Grid-connected wind energy storage energy management method and system based on prediction

Abstract

The application relates to a grid-connected wind and storage energy management method and system based on prediction, a collection module collects historical data; a first time sequence component based on the historical data is decomposed by using an EMD method to obtain a second time sequence component; prediction is carried out based on the second time sequence component, linear fitting prediction is adopted for a linear second time sequence component; reconstruction prediction is carried out based on a phase space for a nonlinear second time sequence component; the second time sequence component after prediction is subjected to an inverse EMD operation to obtain predicted wind speed of wind turbine unit start; the optimal power of the wind turbine is predicted based on the predicted wind speed; the charging and discharging control of the energy storage system of the wind farm is carried out based on the predicted optimal power of the wind turbine, so that the output power of the wind farm in the maximum power point tracking mode of the wind turbine is stabilized. Under the premise of ensuring the prediction accuracy, the data processing amount can be greatly reduced, and the energy management of the grid-connected wind and storage combined system is effectively realized.

Description

Technical Field

[0001] This invention relates to a prediction-based grid-connected wind-storage energy management method and system. Background Technology

[0002] In grid-connected wind-storage integrated system (wind farm) energy management engineering applications, for non-dispatchable generators such as wind turbines, predicting their output and operation can reduce the uncertainty of the microgrid, improve the renewable energy absorption rate of the microgrid, and ensure power supply reliability. For wind turbine prediction, when it is necessary to predict wind turbine power in the ultra-short term (within 15 minutes), smoothing prediction or machine learning methods are usually used. Using linear methods such as smoothing prediction to predict typical nonlinear systems such as the atmosphere (wind) has inherent errors; existing machine learning methods require a large amount of manually labeled historical data to train the network model, requiring the processing of a huge amount of data. Such a huge amount of data and fuzzy label classification are difficult for engineering application-level hardware equipment to meet the requirements. Summary of the Invention

[0003] The purpose of this invention is to provide a prediction-based grid-connected wind and energy storage energy management method and system, which can significantly reduce the amount of data processing while ensuring prediction accuracy, thereby effectively realizing energy management of grid-connected wind and energy storage combined systems.

[0004] Based on the same inventive concept, this invention has two independent technical solutions:

[0005] 1. A prediction-based grid-connected wind-storage energy management method, comprising the following steps:

[0006] Step 1: Collect historical data related to wind turbine energy storage;

[0007] Step 2: Obtain the second time series component from the first time series component based on historical data using the EMD decomposition method;

[0008] Step 3: Make predictions based on the second time series components. For linear second time series components, use linear fitting for prediction; for nonlinear second time series components, reconstruct and predict based on phase space.

[0009] Step 4: Perform inverse EMD operation on the predicted second time series component to obtain the predicted wind speed for wind turbine startup;

[0010] Step 5: Based on the predicted wind speed, predict the optimal power of the wind turbine; based on the predicted optimal power of the wind turbine, perform charge and discharge control on the energy storage system of the wind farm to achieve power output smoothing of the wind farm in the maximum power point tracking mode of the wind turbine.

[0011] Furthermore, in step 1, the historical data includes wind turbine data, meteorological data, and energy storage data.

[0012] Furthermore, step 2 includes the following steps:

[0013] Step 2.1: Extract all maximum and minimum points in the data {x(t)|t=1,2,…,n};

[0014] Step 2.2: Construct its upper and lower envelopes e using cubic interpolation. up (t) and e l o w (t), and their average value m(t);

[0015]

[0016] Step 2.3: Remove {m(t)} from the data to obtain new data ht;

[0017] ht=x(t)-m(t)

[0018] Step 2.4: Repeat steps 2.1 to 2.3k times until the termination condition is met, and obtain the first IMF component {h1t|t=1,2,…,n}, and denote {c1t}={h1t};

[0019] Step 2.5: Remove {c1t} from the data {x(t)} to obtain new data {x′(t)}. Repeat the above process until all IMFs are decomposed, i.e.

[0020]

[0021] In the formula, IMF represents the intrinsic mode, r n (t) represents the residual.

[0022] Furthermore, in step 3, the IMFs with obvious local linearity are linearized by local linearity discrimination and predicted by linear fitting; the phase space of the nonlinear components is constructed and the prediction is reconstructed based on the phase space.

