Offshore wind power generation strategy optimization method introducing SVM (Support Vector Machine) and multi-target nonlinear optimization

Abstract

The invention discloses an offshore wind power generation strategy optimization method introducing SVM and multi-target nonlinear optimization, and the method comprises the following steps: S100, obtaining offshore wind power generation meteorological data of a current time period, including a rated wind speed and a wind direction; s200, classifying the acquired meteorological data, and setting a meteorological condition suitable for power generation as' 1 'and other conditions as'-1 '; s300, inputting the classified data into an SVM (Support Vector Machine) and training a classification model of the data; s400, obtaining predicted meteorological data of a future time period, inputting the data into the trained SVM model, and classifying the meteorological data, so as to select a time period suitable for the operation of the offshore power generation facility; and S500, importing the classified data into a constraint model established according to a local power system, and calculating a power generation strategy in a future time period. According to the method, the optimal power generation time period is selected by classifying the predicted weather data, so that the time-sharing power generation amount of the offshore wind power generation cluster is predicted more accurately.

Description

Technical Field

[0001] This invention relates to an optimization method for offshore wind power generation strategies that incorporates SVM and multi-objective nonlinear optimization, belonging to the field of new energy and energy conservation technology. Background Technology

[0002] Offshore wind power is a power generation technology that has emerged in recent years and is one of the core directions of renewable energy development in China. As this technology matures, its share in the power supply system is increasing. However, due to the influence of natural environmental factors, changes in wind speed or direction at the location of offshore wind turbines often make it difficult to predict whether the power output can meet the load requirements, posing a potential threat to the stability of the power grid.

[0003] Currently, the power generation end of a power system often consists of a combination of various power generation facilities. Therefore, when the power generation quality of wind power does not meet the requirements, other equipment can be sought to meet the load-side requirements. However, this strategy will limit the growth of new energy output in the power system to some extent. Therefore, how to improve the utilization rate of new energy systems while ensuring the stability of the power system is an urgent problem to be solved. Summary of the Invention

[0004] The purpose of this invention is to provide an optimization method for offshore wind power generation strategy that incorporates SVM and multi-objective nonlinear optimization. By classifying predicted weather data, the optimal power generation time period is selected, thereby enabling more accurate prediction of the time-of-use power generation of offshore wind power clusters and reducing its impact on power quality.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] An optimization method for offshore wind power generation strategies, incorporating SVM and multi-objective nonlinear optimization, includes the following steps:

[0007] S100: Acquire meteorological data for offshore wind power generation during the current time period, including rated wind speed and wind direction;

[0008] S200: Classify the acquired meteorological data, marking suitable meteorological conditions for power generation as "1" and other conditions as "-1";

[0009] Suitable meteorological conditions mainly involve specific requirements for wind energy quality, which can be summarized in three aspects: average wind speed, wind power density, and wind direction. Generally, meteorological conditions with an annual average wind speed of 5 m / s or higher, a wind power density of 300 W / m² or higher, and a stable wind direction are suitable for power generation.

[0010] S300: Input the classified data into the SVM and train its classification model;

[0011] S400: Obtain forecast meteorological data for future periods, input this data into a trained SVM model to classify the meteorological data, and thus select the time period suitable for the operation of offshore power generation facilities; the more concentrated and prolonged the occurrence of "1" in step S200, the more suitable it is for power generation.

[0012] S500: Import the categorized data into a constraint model established based on the local power system to calculate the power generation strategy for future periods.

[0013] In the aforementioned offshore wind power generation strategy optimization method incorporating SVM and multi-objective nonlinear optimization, step S200 processes the meteorological data into a training dataset in a feature space, which takes the following form:

[0014] T = {(x1,y1),(x2,y2),...,(x N ,y N )}

[0015] in y i ∈{+1,-1}, i=1,2,...,N, where x i Let x be the i-th feature vector. This vector can be adjusted according to the requirements of meteorological data. At this stage, only the starting wind speed, rated wind speed, and cut-off wind speed are considered. Therefore, x i y is a three-dimensional feature vector; i As a class marker, the meteorological conditions that enable power generation are denoted as +1, and the other meteorological conditions are denoted as -1.

