Wind power plant multi-target collaborative optimization scheduling method and device based on digital twinning
By optimizing the yaw angle control of wind farms using digital twin technology, the problem of improper operation of wind turbines in different wind power scenarios has been solved, improving power generation efficiency and power output, and reducing the impact of wake interference.
Patent Information
- Application Number
- CN202411051458.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-01
AI Technical Summary
In existing technologies, wind turbines in wind farms operate with fixed yaw angle parameters, which can lead to malfunctions or reduced power generation in certain wind power scenarios. Furthermore, wake interference causes significant power loss, affecting power generation efficiency.
By adopting a multi-objective collaborative optimization scheduling method for wind farms based on digital twins, wind turbines are grouped according to wind direction and wind speed data of the wind farm, typical scenarios are generated, and yaw angle control parameters are optimized using reinforcement learning and distributed predictive control algorithms to construct a digital twin model of the wind farm and achieve real-time optimized scheduling.
It improves the power generation efficiency and output of wind farms, reduces power loss caused by wake interference, and optimizes the energy distribution and operating efficiency of wind farms.
Smart Images

Figure CN118971195A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of wind farm management and control, and in particular to a wind farm multi-objective collaborative optimization scheduling method and device based on digital twins. Background Art
[0002] Wind power generation is widely used due to its clean and renewable advantages. The number and scale of wind farms around the world are increasing rapidly, making great contributions to energy utilization and environmental protection. Therefore, more and more wind farms are put into use. In the prior art, when generating electricity through a wind farm, several wind turbines deployed in the wind farm are usually controlled to operate with fixed yaw angle parameters. That is to say, no matter in which power generation environment and wind power scenario, the wind turbines work with the same yaw angle control strategy to output electricity to achieve power generation.
[0003] However, the study found that if all wind turbines operate with the same yaw angle control strategy in any power generation environment and wind power scenario, some wind turbines may not be able to operate normally at their specific yaw angles in certain wind power scenarios, or they may not be able to achieve maximum power output at their specific yaw angles, resulting in reduced power generation efficiency of the wind farm. In addition, when a large number of wind turbines are operating at the same time, there is usually power loss caused by wake interference, which will also lead to a reduction in wind farm power generation. Summary of the invention
[0004] In view of this, the object of the present invention is to provide a wind farm multi-objective collaborative optimization scheduling method and device based on digital twins to improve the power generation efficiency and power generation of the wind farm.
[0005] In a first aspect, an embodiment of the present application provides a wind farm multi-objective collaborative optimization scheduling method based on digital twins, the method comprising:
[0006] The dominant wind direction of the wind farm is divided according to the wind direction data of the wind farm, and the wind turbines in the wind farm are grouped according to the wake characteristics under the dominant wind direction to obtain a plurality of wind turbine groups;
[0007] Generating a plurality of wind power scenarios of the wind farm according to the wind speed data of the wind farm; performing scenario reduction on the generated plurality of wind power scenarios to obtain a typical scenario of the wind farm;
[0008] For each typical scenario, the pre-dispatch instructions are determined based on the reinforcement learning algorithm so that the output power and unit load of the wind farm digital twin model in the typical scenario meet the objective function;
[0009] The pre-scheduling instruction is used as the initial value of the real-time optimization scheduling, and the control parameters are optimized using a distributed predictive control algorithm;
[0010] In this typical scenario, the operation of each wind turbine is controlled based on the optimized control parameters.
[0011] Optionally, dividing the dominant wind direction of the wind farm according to the wind direction data of the wind farm, and grouping the wind turbines in the wind farm according to the wake characteristics under the dominant wind direction to obtain a plurality of wind turbine groups, including:
[0012] Using a hierarchical clustering method to divide the dominant wind direction from the wind direction data;
[0013] Under the dominant wind direction, combined with the specific arrangement of the wind farm, the wind turbines are regarded as nodes of a directed graph, and the wake effects of the upstream turbines on the downstream turbines are regarded as edges of the directed graph;
[0014] Based on the directed graph established above, a community discovery algorithm is used to describe the community structure of the wind farm.
[0015] Optionally, generating a plurality of wind power scenarios of the wind farm according to the wind speed data of the wind farm includes:
[0016] A randomly generated noise signal is used as an input random vector, and the measured wind speed data of the wind farm is used as sample data. A generative adversarial neural network is used to generate scenarios to obtain several wind power scenarios.