[0023] Furthermore, in step 3, the phase space of the nonlinear components is constructed using the following method:

[0024] For a univariate discrete sequence {x(t)|t=1,2,…,n}, construct the phase space using the time delay τ and the embedding dimension m:

[0025] X(t)=[x(t),x(t+τ),...,x(t+(m-1)τ)]

[0026] Where t=1,…,L,L=n-(m-1)τ, we obtain the phase space matrix:

[0027]

[0028] Furthermore, in step 5, when the predicted power is greater than the reference power, i.e. the given grid-connected power of the wind farm, if the energy state of the energy storage system is less than 90%, the wind turbine maintains the maximum power point tracking mode, and at the same time, the energy storage system enters the charging state.

[0029] Furthermore, in step 5, when the predicted power is greater than the reference power, i.e. the given power of the wind farm to the grid, if the energy state of the energy storage system is greater than 90%, the wind turbine speed is limited, i.e., the wind curtailment state is implemented, and the energy storage system no longer performs charging and discharging operations.

[0030] Furthermore, in step 5, when the predicted power is less than the reference power, i.e. the given wind farm grid connection power, if the energy state of the energy storage system is greater than 20%, the wind turbine maintains the maximum power point tracking mode, and at the same time, the energy storage system enters the discharge state.

[0031] Furthermore, in step 5, if the energy state of the energy storage system is less than 20%, the wind turbine will remain in maximum power point tracking mode, and the energy storage system will no longer perform charging and discharging operations.

[0032] 2. A prediction-based grid-connected wind-storage energy management system, comprising:

[0033] The data acquisition module is used to collect historical data related to wind turbine energy storage, including wind turbine data, meteorological data, and energy storage data.

[0034] The prediction module is used to perform the following operations:

[0035] The first time series component based on historical data is decomposed using the EMD method to obtain the second time series component. Prediction is then performed based on the second time series component: linear fitting is used for linear second time series components, while nonlinear second time series components are reconstructed based on phase space. The predicted second time series component is then subjected to inverse EMD to obtain the predicted wind speed for turbine startup. Based on the predicted wind speed, the optimal power of the wind turbine is predicted. Finally, based on the predicted optimal power of the wind turbine, the charging and discharging control of the wind farm's energy storage system is implemented.

[0036] A scheduling instruction transmission module, which is used to transmit scheduling instructions generated by the prediction module;

[0037] A scheduling instruction execution device is used to execute scheduling instructions to achieve power output smoothing of the wind farm in the maximum power point tracking mode of the wind turbine.

[0038] The beneficial effects of this invention are as follows:

[0039] This invention obtains a second time series component from a first time series component based on historical data using EMD (Empirical Mode Decomposition). Prediction is then performed based on this second time series component: linear fitting is used for linear second time series components, while phase space reconstruction is used for nonlinear second time series components. The predicted second time series component is then subjected to inverse EMD to obtain the predicted wind speed for turbine startup. Based on the predicted wind speed, the optimal power of the wind turbine is predicted. This invention significantly reduces data processing volume while maintaining prediction accuracy, avoiding the need for large amounts of hardware for data processing, and combines the advantages of both linear and nonlinear prediction methods. Based on the predicted optimal power of the wind turbine, this invention performs charge and discharge control on the energy storage system of the wind farm to achieve power output smoothing in the maximum power point tracking mode of the wind turbine. This effectively realizes energy management of the grid-connected wind-storage integrated system, ensuring minimum wind curtailment while smoothing wind farm output fluctuations and reducing the impact of power system / load fluctuations.

[0040] Step 2 of this invention, obtaining the second time series component using the EMD decomposition method, includes the following steps:

[0041] Step 2.1: Extract all maximum and minimum points in the data {x(t)|t=1,2,…,n};

[0042] Step 2.2: Construct its upper and lower envelopes e using cubic interpolation. up (t) and e low (t), and their average value m(t);

[0043]

[0044] Step 2.3: Remove {m(t)} from the data to obtain new data ht;

[0045] ht=x(t)-m(t)

[0046] Step 2.4: Repeat steps 2.1 to 2.3k times until the termination condition is met, and obtain the first IMF component {h1t|t=1,2,…,n}, and denote {c1t}={h1t};

[0047] Step 2.5: Remove {c1t} from the data {x(t)} to obtain new data {x′(t)}. Repeat the above process until all IMFs (Intrinsic Modes) are decomposed.

