Wind power plant active power control method, device and equipment based on theoretical power error real-time correction and storage medium

Abstract

The invention provides a wind power plant active power control method, device and equipment based on theoretical power error real-time correction and a storage medium, and relates to the technical field of power grid control. According to the method, the actual theoretical power error of a single unit in the wind power plant power adjusting process is combined, the influence of the wake effect of the unit and other factors is comprehensively considered, and a wind speed-power BP neural network model of each unit cabin of the wind power plant is established. After the wind power plant receives an instruction issued by the dispatching master station, a corresponding control strategy is formulated according to the output capacity margin of each fan and the theoretical power error of each unit, so that the active power control precision and response rate of the wind power plant are improved, and the active power control performance is prevented from exceeding the standard due to multiple rounds of repeated adjustment.

Description

Technical Field

[0001] The present invention relates to the field of power grid control technology, and in particular to a wind farm active power control method, device, equipment and storage medium based on real-time correction of theoretical power error. Background Art

[0002] The AGC control performance of a wind farm is directly related to the frequency stability of the power grid. The control strategies currently used for AGC in wind farms mainly include: proportional active power distribution, average active power change distribution, and similar output margin distribution. Since the former two do not take into account the unit output capacity margin, multiple rounds of repeated adjustments are sometimes required, resulting in a long adjustment time. The latter has significantly improved adjustment, but the adjustment accuracy is affected by the control performance of a single unit. None of the above studies considered the active power upper limit error of the unit. In fact, the theoretical power is the main basis for the output margin, and how to accurately calculate it has always been one of the difficulties.

[0003] In response to the above problems, the current research focuses mainly on the calculation of theoretical power, which roughly includes two categories: benchmark wind turbine method and historical data method. The benchmark wind turbine method is to keep several typical units in normal operation when the wind turbine is limited in power. Under similar wind speed, air density and mechanical environment conditions, its actual power can be equivalent to the theoretical power of adjacent units. The selection of benchmark units is the key to whether the theoretical power can be accurately calculated, and it should be continuously optimized and adjusted during the commissioning stage. Historical data analysis belongs to the category of data mining or deep learning. Unlike simple theoretical power curves, the theoretical power of wind turbines determined at the factory often deviates from the theoretical power curve in actual operation due to mechanical wear, changes in environmental conditions and other reasons. In the data analysis of a large number of test sites, it was found that the active upper limit error tends to be larger or smaller for different units, and the error size of different units is different. For example, the power curve of units with large wind conditions and severe mechanical vibration changes greatly, so the calculated theoretical power is too large. Summary of the invention

[0004] Purpose of the invention: To propose a wind farm active power control method, device, equipment and storage medium based on real-time correction of theoretical power error, so as to solve the above-mentioned problems existing in the prior art.

[0005] In a first aspect of the present invention, a method for controlling active power of a wind farm based on real-time correction of theoretical power error is proposed, and the steps are as follows:

[0006] S1. Collect AGC monitoring data of each fan unit in real time;

[0007] S2, determine whether the active power adjustment instruction issued by the power dispatching master station is received, if so, execute step S3, otherwise return to step S1 to continue monitoring;

[0008] S3, constructing an active power prediction model; using the active power prediction model to predict the upper limit of active power of each unit;

[0009] S4, integrating the AGC monitoring data and the predicted active power upper limit to obtain the theoretical power of each unit;

[0010] S5. Determine the adjustment command value and send it to each unit, and each unit completes power adjustment according to the adjustment command value;

[0011] S6. Reserve a predetermined adjustment waiting time for each unit;

[0012] S7. After the countdown of the adjustment waiting time ends, determine whether the active power deviation value of each unit is less than the adjustment accuracy value required by the dispatching. If so, this round of control process ends; otherwise, return to step S1.

[0013] In a further embodiment of the first aspect, the AGC monitoring data includes the start and stop status of each wind turbine unit, the real-time wind direction and wind speed at different altitudes, the active power of each wind turbine in the unit, the active power of the grid connection point, and the active power adjustment instructions issued by the power dispatching master station.