[0017] Optionally, the reducing the generated wind power scenarios to obtain a typical scenario of the wind farm includes:
[0018] Using an unsupervised clustering method to reduce the generated wind power scenarios to obtain a typical scenario of the wind farm;
[0019] Optionally, for each typical scenario, determining, based on a reinforcement learning algorithm, a pre-dispatching instruction that enables the output power and unit load of the wind farm digital twin model in the typical scenario to meet an objective function includes:
[0020] With output power and unit load as the objective function, and the load of key components of the wind turbine and the limit angle of yaw control as the constraint conditions, an incentive mechanism is established;
[0021] According to the reward mechanism, pre-dispatching instructions for each wind turbine are obtained through self-learning, thereby realizing pre-dispatching of yaw control of the wind farm.
[0022] Optionally, the constraint conditions include an ultimate load limit and a yaw angle limit of the wind turbine generator set;
[0023] The extreme value distribution theory is used to simulate the limit load limit, and the yaw angle limit condition is that the unit load does not exceed the limit load within the limit angle range.
[0024] Optionally, the using the pre-scheduling instruction as the initial value of the real-time optimization scheduling and optimizing the control parameters by using a distributed predictive control algorithm includes:
[0025] When optimizing the control parameters, the constraint conditions are used as constraints in the optimization process, and the objective function is used as the optimization target. The overall optimization calculation of the wind farm is divided into parallel optimization calculations of multiple wind turbine subsystems, and the yaw scheduling instructions of each wind turbine are obtained in real time to achieve optimization of the control parameters.
[0026] Optionally, the method comprises:
[0027] The input wind speed is soft-measured through the unit status, and verified with the lidar wind measurement data to build a digital twin model of the wind farm inflow wind speed;
[0028] The propagation law of the fluid in front of the computer is calculated, and a digital twin model of the wind speed in front of the machine is constructed according to the input wind speed of the wind farm;
[0029] The finite element modeling method is used to combine blade element theory and momentum theory to obtain the aerodynamic characteristics of the wind turbine. The multi-body dynamics method is used to calculate the structural response of the wind turbine to build a single-machine digital twin model.
[0030] Construct a digital twin model of wake characteristics based on the dynamic meandering principle;
[0031] The digital twin model of the wind farm is constructed by combining the wind farm inflow wind speed digital twin model, the front-machine wind speed digital twin model, the single-machine digital twin model, the wake characteristic digital twin model and the wind farm-level energy management twin model.
[0032] Optionally, the method further comprises:
[0033] The data with the highest correlation with the radar wind measurement data in the unit state parameters are selected as input data, the radar wind measurement data are used as output data, and a deep learning neural network method is used to construct the digital twin model of the inflow wind speed of the wind farm;
[0034] A neutral atmospheric boundary layer without a wind turbine is generated by using a large eddy simulation method, a wind turbine is placed in the neutral atmospheric boundary layer, and the interaction between the wind turbine and the atmospheric boundary layer is studied by using a LES coupled actuation line model method, thereby constructing the wind speed digital twin model in front of the wind turbine;
[0035] The blade element momentum theory is used to analyze the aerodynamic performance of the impeller, and the single-machine digital twin model is constructed based on the aerodynamic performance analysis results;
[0036] A passive tracer model of the wake is established based on a dynamic meandering wake model. Based on the evolution of the wake loss, the meandering of the wake, the increase in turbulence intensity caused by the wake, the drop in wind speed near the wake area and the radial expansion of the wake, the passive tracer model is calibrated to obtain the digital twin model of the wake characteristics.
[0037] The wind farm-level energy management twin model is used to dynamically adjust the power output of each wind turbine according to the real-time conditions of the wind farm and the grid demand, so as to optimize the energy distribution and operation efficiency of the entire wind farm.
[0038] In a second aspect, an embodiment of the present application provides a digital twin wind farm multi-objective collaborative optimization scheduling device, the device comprising:
[0039] A wind turbine group grouping module is used to divide the dominant wind direction of the wind farm according to the wind direction data of the wind farm; and to group the wind turbines in the wind farm according to the wake characteristics under the dominant wind direction to obtain a number of wind turbine group groups;
[0040] A typical scene division module is used to generate a number of wind power scenes of the wind farm according to the wind speed data of the wind farm; and perform scene reduction on the generated several wind power scenes to obtain a typical scene of the wind farm;
[0041] A pre-dispatching instruction determination module is used to determine, for each typical scenario, based on a reinforcement learning algorithm, a pre-dispatching instruction that enables the output power and unit load of the wind farm digital twin model in the typical scenario to meet the objective function;
[0042] A control parameter optimization module, used to use the pre-scheduling instruction as the initial value of the real-time optimization scheduling and optimize the control parameters using a distributed predictive control algorithm;
[0043] The wind turbine control module is used to control the operation of each wind turbine based on the optimized control parameters in this typical scenario.