[0048]

[0049] In the formula, IMF represents the intrinsic mode, r n (t) represents the residual.

[0050] In step 3, IMFs with obvious local linearity are linearized by local linearity discrimination and predicted by linear fitting; the phase space of nonlinear components is constructed and reconstructed based on the phase space for prediction.

[0051] The phase space of the nonlinear components is constructed using the following method.

[0052] For a univariate discrete sequence {x(t)|t=1,2,…,n}, construct the phase space using the time delay τ and the embedding dimension m:

[0053] X(t)=[x(t),x(t+τ),...,x(t+(m-1)τ)]

[0054] Where t=1,…,L,L=n-(m-1)τ, we obtain the phase space matrix:

[0055]

[0056] This invention obtains the second time series component through the above-mentioned EMD decomposition method; and constructs the phase space of the nonlinear component through the above method, further ensuring the accuracy of linear fitting prediction and reconstruction prediction based on phase space.

[0057] This invention addresses the following scenarios: When the predicted power exceeds the reference power (i.e., the given grid-connected power of the wind farm), if the energy state of the energy storage system is less than 90%, the wind turbine maintains maximum power point tracking (MPPT) mode, and the energy storage system enters a charging state. When the predicted power exceeds the reference power (i.e., the given grid-connected power of the wind farm), if the energy state of the energy storage system is greater than 90%, the wind turbine speed is limited, resulting in wind curtailment, and the energy storage system ceases charging and discharging operations. When the predicted power is less than the reference power (i.e., the given grid-connected power of the wind farm), if the energy state of the energy storage system is greater than 20%, the wind turbine maintains maximum power point tracking (MPPT) mode, and the energy storage system enters a discharging state. If the energy state of the energy storage system is less than 20%, the wind turbine maintains maximum power point tracking (MPPT) mode, and the energy storage system ceases charging and discharging operations. This invention, through the above-described charging and discharging control of the wind farm's energy storage system, achieves power output smoothing in maximum power point tracking (MPPT) mode, further ensuring minimal wind curtailment while mitigating wind farm output fluctuations and reducing the impact of power system / load fluctuations. Attached Figure Description

[0058] Figure 1 This is a diagram illustrating the architecture of the grid-connected wind-storage energy management system based on prediction, as described in this invention.

[0059] Figure 2This is a schematic diagram of the hybrid prediction method based on EMD decomposition and phase space reconstruction of the present invention.

[0060] Figure 3 This is a flowchart of the hybrid prediction method based on EMD decomposition and phase space reconstruction of the present invention;

[0061] Figure 4 This is a schematic diagram of the power curve of a wind turbine generator;

[0062] Figure 5 This is a schematic diagram of the wind turbine power signal feedback method;

[0063] Figure 6 This is a schematic diagram of the optimal power curve of the wind turbine;

[0064] Figure 7 This is a flowchart of the charging and discharging control of the energy storage system of a wind farm according to the present invention. Detailed Implementation

[0065] The present invention will now be described in detail with reference to the embodiments shown in the accompanying drawings. However, it should be noted that these embodiments are not intended to limit the present invention. Equivalent changes or substitutions in function, method, or structure made by those skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0066] Example 1:

[0067] Prediction-based grid-connected wind-storage energy management method

[0068] A prediction-based grid-connected wind-storage energy management method includes the following steps:

[0069] Step 1: Collect historical data related to wind turbine energy storage.

[0070] The historical data includes wind turbine data, meteorological data, and energy storage data.

[0071] Step 2: Obtain the second time series component from the first time series component based on historical data using the EMD (Empirical Mode Decomposition) method.

[0072] like Figure 2 As shown, the method model can be divided into four objects and three operations. The objects are the original time series, the original time series components (first time series component), the component prediction results, and the prediction results of the original series. The operations are decomposition, prediction, and reconstruction. The core idea is to decompose a complex signal into several relatively simple signals for prediction processing, thereby obtaining the prediction results of the original complex signal. The specific operations are as follows:

[0073] Object 1 (the original time series) is generated from historical data collected by the sampling module and stored in the database. The EMD decomposition method is used to decompose Object 1 (the complex original signal) into several Object 2 (relatively simple sequence components, i.e., the first time series components).