[0014] In a further embodiment of the first aspect, the active power prediction model is constructed based on a BP neural network;

[0015] The input layer of the BP neural network is the historical upper limit of active power of each wind turbine, and the output layer is the predicted upper limit of active power of each wind turbine. The expressions of the input layer and the output layer are as follows:

[0016]

[0017] In the formula, are the sum of the input and output of the i-th neuron in the k-th layer; w ij is the weight coefficient from the jth neuron in the k-1th layer to the ith neuron in the kth layer; θ i k is the threshold of the i-th neuron in the k-th layer.

[0018] In a further embodiment of the first aspect, the current output is compared with the expected output, and if the current output is not within a predetermined floating range of the expected output, a back-propagation process is entered.

[0019] In a further embodiment of the first aspect, the current output is compared with the expected output, and if the current output is not within a predetermined floating range of the expected output, a reverse propagation process is entered, which specifically includes:

[0020] For the weight coefficient w ij Set a random number between the interval (0,1) as the initial value, and the initial layer number k = 1;

[0021] The i-th neuron in the k-th layer As a training sample input to the BP neural network, output the output u of the BP neural network during the kth iteration i k ;

[0022] Calculate the learning error d of the output layer during the kth iteration k :

[0023]

[0024] In the formula, t k It indicates that the expected output value under free power generation condition is calculated according to the preset quality index;

[0025] If the learning error d k If it is within the predetermined floating range, the current weight coefficient is retained;

[0026] Otherwise, the weight coefficients of each layer are modified according to the following formula:

[0027]

[0028] In the formula, μ represents the learning rate of the BP neural network in actual training, and its value is greater than 0.

[0029] In a further embodiment of the first aspect, the adjustment command value is determined according to the following formula and sent to each unit:

[0030]

[0031] Where P i obj is the AGC adjustment target value of the i-th fan; P i real is the active power value of the i-th wind turbine at the current moment; P i the is the theoretical power value of the i-th wind turbine at the current moment, that is, the output value of the BP neural network; P wind real is the active power value of the wind farm at the current moment; P wind the is the theoretical power value of the wind farm at the current moment.

[0032] The second aspect of the present invention provides a wind farm active power control device, which can automatically execute the wind farm active power control method based on real-time correction of theoretical power error disclosed in the first aspect. The device includes:

[0033] The acquisition module is used to collect the AGC monitoring data of each fan unit in real time;

[0034] The first judgment module is used to judge whether the active power regulation instruction issued by the power dispatching master station is received, and if so, it is fed back to the prediction module, otherwise it returns to the acquisition module to continue monitoring;

[0035] A prediction module, which has a built-in active power prediction model, and uses the active power prediction model to predict the active power upper limit of each unit;

[0036] Theoretical power calculation module, used to integrate AGC monitoring data and predicted active power upper limit to obtain the theoretical power of each unit;

[0037] The power regulation module is used to determine the regulation command value and send it to each unit, and each unit completes power regulation according to the regulation command value;

[0038] A waiting module is used to reserve a predetermined length of adjustment waiting time for each unit;

[0039] The second judgment module, after the countdown of the adjustment waiting time ends, determines whether the active power deviation value of each unit is less than the adjustment accuracy value required by the scheduling. If so, this round of control process ends; otherwise, it is fed back to the acquisition module for continued monitoring.

[0040] According to a third aspect of the present invention, an electronic device is proposed, comprising: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the wind farm active power control method based on real-time correction of theoretical power error as described in the first aspect is implemented.

[0041] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, in which at least one executable instruction is stored. When the executable instruction is executed on an electronic device, the electronic device executes the wind farm active power control method based on real-time correction of theoretical power error as described in the first aspect.

[0042] Beneficial effects: The present invention provides a wind farm active power control method, device, equipment and storage medium based on real-time correction of theoretical power error. The core of the invention is to comprehensively consider the theoretical power output of the wind farm and its internal wind turbines (wind turbines) when responding to the automatic generation control (AGC) adjustment command, and optimize power distribution and control through a real-time error correction mechanism, thereby ensuring that the wind farm can participate in the power regulation of the power grid more accurately and efficiently. By improving the AGC regulation performance of the wind farm, it effectively promotes the participation of new energy in the frequency regulation of the power grid, and helps the safe and stable operation of the new power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1This is a flow chart of a wind farm active power control method based on real-time correction of theoretical power error proposed by the present invention.