[0044] Optionally, dividing the dominant wind direction of the wind farm according to the wind direction data of the wind farm, and grouping the wind turbines in the wind farm according to the wake characteristics under the dominant wind direction to obtain a plurality of wind turbine groups, including:
[0045] Using a hierarchical clustering method to divide the dominant wind direction from the wind direction data;
[0046] Under the dominant wind direction, combined with the specific arrangement of the wind farm, the wind turbines are regarded as nodes of a directed graph, and the wake effects of the upstream turbines on the downstream turbines are regarded as edges of the directed graph;
[0047] Based on the directed graph established above, a community discovery algorithm is used to describe the community structure of the wind farm.
[0048] Optionally, generating a plurality of wind power scenarios of the wind farm according to the wind speed data of the wind farm includes:
[0049] A randomly generated noise signal is used as an input random vector, and the measured wind speed data of the wind farm is used as sample data. A generative adversarial neural network is used to generate scenarios to obtain several wind power scenarios.
[0050] Optionally, the reducing the generated wind power scenarios to obtain a typical scenario of the wind farm includes:
[0051] Using an unsupervised clustering method to reduce the generated wind power scenarios to obtain a typical scenario of the wind farm;
[0052] Optionally, for each typical scenario, determining, based on a reinforcement learning algorithm, a pre-dispatching instruction that enables the output power and unit load of the wind farm digital twin model in the typical scenario to meet an objective function includes:
[0053] With output power and unit load as the objective function, and the load of key components of the wind turbine and the limit angle of yaw control as the constraint conditions, an incentive mechanism is established;
[0054] According to the reward mechanism, pre-dispatching instructions for each wind turbine are obtained through self-learning, thereby realizing pre-dispatching of yaw control of the wind farm.
[0055] Optionally, the constraint conditions include an ultimate load limit and a yaw angle limit of the wind turbine generator set;
[0056] The extreme value distribution theory is used to simulate the ultimate load limit, and the yaw angle limit condition is that the unit load does not exceed the ultimate load within the limited angle range.
[0057] Optionally, the using the pre-scheduling instruction as an initial value of the real-time optimization scheduling and optimizing the control parameters using a distributed predictive control algorithm includes:
[0058] When optimizing the control parameters, the constraint conditions are used as constraints in the optimization process, and the objective function is used as the optimization target. The overall optimization calculation of the wind farm is divided into parallel optimization calculations of multiple wind turbine subsystems, and the yaw scheduling instructions of each wind turbine are obtained in real time to achieve optimization of the control parameters.
[0059] Optionally, the device comprises a model building module, wherein the model building module is used to:
[0060] The input wind speed is soft-measured through the unit status, and verified with the lidar wind measurement data to build a digital twin model of the wind farm inflow wind speed;
[0061] The propagation law of the fluid in front of the computer is calculated, and a digital twin model of the wind speed in front of the machine is constructed according to the input wind speed of the wind farm;
[0062] The finite element modeling method is used to combine blade element theory and momentum theory to obtain the aerodynamic characteristics of the wind turbine. The multi-body dynamics method is used to calculate the structural response of the wind turbine to build a single-machine digital twin model.
[0063] Construct a digital twin model of wake characteristics based on the dynamic meandering principle;
[0064] The digital twin model of the wind farm is constructed by combining the wind farm inflow wind speed digital twin model, the front-machine wind speed digital twin model, the single-machine digital twin model, the wake characteristic digital twin model and the wind farm-level energy management twin model.
[0065] Optionally, the model building module is specifically used to:
[0066] The data with the highest correlation with the radar wind measurement data in the unit state parameters are selected as input data, the radar wind measurement data are used as output data, and a deep learning neural network method is used to construct the digital twin model of the inflow wind speed of the wind farm;
[0067] A neutral atmospheric boundary layer without a wind turbine is generated by using a large eddy simulation method, a wind turbine is placed in the neutral atmospheric boundary layer, and the interaction between the wind turbine and the atmospheric boundary layer is studied by using a LES coupled actuation line model method, thereby constructing the wind speed digital twin model in front of the wind turbine;
[0068] The blade element momentum theory is used to analyze the aerodynamic performance of the impeller, and the single-machine digital twin model is constructed based on the aerodynamic performance analysis results;
[0069] A passive tracer model of the wake is established based on a dynamic meandering wake model. Based on the evolution of the wake loss, the meandering of the wake, the increase in turbulence intensity caused by the wake, the drop in wind speed near the wake area and the radial expansion of the wake, the passive tracer model is calibrated to obtain the digital twin model of the wake characteristics.
[0070] The wind farm-level energy management twin model is used to dynamically adjust the power output of each wind turbine according to the real-time conditions of the wind farm and the grid demand, so as to optimize the energy distribution and operation efficiency of the entire wind farm.
