structure, the number of iterations is set to 500, the learning rate is set to 0.001, the batch size is set to 32, the number of training times is set to 300, the batch size is set to 256, and the number of hidden layer neurons is 16.

[0083] Step 4: Collect real-time characteristic data of the wind farm, the real-time characteristic data including the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, judge the current wind farm scenario type according to the real-time characteristic data, and determine the estimated target value of reactive output power of the wind farm under the scenario;

[0084] Step 5: Input the real-time feature data into the trained wind farm reactive output power prediction model to obtain a reactive output power prediction value, adjust the reactive output power prediction value by the reactive output power estimation target value, and obtain an accurate reactive output power value;

[0085] The reactive output power prediction value is adjusted by the reactive output power estimation target value, and the formula for obtaining the precise reactive output power value is as follows:

[0086]

[0087] In the formula, Q adj is the exact value of reactive output power, Q pre Represents the reactive output power prediction value, It represents the estimated target value of reactive output power under the real-time wind farm scenario type, and K represents the adjustment coefficient, where K∈(0,1). The adjustment of reactive power needs to balance different demands. Excessive adjustment may cause system instability. Limiting K to the range of 0 to 1 can ensure that the adjustment process is gradual and controllable.

[0088] Step 6: According to the reactive power regulation range of the doubly-fed wind turbine generator and the energy storage system, determine whether the precise value of the reactive output power meets the regulation range. If it meets the regulation range, output the precise value of the reactive output power.

[0089] The reactive power range of the double-fed wind turbine generator in the wind farm is calibrated as [Q min ,Q max ], where the maximum reactive power of the doubly-fed wind turbine generator is Q max and minimum reactive power Q min The calculations are based on the following formulas:

[0090]

[0091] In the formula, S r is the rated power of the doubly-fed wind turbine generator, and PF is the set power factor;

[0092] The reactive power regulation capability of the energy storage system is calibrated as The maximum reactive power that the energy storage system can output and minimum reactive power The calculations are based on the following formulas:

[0093]

[0094] In the formula, gl cha is the charging power of the energy storage system, PF sto is the power factor of the energy storage system;

[0095] Where, the charging power of the energy storage system is cha The calculation is based on the formula:

[0096]

[0097] Where E is the energy storage capacity of the energy storage system, T cha Represents the charging time, η cha is the charging efficiency, which is generally between 70% and 95%;

[0098] The comprehensive reactive power output capacity of the wind farm is determined based on the reactive power output range of the doubly-fed wind turbine generator and the energy storage system. The overall reactive power regulation capacity range can be expressed as:

[0099]

[0100] In the formula, is the minimum value of the wind farm’s comprehensive reactive power output capacity, It is the maximum value of the comprehensive reactive power output capacity of the wind farm.

[0101] It is judged whether the precise value of reactive output power is within the adjustment range. The logic for outputting the precise value of reactive output power within the adjustment range is as follows:

[0102] when When determining that the calculated precise value of reactive output power meets the requirements, the precise value is output;

[0103] when or When , it indicates that the precise value is abnormal and is no longer within the wind farm's energy regulation range. Check whether the process of obtaining the reactive output power forecast value and calculating the reactive output power estimated target value is disturbed, and regenerate the precise reactive output power value.

[0104] See also Figure 2 The present invention also provides a system for determining reactive output power of a doubly-fed wind farm, wherein the system for determining reactive output power of a doubly-fed wind farm is used to execute a method for determining reactive output power of a doubly-fed wind farm, comprising:

[0105] A historical data collection module is used to collect several groups of wind farm historical time series data, including the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, and record the reactive output power value at the corresponding timestamp;

[0106] A wind farm scenario classification module is used to classify wind farm scenarios according to historical time series data to obtain different wind farm scenario types, wherein the scenario types include level 1 wind speed load scenario, level 2 wind speed load scenario and level 3 wind speed load scenario, and based on different scenario types, calculate the estimated target value of reactive output power of the wind farm under different scenarios;

[0107] The prediction model training module is used to establish a wind farm reactive output power prediction model, using the collected historical time series data as sample data, the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature, and the load demand as input data, and the corresponding reactive output power value as a label to train the model;

[0108] A real-time scenario determination module is used to collect real-time characteristic data of the wind farm, the real-time characteristic data including the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, and to determine the current wind farm scenario type according to the real-time characteristic data, and determine the estimated target value of the reactive output power of the wind farm under the scenario;

[0109] The power prediction value correction module is used to input the real-time feature data into the trained wind farm reactive output power prediction model to obtain the reactive output power prediction value, and adjust the reactive output power prediction value by the reactive output power estimation target value to obtain the precise reactive output power value;

[0110] The power precise value output module is used to determine whether the precise value of reactive output power meets the adjustment range according to the reactive power adjustment range of the doubly-fed wind turbine generator and the energy storage system, and output the precise value of reactive output power if it meets the adjustment range.