[0016] In the aforementioned offshore wind power generation strategy optimization method incorporating SVM and multi-objective nonlinear optimization, the training of the SVM model in step S300 follows the principles described below.

[0017]

[0018] Where α i C is a Lagrange multiplier, and C is a penalty parameter;

[0019] Based on the above formula, construct a convex quadratic programming problem and obtain α. i The optimal solution is

[0020] According to α * calculate Where ω * For x in the hyperpartial expression i Parameters;

[0021] From α * Choose any parameter that meets the above conditions and substitute it into the calculation. Thus, the expression for the hyperplane is obtained as: ω* ·x+b * =0

[0022] The final decision function is: f(x) = sign(ω) * ·x+b * ).

[0023] In the aforementioned offshore wind power generation strategy optimization method that incorporates SVM and multi-objective nonlinear optimization, step S400 inputs the predicted meteorological data for a future period into the trained SVM model for classification, and filters out the time period suitable for the operation of wind power generation facilities. This time period will be represented by the start-stop state variables in the constraint model, where "1" represents the operating state and "0" represents the shut-off state.

[0024] In the aforementioned offshore wind power generation strategy optimization method incorporating SVM and multi-objective nonlinear optimization, step S500 establishes a constraint model based on the regional power generation facility configuration. Building upon offshore wind power and thermal power, it configures thermal power, wind power…hydropower…according to regional conditions to meet local load requirements. The objective formula for this model is…

[0025]

[0026] Where c g For the cost of thermal power generation, p g,t Let c be the power generation of thermal power facility g at time t. w For the cost of generating electricity from offshore wind power, p w,t Let c be the power generation of the wind power facility w at time t. s p represents the cost of generating electricity from hydroelectric power. s,t Let be the amount of electricity generated by hydropower facility s at time t;

[0027] Power conservation constraint is

[0028]

[0029] Where u w,t Let p be the start-up / shutdown variable of the wind turbine generator w at time t. This variable is determined by the classification results of SVM, which shows the period of continuous operation. Based on this, the optimal operating time is further determined through the above constraints. l,t Let be the power of load l at time t.

[0030] In the aforementioned offshore wind power generation strategy optimization method incorporating SVM and multi-objective nonlinear optimization, the constraints on power generation facilities in step S500 constraint model are as follows:

[0031] 0≤p w,t ≤p w,t,max

[0032]

[0033] 0≤s u,g,t ≤min(p g,max -p g,t ,v g,up ×t)

[0034] 0≤s d,g,t ≤min(p g,t -p g,min ,v g,down ×t)

[0035] u g,t p g,t,min ≤p g,t ≤u g,t p g,t,max

[0036] u g,t Q g,t,min ≤Qg ,t ≤u g,t Q g,t,max

[0037] p g,t -p g,t-1 ≤tv g,up

[0038] p g,t-1 -p g,t ≤tv g,down

[0039] Where p w,t,max For the maximum active power of the wind power cluster w, s u,g,t and s d,g,t For the positive and negative rotational reserve constraints of thermal power facilities, w u and w d These are the integrated wind power demand coefficients for positive and negative rotational reserves, respectively. For the integrated power of the wind power cluster w, v g,up and v g,down These represent the rate of increase and rate of decrease of thermal power output, u g,t These are the start-up and shutdown indication parameters for thermal power facilities.

[0040] Compared with existing technologies, this invention selects the optimal power generation time period by classifying predicted weather data, thereby achieving more accurate prediction of the time-of-use power generation of offshore wind power clusters and reducing its impact on power quality.

[0041] By incorporating meteorological data, the accuracy of wind power forecasting is improved in the short term. Simultaneously, this approach optimizes dispatch planning and reduces the overall operating costs of thermal power, hydropower, and energy storage facilities. To a certain extent, this approach enhances the stability of the power system, reducing the probability of large-scale grid disconnections from offshore wind turbines through meteorological data forecasting. Attached Figure Description

[0042] Figure 1 This is a flowchart of one embodiment of the present invention.