[0071] In a third aspect, an embodiment of the present application provides a computer device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the computer device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the multi-objective collaborative optimization scheduling method for wind farms based on digital twins are performed in any one of the optional implementations of the first aspect above.
[0072] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the multi-objective collaborative optimization scheduling method for wind farms based on digital twins are executed in any optional implementation manner in the first aspect above.
[0073] The technical solution provided by this application includes but is not limited to the following beneficial effects:
[0074] The present application first divides the dominant wind direction of the wind farm according to the wind direction data of the wind farm, and groups the wind turbines in the wind farm according to the wake characteristics under the dominant wind direction to obtain a number of wind turbine groups; generates a number of wind power scenarios of the wind farm according to the wind speed data of the wind farm; and performs scenario reduction on the generated several wind power scenarios to obtain a typical scenario of the wind farm; in the above steps, by grouping the wind turbines in the wind farm to obtain a number of wind turbine groups, it is possible to achieve unified optimization of the control strategies of the associated wind turbines, and by dividing the typical wind power scenarios of the wind farm, provide a basis for the subsequent optimization of the yaw angles in each wind power scenario.
[0075] Then, for each typical scenario, the pre-dispatching instructions are determined based on the reinforcement learning algorithm so that the output power and unit load of the digital twin model of the wind farm in the typical scenario meet the objective function; the pre-dispatching instructions are used as the initial value of the real-time optimization scheduling, and the control parameters are optimized using the distributed predictive control algorithm. The yaw angle control parameters that are adapted to the scenario can be determined for different wind power scenarios, thereby improving the adaptability and pertinence of the yaw angle control parameters.
[0076] Finally, in this typical scenario, controlling the operation of each wind turbine based on the optimized control parameters can avoid the power loss caused by the diversity and complexity of wind power scenarios and the mutual interference of wakes between wind turbines.
[0077] By adopting the above scheme, based on the wake interference situation of the wind farm, the yaw angle control parameters of the wind farm are optimized for different wind power scenarios, and then the operation of each wind turbine is controlled based on the optimized control parameters, which can reduce the loss of power generation in the wind farm and improve the power generation efficiency and power generation of the wind farm.
[0078] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0079] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.
[0080] Figure 1 A flowchart of a wind farm multi-objective collaborative optimization scheduling method based on digital twins provided in the first embodiment of the present invention is shown;
[0081] Figure 2 A flow chart of a method for grouping wind turbine groups provided in the first embodiment of the present invention is shown;
[0082] Figure 3 A flowchart of a method for generating a pre-scheduling instruction provided in the first embodiment of the present invention is shown;
[0083] Figure 4 A schematic diagram of the structure of a wind farm multi-objective collaborative optimization scheduling device based on digital twins provided in the second embodiment of the present invention is shown;
[0084] Figure 5 A schematic diagram of the structure of a computer device provided in Embodiment 3 of the present invention is shown. DETAILED DESCRIPTION
[0085] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.
[0086] Embodiment 1
[0087] To facilitate understanding of this application, Figure 1The flowchart of a digital twin wind farm multi-objective collaborative optimization scheduling method provided in the first embodiment of the present invention is described in detail to explain the first embodiment of the present application.
[0088] See also Figure 1 As shown, Figure 1 A flowchart of a wind farm multi-objective collaborative optimization scheduling method based on digital twins provided in the first embodiment of the present invention is shown, wherein the method comprises steps S101 to S105:
[0089] S101: dividing the dominant wind direction of the wind farm according to wind direction data of the wind farm, and grouping the wind turbines in the wind farm according to wake characteristics under the dominant wind direction to obtain a plurality of wind turbine groups.
[0090] Specifically, a hierarchical clustering algorithm is used to divide the dominant wind direction. Under the dominant wind direction, a community discovery algorithm (such as Louvain algorithm, Girvan-Newman algorithm, spectral clustering algorithm, etc.) is used to describe the community structure in the directed graph of the wind farm, and the wind turbines in the wind farm are grouped to obtain several wind turbine groups.
[0091] S102: generating a plurality of wind power scenarios of the wind farm according to the wind speed data of the wind farm; and performing scenario reduction on the generated plurality of wind power scenarios to obtain a typical scenario of the wind farm.
[0092] Specifically, data-driven neural network models (such as variational autoencoders, long short-term memory neural networks, generative adversarial neural networks, etc.) are used for scenario generation, and unsupervised clustering methods (including but not limited to self-organizing neural network SOM, K-means++ algorithm, Gaussian mixture model GMM, etc.) are used for scenario reduction to extract typical scenarios of wind farm output.