[0111] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0112] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0113] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0114] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for determining reactive output power of a doubly-fed wind farm, characterized in that: The specific steps include: Collecting several groups of wind farm historical time series data, the historical time series data including the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, and recording the reactive output power value at the corresponding timestamp; According to the historical time series data, the wind farm scenarios are classified to obtain different wind farm scenario types, wherein the scenario types include a level 1 wind speed load scenario, a level 2 wind speed load scenario, and a level 3 wind speed load scenario. Based on the different scenario types, the estimated target value of the reactive output power of the wind farm under different scenarios is calculated; Based on different scenario types, the logic for calculating the estimated target value of reactive output power of wind farms in different scenarios is as follows: The formula for calculating the estimated target value of reactive output power in the first-level wind speed load scenario is: In the formula, is the estimated target value of reactive output power under the first-level wind speed load scenario, is the load power of the power grid, represents the expected power factor, is the safety margin, is the weight coefficient of the safety margin under the first-level wind speed load scenario; By setting different safety margin weight coefficients and , the estimated reactive output power target value under the second and third wind speed load scenarios is calculated and ; Establish a wind farm reactive output power prediction model, use the collected historical time series data as sample data, use the doubly fed unit speed, cut-in wind speed, cut-out wind speed, ambient temperature, and load demand as input data, and use the corresponding reactive output power value as a label to train the model; Collecting real-time characteristic data of the wind farm, the real-time characteristic data including the rotation speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, judging the current wind farm scenario type according to the real-time characteristic data, and determining the estimated target value of the reactive output power of the wind farm under the scenario; The real-time feature data is input into the trained wind farm reactive output power prediction model to obtain a reactive output power prediction value, and the reactive output power prediction value is adjusted by the reactive output power estimation target value to obtain an accurate reactive output power value; The exact value of reactive output power is calculated based on the formula: In the formula, is the precise value of reactive output power, Represents the predicted value of reactive output power, Indicates the estimated target value of reactive output power under the real-time wind farm scenario type. represents the adjustment coefficient, where ; According to the reactive power regulation range of the doubly-fed wind turbine generator and the energy storage system, it is determined whether the precise value of the reactive output power meets the regulation range. If it meets the regulation range, the precise value of the reactive output power is output.

2. The method for determining reactive output power of a doubly-fed wind farm according to claim 1, characterized in that: Collect several groups of wind farm historical time series data, and organize a group of the historical time series data including the speed of the double-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand into a time series data vector ,in Indicates The time series data vector at the moment, They correspond to the five data of the doubly-fed unit speed, cut-in wind speed, cut-out wind speed, ambient temperature and load demand. is the index of the time series data vector, where , is the number of time series data vectors.

3. The method for determining reactive output power of a doubly-fed wind farm according to claim 2, characterized in that: According to the historical time series data, the wind farm scenarios are classified and the logic of obtaining different wind farm scenario types is as follows: first, the time series data vector The data in is standardized, and the formula for standardization is: In the formula, is the normalized value of the unit speed, is the average speed of all units, is the standard deviation of the unit speed; The time series data vector After the remaining data in is standardized, the standardized time series data vector is obtained. , based on the standard time series data vector ,use The mean clustering algorithm is used to classify wind farm scenarios. The specific steps include determining the number of clusters to be 3, selecting 3 points as the initial cluster centers, calculating the Euclidean distance between each standard time series data vector and all cluster centers, and assigning the standard time series data vector to the cluster center with the nearest distance. The formula for calculating the Euclidean distance is: In the formula, Represents a standard time series data vector With The distance between the initial cluster centers, Indicates The feature vectors of the initial cluster centers, is the index of the number of clusters, where ; After assigning the samples to the nearest initial cluster center, the cluster center is updated. For each cluster, the mean of all samples assigned to the cluster is calculated and the mean is used as the new cluster center. The formula for updating the cluster center is: In the formula, Indicates that the A set of standard time series data vectors for clusters, Indicates The number of standard time series data vectors for clusters, is the updated cluster center feature vector; repeat the above steps of allocating standard time series data vectors and updating cluster centers until the cluster center no longer changes or reaches the preset maximum number of iterations, and define the three clusters as the first-level wind speed load scenario, the second-level wind speed load scenario, and the third-level wind speed load scenario, respectively. According to the clustering results, the time series data vector Mark and get each time series data vector The cluster to which it belongs.