[0043] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Detailed Implementation

[0044] Embodiment 1 of the present invention: An optimization method for offshore wind power generation strategy by introducing SVM and multi-objective nonlinear optimization, characterized by comprising the following steps:

[0045] S100: Acquire meteorological data for offshore wind power generation during the current time period, including rated wind speed and wind direction;

[0046] S200: Classify the acquired meteorological data, assigning "1" to meteorological conditions suitable for power generation and "-1" to all other conditions; the specific method is as follows: The meteorological data is processed into a training dataset in a feature space, in the following form:

[0047] T = {(x1,y1),(x2,y2),...,(x N ,y N )}

[0048] in y i ∈{+1,-1}, i=1,2,...,N, where x i Let x be the i-th feature vector. This vector can be adjusted according to the requirements of meteorological data. At this stage, only the starting wind speed, rated wind speed, and cut-off wind speed are considered. Therefore, x i y is a three-dimensional feature vector; i As a class marker, the meteorological conditions that enable power generation are denoted as +1, and the other meteorological conditions are denoted as -1.

[0049] S300: Input the classified data into the SVM and train its classification model; the training of the SVM model follows these principles.

[0050]

[0051] Where α i C is a Lagrange multiplier, and C is a penalty parameter;

[0052] Based on the above formula, construct a convex quadratic programming problem and obtain α. i The optimal solution is

[0053] According to α * calculate Where ω * For x in the hyperpartial expression i Parameters;

[0054] From α * Choose any parameter that meets the above conditions and substitute it into the calculation. Thus, the expression for the hyperplane is obtained as: ω * ·x+b * =0

[0055] The final decision function is: f(x) = sign(ω) * ·x+b * ).

[0056] S400: Obtain forecast meteorological data for future periods, input this data into a trained SVM model to classify the meteorological data, input forecast meteorological data for a future period into a trained SVM model for classification, and filter out the time periods suitable for the operation of wind power generation facilities. This time period will be represented by the start-stop state variables in the constraint model, where "1" represents the operating state and "0" represents the shut-off state.

[0057] S500: Import the categorized data into a constraint model established based on the local power system to calculate the power generation strategy for future periods. The specific method is as follows: Establish a constraint model based on the regional power generation facility configuration. Building upon offshore wind power and thermal power, configure thermal power, wind power…and hydropower according to local conditions to meet local load requirements. The objective formula for this model is…

[0058]

[0059] Where c g For the cost of thermal power generation, p g,t Let c be the power generation of thermal power facility g at time t. w For the cost of generating electricity from offshore wind power, p w,t Let c be the power generation of the wind power facility w at time t. s p represents the cost of generating electricity from hydroelectric power. s,t Let be the amount of electricity generated by hydropower facility s at time t;

[0060] Power conservation constraint is

[0061]

[0062] Where uw,t Let p be the start-up / shutdown variable of the wind turbine generator w at time t. This variable is determined by the classification results of SVM, which shows the period of continuous operation. Based on this, the optimal operating time is further determined through the above constraints. l,t Let be the power of load l at time t.

[0063] The constraints on power generation facilities in the constraint model are:

[0064] 0≤p w,t ≤p w,t,max

[0065]

[0066] 0≤s u,g,t ≤min(p g,max -p g,t ,v g,up ×t)

[0067] 0≤s d,g,t ≤min(p g,t -p g,min ,v g,down ×t)

[0068] u g,t p g,t,min ≤p g,t ≤u g,t p g,t,max

[0069] u g,t Q g,t,min ≤Q g,t ≤u g,t Q g,t,max

[0070] p g,t -p g,t-1 ≤tv g,up

[0071] p g,t-1 -p g,t ≤tv g,down

[0072] Where p w,t,max For the maximum active power of the wind power cluster w, s u,g,t and s d,g,t For the positive and negative rotational reserve constraints of thermal power facilities, w u and w d These are the integrated wind power demand coefficients for positive and negative rotational reserves, respectively. For the integrated power of the wind power cluster w, v g,up and v g,down These represent the rate of increase and rate of decrease of thermal power output, u g,tThese are the start-up and shutdown indication parameters for thermal power facilities.

[0073] Tests have shown that the optimized method of this invention can improve the overall reliability of the power grid, prevent large-scale disconnection of wind turbines from the grid, and significantly improve power generation efficiency.