[0093] S103: For each typical scenario, determine the pre-dispatching instructions based on the reinforcement learning algorithm so that the output power and unit load of the wind farm digital twin model in the typical scenario meet the objective function.
[0094] Specifically, a reward mechanism is established with output power and unit load as objective functions, and with loads of key components of wind turbines and limit angles of yaw control as constraints; based on the reward mechanism, pre-dispatching instructions for each wind turbine are obtained through self-learning to achieve pre-dispatching of yaw control of wind farms.
[0095] S104: Using the pre-scheduling instruction as the initial value of the real-time optimization scheduling, and optimizing the control parameters using a distributed predictive control algorithm.
[0096] Specifically, when optimizing the control parameters, the constraint conditions are used as the constraints in the optimization process, and the objective function is used as the optimization goal. The overall optimization calculation of the wind farm is divided into parallel optimization calculations of multiple wind turbine subsystems, and the yaw scheduling instructions of each wind turbine are obtained in real time to achieve the optimization of the control parameters.
[0097] S105: In this typical scenario, the operation of each wind turbine is controlled based on the optimized control parameters.
[0098] Specifically, each wind turbine is controlled to operate at a yaw angle with an optimized control parameter, so as to achieve minimum loss and maximum power generation in each typical scenario.
[0099] In an alternative embodiment, see Figure 2 As shown, Figure 2 A flow chart of a method for grouping wind turbines provided in the first embodiment of the present invention is shown, wherein the dominant wind direction of the wind farm is divided according to the wind direction data of the wind farm, and the wind turbines in the wind farm are grouped according to the wake characteristics under the dominant wind direction to obtain a plurality of wind turbine groups, including steps S201 to S202:
[0100] S201: using a method based on hierarchical clustering to divide the dominant wind direction from the wind direction data;
[0101] S202: Under the dominant wind direction, in combination with the specific arrangement of the wind farm, the wind turbines are regarded as nodes of a directed graph, and the wake effects of the upstream turbines on the downstream turbines are regarded as edges of the directed graph;
[0102] S203: Based on the directed graph established above, a community discovery algorithm is used to describe the community structure of the wind farm.
[0103] Specifically, a hierarchical clustering algorithm is used to divide the dominant wind direction. First, all wind direction data are preprocessed to limit the wind direction change to between positive and negative 180 degrees. After standardization, all individual data points are treated as one class, and the Euclidean distance and Fréchet distance between any two points are calculated to construct the similarity measurement matrix F(A,B):
[0104] F(A,B)=infmax{d(A(i),B(j))}
[0105] In the formula: A(i) and B(j) are two different classes, i and j are the serial numbers of a point in the class, that is, any two nodes, inf is used to find the lower bound, d is the Euclidean distance between two points, and the physical meaning of the Fleche distance is: the lower bound of the maximum value of the Euclidean distance between any two points in the two classes A(i) and B(j). The two classes with the closest distance are merged, and the similarity is updated using the Fleche distance. This step is repeated until all data are clustered into one class to obtain a clustering graph.
[0106] A Louvain algorithm based on multi-level optimization modularity is used to cluster the wind turbines in the wind farm in combination with the specific layout of the wind farm and the wake effect. Individual wind turbines can be regarded as nodes of a directed graph, and the wake impact of upstream units on downstream units can be regarded as edges of the directed graph. The set of directed edges in the wind farm can be obtained based on the wake effect.
[0107] The modularity Q in the community structure of a directed graph is defined as the expected difference between the proportion of edges connecting internal nodes in the original directed network and the proportion of edges connecting internal nodes in the directed random network. The greater the modularity, the higher the quality of the clustering. The expression of Q is as follows:
[0108]
[0109] In the formula, i and j represent any two nodes in the directed graph, m represents the number of edges in the network, and k i is the degree of node i, k j is the degree of node j, A ij is an element in the adjacency matrix, representing whether nodes i and j are connected; δ(i,j) is a judgment function, if i and j are in the same community, its value is 1, otherwise it is 0.
[0110] First, all nodes are divided into separate communities. Second, a vertex is merged with an adjacent vertex. If the modularity is greater than zero, the vertex is merged into the community where the adjacent vertex is located. This step is repeated until the communities where all vertices are located no longer change. Then, all communities are compressed into one vertex, and the weights of all vertices in the original community are converted into the weights of the new vertex. The weights of all edges between the original communities are converted into the weights of the new edges. The above three steps are repeated until the community no longer changes, and different wind turbine groups in the entire wind farm are obtained.
[0111] In an optional implementation manner, generating a plurality of wind power scenarios of the wind farm according to the wind speed data of the wind farm includes:
[0112] A randomly generated noise signal is used as an input random vector, and the measured wind speed data of the wind farm is used as sample data. A generative adversarial neural network is used to generate scenarios to obtain several wind power scenarios.