[0074] Includes the following steps:

[0075] Step 2.1: Extract all maximum and minimum points in the data {x(t)|t=1,2,…,n};

[0076] Step 2.2: Construct its upper and lower envelopes e using cubic interpolation. up (t) and e low (t), and their average value m(t);

[0077]

[0078] Step 2.3: Remove {m(t)} from the data to obtain new data ht;

[0079] ht=x(t)-m(t)

[0080] Step 2.4: Repeat steps 2.1 to 2.3k times until the termination condition is met, and obtain the first IMF component {h1t|t=1,2,…,n}, and denote {c1t}={h1t};

[0081] Step 2.5: Remove {c1t} from the data {x(t)} to obtain new data {x′(t)}. Repeat the above process until all IMFs are decomposed, i.e.

[0082]

[0083] In the formula, IMF represents the intrinsic mode, r n (t) represents the residual.

[0084] Step 3: Make predictions based on the second time series components. For linear second time series components, use linear fitting for prediction; for nonlinear second time series components, reconstruct and predict based on phase space.

[0085] like Figure 2 As shown, operation two (prediction) is performed on object 2. The IMF (Intrinsic Mode) with obvious local linearity is linearized by local linearity discrimination and linear fitting is used for prediction. The phase space of the nonlinear component is constructed and the prediction is reconstructed based on the phase space, finally obtaining object 3 (the predicted second time series component).

[0086] The phase space of the nonlinear components is constructed using the following method.

[0087] For a univariate discrete sequence {x(t)|t=1,2,…,n}, construct the phase space using the time delay τ and the embedding dimension m:

[0088] X(t)=[x(t),x(t+τ),...,x(t+(m-1)τ)]

[0089] Where t=1,…,L,L=n-(m-1)τ, we obtain the phase space matrix:

[0090]

[0091] The time delay τ and the embedding dimension m are key parameters in coordinate delay phase space reconstruction techniques. For an ideal sequence described by Takens' theorem, these two parameters can take any value. However, since the length n of a real sequence is finite and there is unavoidable noise, they need to be analyzed and determined.

[0092] By reconstructing the phase space of the wind speed sequence, a phase space that is consistent with the original one-dimensional dynamic system in a topological sense is obtained. In this way, the system state can be judged, analyzed and predicted within this phase space.

[0093] Step 4: As Figure 2 As shown, the predicted second time series component is subjected to inverse EMD operation to obtain the predicted wind speed for the start-up of the wind turbine unit.

[0094] like Figure 2 As shown, an anti-EMD operation (reconstruction) is performed on object 3 to reconstruct object 4, which is the prediction result of the original sequence. Figure 3 As shown, for IMF components with obvious linear characteristics, their local linearization is performed, and then the sequence of this part is predicted by least squares fitting. For IMF components with indistinct linear characteristics, a phase space is constructed based on historical data sequences, and reconstruction prediction is achieved within the phase space. Ultimately, a hybrid prediction of linear and nonlinear characteristics is achieved.

[0095] Step 5: Based on the predicted wind speed, predict the optimal power of the wind turbine; based on the predicted optimal power of the wind turbine, perform charge and discharge control on the energy storage system of the wind farm to achieve power output smoothing of the wind farm in the maximum power point tracking (MPPT) mode.

[0096] Wind turbine power prediction: Estimating wind turbine output based on predicted wind speed is mainly based on Euler's theory of wind turbine energy harvesting. The optimal power of the wind turbine is calculated according to the law of conservation of energy.

[0097]

[0098] Where, ρ a It is air density; A wIt is the area swept by the blade; C p The unit's power factor has a theoretical Betz limit of approximately 0.593; v is the turbine head wind speed, and its power curve is shown below. Figure 4 As shown, v0 is the starting wind speed of the wind turbine unit, and P limit Let C be the limiting power of the wind turbine unit. Based on Euler's theory, assuming constant air density and neglecting blade oscillation, then given a wind turbine model (i.e., given C...), p Therefore, given the wind speed at the turbine head, the maximum output power of the wind turbine generator can be determined.

[0099] Wind turbine MPPT control: The wind turbine MPPT control algorithm in this patent adopts the power signal feedback method. Given the optimal power curve of the wind turbine, the corresponding optimal active power is obtained after inputting the speed. The optimal active power is then used to directly calculate the reference value of the wind turbine's electromagnetic torque, and subsequently, the reference value of the D-axis rotor current. Its control diagram is shown below. Figure 5 As shown:

[0100] Mechanical torque is

[0101]

[0102] Among them, w m This refers to the mechanical rotation speed.