Claims

1. A method for offshore wind power generation strategy optimization by introducing SVM and multi-objective nonlinear optimization, characterized in that, The method comprises the following steps: S100: Obtain offshore wind power meteorological data of a current time period, including rated wind speed and wind direction; S200: Classify the obtained meteorological data, and count meteorological conditions suitable for power generation as "1", and count the rest as "-1"; S300: Input the classified data into an SVM and train a classification model thereof; S400: Obtain predicted meteorological data of a future time period, input the data into the trained SVM model, classify the meteorological data, and thus select a time period suitable for operation of offshore power generation facilities; S500: Import the classified data into a constraint model established according to a local power system, and calculate a power generation strategy of the future time period.

2. The offshore wind power generation strategy optimization method incorporating SVM and multi-objective nonlinear optimization according to claim 1, characterized in that, In step S200, the meteorological data is processed into a training data set in a feature space, and the form is as follows: T = {(xl,yl), (x2,y2),..., (xn,yn)} N N )}​ wherein y i ∈{+1,-1},i=1,2,...,N, wherein x i is the ith feature vector, which is adjusted according to the requirements of the meteorological data, taking into account the cut-in wind speed, the rated wind speed and the cut-out wind speed, x i is a three-dimensional feature vector; y i is the class label, noting that the generating meteorological conditions are marked +1 and the remaining meteorological conditions are marked -1.

3. The offshore wind power generation strategy optimization method incorporating SVM and multi-objective nonlinear optimization of claim 1, wherein, In step S300, the training of the SVM model follows the following principles where a i is a Lagrange multiplier and C is a penalty parameter; The optimal solution of a is obtained by constructing a convex quadratic programming problem according to the above formula i ​ According to a * Computing where ω * is a parameter of the super-tangent expression for x i . From the α * The parameter that meets the above conditions is brought into the calculation So the expression of the hyperplane is: ω * ·x+b * = 0 The final decision function is: f(x) = sign(ω * ·x + b * ).

4. The offshore wind power generation strategy optimization method incorporating SVM and multi-objective nonlinear optimization of claim 1, wherein, In step S400, the predicted meteorological data of a future time period is input into the trained SVM model for classification, and a time period suitable for operation of wind power generation facilities is selected. The time period is represented by a start-stop state variable in the constraint model, and "1" represents an operation state and "0" represents a shutdown state.

5. The offshore wind power generation strategy optimization method incorporating SVM and multi-objective nonlinear optimization according to claim 1, characterized in that, In step S500, a constraint model is established according to the configuration of regional power generation facilities. On the basis of offshore wind power and thermal power, thermal power, wind power, and hydroelectric power are configured according to the regional situation to meet the load requirements of the region. The objective formula of the model is wherein c g is the cost of thermal power generation, p g,t is the power generation of the thermal power facility g at time t, c w is the cost of offshore wind power generation, p w,t is the power generation of the wind power facility w at time t, c s is the cost of hydroelectric power generation, p s,t is the power generation of the hydroelectric power facility s at time t; The power conservation constraint is where u w,t is the start-stop variable of the wind turbine w at time t, which is derived from the classification result of the SVM for the time period during which the wind turbine can be operated continuously, and on the basis of which the optimal operating time is further determined by the above constraints, p l,t is the power of the load l at time t.

6. The offshore wind power generation strategy optimization method incorporating SVM and multi-objective nonlinear optimization of claim 5, wherein, In step S500, the constraint model of the power generation facility is 0 < p w,t ≤ p w,t,max 0 ≤ s u,g,t ≤ min(p g,max -p g,t ,v g,up × t) 0 ≤ s d,g,t ≤ min(p g,t -p g,min ,v g,down × t) u g,t p g,t,min ≤p g,t ≤u g,t p g,t,max u g,t Q g,t,min ≤Q g,t ≤u g,t Q g,t,max p g,t -p g,t-1 ≤tv g,up p g,t-1 -p g,t ≤tv g,down where p w,t,max is the maximum active power of the wind power cluster w, s u,g,t and s d,g,t are the positive and negative spinning reserve constraints of the thermal power plant, w u and w d are the wind power integration demand factors of the positive and negative spinning reserve, respectively, is the integrated power of the wind power cluster w, v g,up and v g,down are the ramp-up and ramp-down rates of the thermal power output, u g,t is the start-stop indication parameter of the thermal power plant operation.

Images