[0113] Specifically, the wind farm output generation scenario using a generative adversarial neural network includes a generator network and a discriminator network. The generator network inputs a random vector to generate simulated data similar to the sample data, and passes the simulated data to the discriminator network. The discriminator network inputs the sample data and the simulated data output by the generator network, and outputs the probability that the simulated data belongs to the sample data. The random vector is a randomly generated noise signal, and the sample data is the measured data of the wind farm, including time series wind speed information in the three directions of u, v, and w at different positions on the wind speed plane. The generated data consistent with the data sample is achieved by continuously updating the parameters of the generator and the discriminator.
[0114] The loss function Loss(G,D) of the generative adversarial neural network is as follows:
[0115]
[0116] Among them, n real is the number of real data, n fake is the number of generated data, x i is the real data, D(x i ) is the output of the discriminant network based on sample data, p is noise data, G(p) represents the simulated data generated by the generator, and D(G(p)) represents the output of the discriminator for the simulated data. The first term represents the loss of samples from real data on the discriminator, and the second term represents the loss of samples from the generator. Fix the parameters of the generator network and update the parameters of the discriminant network, that is, calculate the JS divergence between the sample data and the simulated data; fix the parameters of the discriminant network and update the parameters of the generator network, that is, minimize the JS divergence, so that the two data distribution patterns are closer. Update the parameters of the two networks alternately until the training is completed.
[0117] In an optional implementation manner, the reducing the generated wind power scenarios to obtain a typical scenario of the wind farm includes:
[0118] An unsupervised clustering method is used to reduce the generated wind power scenarios to obtain a typical scenario of the wind farm.
[0119] Specifically, the unsupervised K-means clustering method is used to reduce the wind power scenarios generated above. Sample points are randomly selected from the sample set as the initial cluster center, and the Euclidean distances from other sample points to the cluster center are calculated. i ) Select the remaining k 3 -1 cluster center, probability P(x i) is the ratio of the Euclidean distance of the i-th sample point from the initial cluster center to the sum of the Euclidean distances of all remaining sample points from the initial cluster center. Set different numbers of cluster centers, perform clustering under different values, and select the value with the largest silhouette coefficient as the number of cluster centers k 3 . Calculate the mean of all data clusters and set it as the new cluster center, and repeat this step until the cluster center does not change. Select representative samples in the cluster as typical scenes to achieve the reduction of complex scenes.
[0120] In an alternative embodiment, see Figure 3 As shown, Figure 3 A flowchart of a method for generating a pre-dispatching instruction provided in the first embodiment of the present invention is shown, wherein for each typical scenario, based on a reinforcement learning algorithm, a pre-dispatching instruction is determined that enables the output power and unit load of the wind farm digital twin model in the typical scenario to meet the objective function, including steps S301 to S302:
[0121] S301: Establish a reward mechanism with output power and unit load as objective functions, and loads of key components of the wind turbine and the limit angle of yaw control as constraints.
[0122] S302: According to the reward mechanism, pre-dispatching instructions for each wind turbine are obtained through self-learning to implement pre-dispatching of yaw control of the wind farm.
[0123] Specifically, based on the reinforcement learning RL algorithm, for different typical scenarios, the output power and unit load of the digital twin model of the wind farm are taken as objective functions, and the loads of key components of the wind turbine and the limit angle of yaw control are taken as constraints. An incentive mechanism is established, and the yaw control angles of each unit in the field are obtained through self-learning when the output power of the entire field is maximum and the load of the wind turbine is minimum, so as to realize the pre-scheduling of the yaw control of the wind farm; the loads of key components include: multi-modal load limits of wind turbine components (such as blade root out-of-plane bending moment, tower bottom bending moment, intermediate shaft torque, etc.); a reinforcement learning table is obtained, and the initial yaw angle is set according to the table in different scenarios of the wind farm.
[0124] In an optional embodiment, the constraints include the ultimate load limit and yaw angle limit of the wind turbine set; wherein the ultimate load limit is obtained by simulating the extreme value distribution theory, and the yaw angle limit condition is that the set load does not exceed the ultimate load within the limited angle range.
[0125] Specifically, the constraints include the ultimate load limit and yaw angle limit of the wind turbine. The ultimate load is simulated using the extreme value distribution theory. First, the ultimate load distribution is fitted according to the operating load data of the wind turbine (Type I, Type II and Type III distribution models can be used), and the selected distribution model is evaluated and verified using the normality test method. Finally, combined with the design service life of the wind turbine, the probability of the ultimate load state of the computer group under different conditions is calculated to obtain the ultimate load value. The yaw angle restriction condition is: within the restricted angle range, the unit load does not exceed the ultimate load.