[0044] Figure 2 This is the main electrical wiring diagram of the wind farm tested by the algorithm of the present invention.

[0045] Figure 3 This is a graph showing the convergence process of the active upper limit model calculation of the BP neural network cited in the present invention.

[0046] Figure 4 This is a graph of the AGC regulation command of the wind farm tested by the present invention.

[0047] Figure 5 These are the performance parameter values ​​of active power regulation (AGC) of wind farms under different methods, including response time and regulation time. DETAILED DESCRIPTION

[0048] In the following description, a large number of specific details are provided to provide a more thorough understanding of the present invention. However, it is apparent to those skilled in the art that the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present invention, some technical features known in the art are not described.

[0049] The wind farm AGC consists of the wind farm layer AGC system and the unit layer AGC module. The wind farm layer AGC receives the active power regulation target value instruction issued by the dispatching agency, and combines the real-time operation information of the wind farm, and uses the set allocation strategy to allocate and issue the active power control target value of each unit to each unit in the field, thereby adjusting the overall active power of the wind farm, meeting the system dispatching requirements, and reducing the peak load regulation pressure of the power grid.

[0050] This application selects several typical wind farms and wind farm layers of single units for statistical analysis of active upper limit errors, and obtains relevant statistical indicators such as active upper limit error distribution curve, mean absolute error, root mean square error, etc. of each unit. On this basis, an AGC control strategy based on BP neural network active upper limit error correction is proposed, and the historical data of the actual wind farm in free power generation mode is used as the BP neural network training sample, so as to obtain a BP neural network model suitable for real-time active upper limit correction.

[0051] The AGC control strategy disclosed in this embodiment is based on Figure 1 The process execution steps can be divided into the station level and the unit level. The specific implementation steps are as follows:

[0052] Step 1: The AGC monitoring data at the station level includes the start and shutdown status of each unit, the real-time wind direction and speed at 50 meters, 70 meters and the nacelle, the active power of the wind turbine, the active power of the station grid connection point, etc.

[0053] Step 2: Determine whether the active power adjustment command issued by the power dispatching master station is received (Generally, the station adjusts the power curve). If yes, execute step 3, otherwise return to step 1.

[0054] Step 3: Start responding to the AGC adjustment instructions issued by the dispatching agency and start the active power control strategy formulation program.

[0055] Step 4: Based on the real-time wind direction and wind speed data of each wind turbine at 50 meters, 70 meters and the cabin collected in Step 1, and combined with the BP neural network trained above, an active power prediction model for each wind turbine is established (the historical active power of each wind turbine is used as the input layer, and the theoretical power is used as the output layer, which is a mapping relationship of multiple inputs and single outputs), and the real-time predicted theoretical power of each wind turbine is obtained.

[0056] Step 5: Use the “equal capacity margin method” to determine the adjustment command value of each fan and send it to each fan; the calculation method of the equal capacity margin method is as follows:

[0057]

[0058] Where P i obj is the AGC adjustment target value of the i-th fan; P i real is the active power value of the i-th wind turbine at the current moment; P i the is the theoretical power value of the i-th wind turbine at the current moment, that is, the output value of the BP neural network; P wind real is the active power value of the wind farm at the current moment; P wind the is the theoretical power value of the wind farm at the current moment.

[0059] Step 6: Wait for 30 seconds. This stage is the wind turbine power adjustment time. Generally, the wind turbine pitch angle and wind turbine yaw system are started.

[0060] Step 7: After 30 seconds, determine the real-time active power P of the wind farm grid connection point wind real and Is the deviation less than the regulation accuracy value required by the dispatch (such as 1 MW, 1% of the rated active power of the wind farm)? If so, the program ends; if not, it returns to Step 1 again.