4. The method for determining reactive output power of a doubly-fed wind farm according to claim 1, characterized in that: Based on the long short-term memory network model LSTM, a wind farm reactive output power prediction model is established, and the hyperparameters of the LSTM model are set. The hyperparameters of the LSTM model include: the number of network layers, the number of iterations, the learning rate, the batch size, the number of training times, the batch processing number, and the number of hidden layer neurons; The number of network layers is set to a four-layer network structure, the number of iterations is set to 500, the learning rate is set to 0.001, the batch size is set to 32, the number of training times is set to 300, the batch size is set to 256, and the number of hidden layer neurons is 16.

5. The method for determining reactive output power of a doubly-fed wind farm according to claim 1, characterized in that: The reactive power range of the doubly-fed wind turbine generator in the wind farm is calibrated as follows: , where the maximum reactive power of the doubly-fed wind turbine generator is and minimum reactive power The calculations are based on the following formulas: In the formula, is the rated power of the doubly-fed wind turbine generator, is the set power factor; The reactive power regulation capability of the energy storage system is calibrated as , where the maximum reactive power that the energy storage system can output is and minimum reactive power The calculations are based on the following formulas: In the formula, is the charging power of the energy storage system, is the power factor of the energy storage system; The comprehensive reactive power output capacity of the wind farm is determined based on the reactive power output range of the doubly-fed wind turbine generator and the energy storage system. The overall reactive power regulation capacity range can be expressed as: In the formula, is the minimum value of the wind farm’s comprehensive reactive power output capacity, It is the maximum value of the comprehensive reactive power output capacity of the wind farm.

6. The method for determining reactive output power of a doubly-fed wind farm according to claim 5, characterized in that: The logic for judging whether the precise value of reactive output power is within the adjustment range and outputting the precise value of reactive output power if it is within the adjustment range is as follows: when When determining that the calculated precise value of reactive output power meets the requirements, the precise value is output.

7. A system for determining reactive output power of a doubly-fed wind farm, characterized in that: The system for determining reactive output power of a doubly-fed wind farm is used to execute the method for determining reactive output power of a doubly-fed wind farm according to any one of claims 1 to 6, comprising: A historical data collection module is used to collect several groups of wind farm historical time series data, including the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, and record the reactive output power value at the corresponding timestamp; A wind farm scenario classification module is used to classify wind farm scenarios according to historical time series data to obtain different wind farm scenario types, wherein the scenario types include level 1 wind speed load scenario, level 2 wind speed load scenario and level 3 wind speed load scenario, and based on different scenario types, calculate the estimated target value of reactive output power of the wind farm under different scenarios; The prediction model training module is used to establish a wind farm reactive output power prediction model, using the collected historical time series data as sample data, the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature, and the load demand as input data, and the corresponding reactive output power value as a label to train the model; A real-time scenario determination module is used to collect real-time characteristic data of the wind farm, the real-time characteristic data including the speed of the doubly-fed unit, the cut-in wind speed, the cut-out wind speed, the ambient temperature and the load demand, and to determine the current wind farm scenario type according to the real-time characteristic data, and determine the estimated target value of the reactive output power of the wind farm under the scenario; The power prediction value correction module is used to input the real-time feature data into the trained wind farm reactive output power prediction model to obtain the reactive output power prediction value, and adjust the reactive output power prediction value by the reactive output power estimation target value to obtain the precise reactive output power value; The power precise value output module is used to determine whether the precise value of reactive output power meets the adjustment range according to the reactive power adjustment range of the doubly-fed wind turbine generator and the energy storage system, and output the precise value of reactive output power if it meets the adjustment range.

Images