[0126] In an optional implementation, the pre-scheduling instruction is used as the initial value of the real-time optimization scheduling, and the control parameters are optimized using a distributed predictive control algorithm, including:
[0127] When optimizing the control parameters, the constraint conditions are used as constraints in the optimization process, and the objective function is used as the optimization target. The overall optimization calculation of the wind farm is divided into parallel optimization calculations of multiple wind turbine subsystems, and the yaw scheduling instructions of each wind turbine are obtained in real time to achieve optimization of the control parameters.
[0128] Specifically, the yaw control parameters indicated by the pre-dispatch instruction are used as the initial values of the real-time optimization dispatch, and the DMPC distributed predictive control algorithm is used to perform real-time optimization of the wind farm digital twin. Each wind turbine is regarded as a subsystem, and each wind turbine subsystem controller uses the wind farm model to optimize its corresponding input. Each subcontroller considers the information of the remaining subsystems in the calculation. During the optimization process, all wind turbine subcontrollers use the global objective function of the wind farm, and use the aforementioned constraints as constraints in the optimization process. The overall optimization calculation of a wind farm is divided into multiple subsystems for parallel optimization calculations, and the yaw dispatch instructions of each unit in the wind farm are obtained in real time to optimize the control parameters.
[0129] In an alternative embodiment, the method comprises:
[0130] The input wind speed is soft-measured through the unit status, and the laser radar wind measurement data is combined for verification to construct a digital twin model of the inflow wind speed of the wind farm; the propagation law of the fluid in front of the computer is calculated, and the digital twin model of the wind speed in front of the machine is constructed according to the input wind speed of the wind farm; the finite element modeling method is adopted, combined with the blade theory and momentum theory to obtain the aerodynamic characteristics of the wind turbine, and the structural response of the wind turbine is calculated using the multi-body dynamics method to construct a single-machine digital twin model; the digital twin model of the wake characteristics is constructed based on the dynamic winding principle; the digital twin model of the wind farm inflow wind speed, the digital twin model of the wind speed in front of the machine, the single-machine digital twin model, the digital twin model of the wake characteristics and the wind farm field-level energy management twin model are combined to construct the digital twin model of the wind farm.
[0131] Specifically, the twin modeling of the wind speed in front of the wind farm performs soft measurement of the input wind speed through the unit status, and combines it with the lidar wind measurement data for verification to achieve accurate measurement of the input wind speed of the wind farm, and the propagation law of the fluid in front of the computer is calculated to reconstruct the inflow wind speed in front of the wind turbine according to the input wind speed of the wind farm; the dynamic twin modeling of a single unit adopts the finite element modeling method, combines the blade theory and momentum theory to obtain the aerodynamic characteristics of the wind turbine, and uses the multi-body dynamics method to calculate the structural response of the wind turbine; based on the dynamic winding principle, a digital twin model of the wake characteristics of the wind farm is established; the above model is combined with the field-level energy management twin model of the wind farm to obtain a digital twin model of the wind farm.
[0132] The wind farm twin model mentioned above is compiled for the working conditions of parallel operation in the Linux RT environment, and localized real-time simulation is performed on a high-performance server with a multi-core CPU.
[0133] In an optional embodiment, the method further comprises:
[0134] The data with the highest correlation with the radar wind measurement data in the unit state parameters are selected as input data, and the radar wind measurement data is used as output data. The digital twin model of the inflow wind speed of the wind farm is constructed using the deep learning neural network method.
[0135] Specifically, historical operating data are collected as sample data, the lidar data therein is calibrated, and the correlation between the unit state parameters (such as speed, torque, pitch angle, power generation power, blade root load, drive shaft load, etc.) in the historical operating data and the radar wind measurement data is analyzed. The data with the highest correlation is selected as input data, and the radar wind measurement data is used as output data. The deep learning neural network method is used to estimate the wind speed at the lidar measurement position (the common value is 200 meters) as the output.
[0136] A neutral atmospheric boundary layer without wind turbines is generated by large eddy simulation method. The wind turbine is placed in the neutral atmospheric boundary layer, and the interaction between the wind turbine and the atmospheric boundary layer is studied by adopting LES coupled actuation line model method, thereby constructing the digital twin model of wind speed in front of the wind turbine.