[0061] The output of a single wind turbine is still difficult to estimate even if factors such as wind speed error and wind speed wave are considered at the same time. This patent uses the BP neural network widely used in the field of machine learning to predict the upper limit of the active power of each unit. The BP neural network (Back-ProPagation Network), also known as the back-lead propagation neural network, consists of an input layer, a hidden layer and an output layer. Its characteristics are that each layer of neurons is fully connected only to adjacent neurons, there is no connection between neurons in the same layer, and there is no feedback connection between neurons in each layer. The input layer selects the upper limit of the active power of each wind turbine. In this embodiment, the expression of the BP neural network is as follows:

[0062]

[0063] In the formula, are the sum of the input and output of the i-th neuron in the k-th layer; w ij is the weight coefficient from the jth neuron in the k-1th layer to the ith neuron in the kth layer; θ i k is the threshold of the i-th neuron in the k-th layer.

[0064] The backpropagation algorithm includes forward propagation and backward propagation. Forward propagation is that the input samples are processed layer by layer from the input layer through the hidden units. After passing through all the hidden layers, they are transmitted to the output layer. In the process of layer-by-layer processing, the state of each layer of neurons only affects the state of the neurons in the next layer.

[0065] At the output layer, the current output is compared with the expected output. If the current output is not equal to the expected output, the back propagation process begins. The specific implementation steps are as follows:

[0066] For the weight coefficient w ij Set a random number between the interval (0,1) as the initial value, and the initial number of layers k = 1.

[0067] The i-th neuron in the k-th layer As a training sample input to the BP neural network, output the output u of the BP neural network during the kth iteration i k .

[0068] Calculate the learning error of the output layer during the kth iteration t k It indicates that the expected output value under free power generation condition is calculated according to the preset quality index.

[0069] If the learning error d k If it is within the predetermined floating range, the current weight coefficient is retained;

[0070] Otherwise, the weight coefficients of each layer are modified according to the following formula:

[0071]

[0072] In the formula, μ represents the learning rate of the BP neural network in actual training, and its value is greater than 0.

[0073] Take the actual data of a typical wind farm as an example for analysis. The wiring diagram of the wind farm is shown in Figure 2 , the unit parameters are shown in Table 1.

[0074] Table 1 2.0MW wind turbine parameters

[0075] project parameter project parameter Rated voltage 690V Rated frequency 50Hz Cut-in wind speed 0.5m / s Rated wind speed 10m / s Cut-out wind speed 15m / s Transient Reactance 0.3496 Manufacturer Goldwind Technology Power Factor -0.95～0.95

[0076] Due to the real-time changes in the operating conditions of wind farms, it is difficult to accurately compare the advantages and disadvantages of different algorithms. Therefore, the real-time data of the booster station, collection system and each unit in the power station on typical days from January to December 2019 were selected, including 2678400 sections: wind speed, theoretical power, grid connection point power, collection system active and reactive power, grid connection point active and reactive power. The method described in this patent is used to simulate offline data and compare it with the average distribution of active power changes and the distribution of similar output margins. The simulation program based on the Matla / Simulink platform was tested on a PC with a CPU of 3.2GHz main frequency and 8G memory.

[0077] The method of correcting the upper limit of active power using the BP neural network proposed in this patent, because the training samples are massive data, it is easy to cause program memory overflow, and the prediction errors of different working conditions are different, so the samples are first classified. The specific classification method can be determined according to the historical data of the power station. The power station classification selected in this patent is shown in Table 2. It can be seen that according to the power change trend, season and climate of the grid-connected point, there are 27 types of sample sets, and different models are numbered with 3-digit numbers. The first, second and third digits represent the corresponding options for power change trend, season and climate respectively. For ease of comparison, the above 27 types of sample model methods are compared with the unclassified BP neural network model, as shown in Table 3, where the calculation result △P=|Prel-Pcal|, Prel is the actual active upper limit value, and Pcal is the BP neural network calculation value, in MW. It can be seen that compared with the unclassified calculation method, the errors of the historical data sample classification method proposed in this patent are all reduced, among which the 231 model has a maximum reduction of 51.09%. The iterative convergence process of the BP neural network of this sample is shown as follows Figure 3 As shown, Figure 3 In the figure, (a) shows the convergence process of the mean square error of a training sample set, and (b) shows the iterative convergence process of a training sample set.