[0103] The tip speed ratio is

[0104] λ=w m R / v

[0105] The wind energy utilization coefficient is obtained analytically and is represented by an approximate nonlinear function as follows:

[0106]

[0107]

[0108] Where β is the blade pitch angle, assumed to be 0, and R represents the blade length. At a given wind speed, there exists a unique rotational speed that maximizes the wind energy conversion efficiency, i.e., the maximum power point (MPP). When wind speeds differ, such as... Figure 6 As shown, by adjusting the rotational speed, the tip speed ratio is kept optimal, thereby achieving the maximum power of the fan.

[0109] like Figure 7As shown, the energy management system estimates wind turbine output based on predicted wind speed, and then controls the charging and discharging of the wind farm's energy storage equipment to achieve power smoothing in the maximum power point tracking (MPPT) mode. This ensures minimal wind curtailment while mitigating wind farm output fluctuations and reducing the impact of power system / load fluctuations. To address the issue of wind farm output fluctuations, this patent proposes an energy management method with the objective of smoothing power fluctuations: estimating wind turbine output based on predicted wind speed, and then controlling the charging and discharging of the wind farm's energy storage equipment to achieve power smoothing in the maximum power point tracking (MPPT) mode. This ensures minimal wind curtailment while minimizing the impact of wind farm output fluctuations on the power system / load.

[0110] When the predicted power is greater than the reference power (i.e., the power of the given wind farm to the grid), if the energy state of the energy storage system is less than 90%, the wind turbine will maintain the maximum power point tracking mode, and the energy storage system will enter the charging state.

[0111] When the predicted power is greater than the reference power, i.e. the given power of the wind farm to the grid, if the energy state of the energy storage system is greater than 90%, the wind turbine speed is limited, i.e., the wind curtailment is implemented, and the energy storage system no longer performs charging and discharging operations.

[0112] When the predicted power is less than the reference power (i.e., the power of the given wind farm to the grid), if the energy state of the energy storage system is greater than 20%, the wind turbine will maintain the maximum power point tracking mode, and the energy storage system will enter the discharge state.

[0113] If the energy state of the energy storage system is less than 20%, the wind turbine will remain in maximum power point tracking mode, and the energy storage system will no longer perform charging and discharging operations.

[0114] Example 2:

[0115] A prediction-based grid-connected wind-storage energy management system

[0116] like Figure 1 As shown, the prediction-based grid-connected wind-storage energy management system includes:

[0117] The data acquisition module is used to collect historical data, including wind turbine data, meteorological data, and energy storage data. Each data acquisition module collects data from the wind turbine system, meteorological station, and energy storage (BMS) of the target wind farm through a communication system, stores the data in a database, and updates the data in real time.

[0118] The prediction module reads the latest data from the database, generates scheduling instructions, and transmits them to the database. The instruction execution device then runs according to the latest scheduling instructions. The prediction module is used to perform the following operations:

[0119] The first time series component based on historical data is decomposed using the EMD method to obtain the second time series component. Prediction is then performed based on the second time series component: linear fitting is used for linear second time series components, while nonlinear second time series components are reconstructed based on phase space. The predicted second time series component is then subjected to inverse EMD to obtain the predicted wind speed for turbine startup. Based on the predicted wind speed, the optimal power of the wind turbine is predicted. Finally, based on the predicted optimal power of the wind turbine, the charging and discharging control of the wind farm's energy storage system is implemented.

[0120] A scheduling instruction transmission module, which is used to transmit scheduling instructions generated by the prediction module;

[0121] A scheduling instruction execution device is used to execute scheduling instructions to achieve power output smoothing of the wind farm in the maximum power point tracking mode of the wind turbine.

[0122] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

[0123] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within the present invention.