[0078] Table 2 Classification of historical data training samples

[0079] Classification principle Option 1 Option 2 Option 3 Power change trend rise decline smooth season Spring / Autumn summer winter climate clear rain Negative

[0080] Table 3 BP neural network correction calculation results under different categories

[0081] Model Calculation result / △P Model Calculation result / △P 111 1.12 / 0.78 223 0.99 / 0.49 112 1.32 / 0.82 231 0.92 / 0.45 113 1.01 / 0.56 232 1.04 / 0.61 121 1.04 / 0.60 233 1.10 / 0.67 122 1.08 / 0.72 311 1.04 / 0.52 123 1.11 / 0.65 312 1.05 / 0.70 131 1.08 / 0.59 313 0.97 / 0.65 132 0.98 / 0.49 321 1.12 / 0.70 133 0.95 / 0.47 322 1.04 / 0.65 211 1.12 / 0.56 323 0.94 / 0.49 212 1.20 / 0.66 331 0.98 / 0.52 213 1.18 / 0.70 332 1.02 / 0.64 221 1.05 / 0.64 333 1.01 / 0.56 222 0.87 / 0.45 / /

[0082] After training the "wind speed-power" model of the wind farm and wind turbine, the model is introduced into the AGC system. This patent has the following contents: Figure 4 As shown in the figure, the active power output of the wind farm drops from 80% to 20%, with each drop of 20%, and continues to run for 4 minutes at each control point. Then the active power increases from 20% to 80%, with each increase of 20%, and continues to run for 4 minutes at each control point.

[0083] In each active power setting value adjustment period, the response time and adjustment time are calculated respectively. The average distribution of active power variation is referred to as ADAPV, the average distribution of active power variation based on BP neural network correction is referred to as ADAPV-TPR, the similar output margin distribution is referred to as DSOM, and the similar output margin distribution based on BP neural network correction is referred to as DSOM-TPR. The classification method is used to simulate and correct the historical data of the power station. The results are as follows Figure 5 shown. Figure 5 In the figure, (a) shows the response time of different strategies, and (b) shows the adjustment time of different strategies. It can be seen that when the power decreases, the response time performance of different control strategies is equivalent. When the power increases, as the power gradually increases, especially when the available power of some units cannot reach the control target issued by the AGC, the corresponding adjustment cycle cannot enter the dead zone, resulting in an excessive adjustment time.

[0084] Figure 5 In the example of the regulation from 60% rated active power to 80% rated active power, the average response time of ADAPV is 8.90s, while that of ADAPV-TPR is 7.69s, which can be reduced by 13.6%. After adopting DSOM, the average response time is 4.01s, and that of DSOM-TPR is 2.32s, which can be reduced by 42.14%. At the same time, the average regulation time of ADAPV is 54.3s, and that of APAPV-TPR is 36.7s, which can be reduced by 32.41%. After adopting DSOM, the average regulation time is 28.9s, and that of DSOM-TPR is 23.2s, which can be reduced by 19.72%.

[0085] The above simulation results show that DSOM has less response time and adjustment time than ADAPV. This is because the ADAPV strategy only considers the theoretical active power upper limit, which sometimes exceeds the minimum power limit of the unit. In terms of adjustment rate, the rising power is significantly slower than the falling power. This is due to the mechanical vibration protection of the unit. When the rising rate is too fast, the unit will be damaged. Therefore, the manufacturer sets an artificial rising rate limit for the unit.

[0086] It should be noted that the wind farm active power control method disclosed in the above embodiment can be implemented in whole or in part by software, hardware, firmware or any other combination in actual application. When implemented by software, the above embodiment can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0087] As described above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes in form and details may be made without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A wind farm active power control method based on real-time correction of theoretical power error, characterized in that: The steps include: S1. Collect AGC monitoring data of each fan unit in real time; S2, determine whether the active power adjustment instruction issued by the power dispatching master station is received, if so, execute step S3, otherwise return to step S1 to continue monitoring; S3, constructing an active power prediction model; using the active power prediction model to predict the upper limit of active power of each unit; S4, integrating the AGC monitoring data and the predicted active power upper limit to obtain the theoretical power of each unit; S5. Determine the adjustment command value and send it to each unit, and each unit completes power adjustment according to the adjustment command value; S6. Reserve a predetermined adjustment waiting time for each unit; S7. After the countdown of the adjustment waiting time ends, determine whether the active power deviation value of each unit is less than the adjustment accuracy value required by the dispatching: If yes, the control process ends; otherwise, it returns to step S1.

2. The wind farm active power control method based on real-time correction of theoretical power error according to claim 1 is characterized in that: The AGC monitoring data includes the start and stop status of each wind turbine unit, the real-time wind direction and wind speed at different altitudes, the active power of each wind turbine in the unit, the active power of the grid connection point, and the active power regulation instructions issued by the power dispatching master station.

3. The wind farm active power control method based on real-time correction of theoretical power error according to claim 1 is characterized in that: The active power prediction model is constructed based on BP neural network; The input layer of the BP neural network is the historical upper limit of active power of each wind turbine, and the output layer is the predicted upper limit of active power of each wind turbine. The expressions of the input layer and the output layer are as follows: In the formula, are the sum of the input and output of the i-th neuron in the k-th layer; w ij is the weight coefficient from the jth neuron in the k-1th layer to the ith neuron in the kth layer; θ i k is the threshold of the i-th neuron in the k-th layer.

4. The wind farm active power control method based on real-time correction of theoretical power error according to claim 3 is characterized in that: The current output is compared with the expected output. If the current output is not within the predetermined floating range of the expected output, the back propagation process is entered.

5. The wind farm active power control method based on real-time correction of theoretical power error according to claim 4 is characterized in that: The current output is compared with the expected output. If the current output is not within the predetermined floating range of the expected output, a reverse propagation process is entered, which specifically includes: For the weight coefficient w ij Set a random number between the interval (0,1) as the initial value, and the initial layer number k = 1; The i-th neuron in the k-th layer As a training sample input to the BP neural network, output the output u of the BP neural network during the kth iteration i k ; Calculate the learning error d of the output layer during the kth iteration k : Where, t k It indicates that the expected output value under free power generation condition is calculated according to the preset quality index; If the learning error d k If it is within the predetermined floating range, the current weight coefficient is retained; Otherwise, the weight coefficients of each layer are modified according to the following formula: In the formula, μ represents the learning rate of the BP neural network in actual training, and its value is greater than 0.

6. The wind farm active power control method based on real-time correction of theoretical power error according to claim 1 is characterized in that: The adjustment command value is determined according to the following formula and sent to each unit: Where P i obj is the AGC adjustment target value of the i-th fan; P i real is the active power value of the i-th wind turbine at the current moment; P i the is the theoretical power value of the i-th wind turbine at the current moment, that is, the output value of the BP neural network; P wind real is the active power value of the wind farm at the current moment; P wind the is the theoretical power value of the wind farm at the current moment.

7. A wind farm active power control device, used for automatically executing the wind farm active power control method based on real-time correction of theoretical power error as described in any one of claims 1 to 6, characterized in that: The wind farm active power control device comprises: The acquisition module is used to collect the AGC monitoring data of each fan unit in real time; The first judgment module is used to judge whether the active power regulation instruction issued by the power dispatching master station is received, and if so, it is fed back to the prediction module, otherwise it returns to the acquisition module to continue monitoring; A prediction module, which has a built-in active power prediction model, and uses the active power prediction model to predict the active power upper limit of each unit; Theoretical power calculation module, used to integrate AGC monitoring data and predicted active power upper limit to obtain the theoretical power of each unit; The power regulation module is used to determine the regulation command value and send it to each unit, and each unit completes power regulation according to the regulation command value; A waiting module is used to reserve a predetermined length of adjustment waiting time for each unit; The second judgment module, after the countdown of the adjustment waiting time ends, determines whether the active power deviation value of each unit is less than the adjustment accuracy value required by the scheduling. If so, the control process ends; otherwise, it is fed back to the acquisition module for continued monitoring.

8. An electronic device, characterized in that: The device comprises: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the method for controlling active power of a wind farm based on real-time correction of theoretical power error according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The storage medium stores at least one executable instruction. When the executable instruction is executed on the electronic device, the electronic device executes the wind farm active power control method based on real-time correction of theoretical power error according to any one of claims 1 to 6.

Images