Claims

1. A prediction-based grid-connected wind-storage energy management method, characterized in that, Includes the following steps: Step 1: Collect historical data related to wind turbine energy storage; Step 2: Obtain the second time series component from the first time series component based on historical data using the EMD decomposition method; Step 3: Make predictions based on the second time series components, which include both linear and nonlinear types; for linear second time series components, linear fitting is used for prediction; for nonlinear second time series components, reconstruction prediction is performed based on phase space. Step 4: Perform inverse EMD operation on the predicted second time series component to obtain the predicted wind speed for wind turbine startup; Step 5: Based on the predicted wind speed, predict the optimal power of the wind turbine; based on the predicted optimal power of the wind turbine, perform charge and discharge control on the energy storage system of the wind farm to achieve power output smoothing of the wind farm in the maximum power point tracking mode of the wind turbine.

2. The prediction-based grid-connected wind-storage energy management method according to claim 1, characterized in that: In step 1, the historical data includes wind turbine data, meteorological data, and energy storage data.

3. The prediction-based grid-connected wind-storage energy management method according to claim 2, characterized in that: Step 2 includes the following steps: Step 2.1: Extract Data The maximum and minimum points in the equation; where x(t) represents the t-th data point and n represents the number of data points extracted; Step 2.2: Construct its upper and lower envelopes using cubic interpolation. and and their average values ; Step 2.3: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require Remove data from the original data to obtain new data. ; Step 2.4: Repeat steps 2.1 through 2.

3. Once the termination condition is met, the first IMF component is obtained. ,remember ; Step 2.5: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require From the data Remove from the middle to obtain new data. Repeat the above process until all IMFs are extracted, i.e. In the formula, IMF represents intrinsic modes. r n ( t ) represents the residual. C i This represents the i-th IMF component.

4. The prediction-based grid-connected wind-storage energy management method according to claim 3, characterized in that: In step 3, IMFs with obvious local linearity are linearized by local linearity discrimination and predicted by linear fitting; the phase space of nonlinear components is constructed and reconstructed based on the phase space for prediction.

5. The prediction-based grid-connected wind-storage energy management method according to claim 4, characterized in that: In step 3, the phase space of the nonlinear components is constructed using the following method: For a univariate discrete sequence Through time delay and embedding dimension Construct phase space: in, , The phase space matrix is ​​obtained as follows: 。 6. The prediction-based grid-connected wind-storage energy management method according to claim 1, characterized in that, In step 5, when the predicted power is greater than the reference power (i.e., the power of the given wind farm to the grid), if the energy state of the energy storage system is less than 90%, the wind turbine will maintain the maximum power point tracking mode, and the energy storage system will enter the charging state.

7. The prediction-based grid-connected wind-storage energy management method according to claim 6, characterized in that: In step 5, when the predicted power is greater than the reference power, i.e. the given wind farm grid connection power, if the energy state of the energy storage system is greater than 90%, the wind turbine speed is limited, i.e., the wind curtailment state is entered, and the energy storage system no longer performs charging and discharging operations.

8. The prediction-based grid-connected wind-storage energy management method according to claim 7, characterized in that: In step 5, when the predicted power is less than the reference power, i.e. the given grid-connected power of the wind farm, if the energy state of the energy storage system is greater than 20%, the wind turbine will maintain the maximum power point tracking mode, and at the same time, the energy storage system will enter the discharge state.

9. The prediction-based grid-connected wind-storage energy management method according to claim 8, characterized in that: In step 5, if the energy state of the energy storage system is less than 20%, the wind turbine will remain in maximum power point tracking mode, and the energy storage system will no longer perform charging and discharging operations.

10. A prediction-based grid-connected wind-storage energy management system, characterized in that, include: The data acquisition module is used to collect historical data related to wind turbine energy storage. The prediction module is used to perform the following operations: obtaining a second time series component from a first time series component based on historical data using the EMD decomposition method; Prediction is made based on the second time series component, which includes both linear and nonlinear types. Linear fitting prediction is used for the linear second time series component. Reconstruction prediction is performed based on the phase space for the nonlinear second time series component. The predicted second time series component is then subjected to inverse EMD operation to obtain the predicted wind speed for the wind turbine startup. Based on the predicted wind speed, the optimal power of the wind turbine is predicted; based on the predicted optimal power of the wind turbine, the charging and discharging control of the energy storage system of the wind farm is performed. A scheduling instruction transmission module, which is used to transmit the scheduling instructions generated by the prediction module; A scheduling instruction execution device is used to execute scheduling instructions to achieve power output smoothing of the wind farm in the maximum power point tracking mode of the wind turbine.

Images