[0137] Specifically, the large eddy simulation method is used to generate a neutral atmospheric boundary layer without wind turbines, and the wind turbine is placed in the neutral atmospheric boundary layer. The LES coupled actuation line model method is used to study the interaction between the wind turbine and the atmospheric boundary layer. The turbulence is divided into large-scale vortices and small-scale vortices by the filtering method. The large-scale vortices are directly solved by the NS equations, and the small-scale vortices are simulated by the sub-grid model. The surface stress and temperature flux of the atmospheric boundary layer are simulated according to the surface roughness height. The inflow wind speed in front of the wind turbine is reconstructed by the above method, and the evolution process of atmospheric turbulence in front of the wind turbine is analyzed.
[0138] The blade element momentum theory is used to analyze the aerodynamic performance of the impeller, and the single-machine digital twin model is constructed based on the aerodynamic performance analysis results.
[0139] Specifically, the blade element momentum theory is used to analyze the aerodynamic performance of the impeller to achieve refined modeling of the impeller's aerodynamic characteristics. The blade element theory divides the entire blade into several aerodynamic characteristic regions, namely blade element units, along the radial direction according to the airfoil characteristic parameters such as blade chord length and thickness; the momentum theory is based on the law of conservation of momentum to obtain the relationship between the force, torque, momentum, and angular momentum of the unit's impeller. Combining the blade element theory and momentum theory, the overall aerodynamic characteristics of the impeller can be obtained by superimposing the aerodynamic characteristics of several blade elements.
[0140] A passive tracer model of the wake is established based on the dynamic meandering wake model. Based on the evolution of the wake loss, the wake meandering, the increase in turbulence intensity caused by the wake, the drop in wind speed in the near-wake area and the radial expansion of the wake, the passive tracer model is response-corrected to obtain the digital twin model of the wake characteristics.
[0141] Specifically, a passive tracer model of the wake was established based on the DWM dynamic meandering wake model, taking into account the evolution of the wake loss, the wake meandering, and the increase in turbulence intensity caused by the wake. On the basis of DWM, the wind speed drop in the near-wake area (pressure gradient area) and the radial expansion of the wake were considered, and a response correction was performed to improve the accuracy of the far wake. The overlapping areas of all wakes were calculated, and the wake losses in the overlapping areas were merged.
[0142] In the DWM model, small turbulent eddies with a vortex scale less than twice the diameter of the wind rotor are used to calculate the evolution of the wake loss, and large turbulent eddies with a vortex scale greater than twice the diameter of the wind rotor are used to calculate the wake meandering (increasing the turbulence intensity and increasing the wind turbine load).
[0143] Wake losses are superimposed in the axial direction based on the root-squared-sum (RSS) method. The transverse components (radial wake losses) are superimposed by a vector sum. The RSS method assumes that the local kinetic energy of the axial losses in the combined wake is equal to the sum of the local energies of the axial losses of each wake at a given wind data point. The RSS method is applicable only to scalar arrays. This method is applicable to axial losses because overlapping wakes are likely to have similar axial directions. Since any given radial direction depends on the azimuth in an axisymmetric coordinate system, a vector sum is applied to the transverse components (radial wake losses).
[0144] The wind farm-level energy management twin model is used to dynamically adjust the power output of each wind turbine according to the real-time conditions of the wind farm and the grid demand, so as to optimize the energy distribution and operation efficiency of the entire wind farm.
[0145] Specifically, based on the active load command given by the power grid and combined with the power output in the wind farm under the current yaw control, the multi-objective grey wolf algorithm is used to optimize the power output of each unit. The control target is the allowable power generation. The digital twin model and real-time data of the wind turbine are used to comprehensively consider the fatigue load of the wind turbine under different power outputs. The unit output is dynamically adjusted through variable pitch control to increase or reduce the power of a single unit, so that the total output power of the wind farm meets the requirements of the power grid command.
[0146] The field-level energy management digital twin model comprehensively manages and configures the energy of the wind farm, executes the active load command given by the power grid, and achieves the purpose of automatically controlling the online power of the wind farm. Taking the maximum allowable power generation as the control target, the wind farm online load is dispatched and controlled through allocation calculation, so that the wind farm online load is automatically controlled within the rated capacity. The field-level energy management twin model of the wind farm can dynamically adjust the power output of each wind turbine according to the real-time conditions of the wind farm and the needs of the power grid to optimize the energy distribution and operation efficiency of the entire wind farm. According to the active load command given by the power grid, combined with the power output and the power output in the wind farm under the current yaw control, the multi-objective gray wolf algorithm is used to optimize the power output of each unit. Taking the allowable power generation as the control target, the digital twin model and real-time data of the wind turbine are used to comprehensively consider the fatigue load of the wind turbine under different power outputs, and the unit output is dynamically adjusted through the pitch control method to increase or reduce the power of a single unit, so that the total output power of the wind farm meets the grid instruction requirements.
[0147] In practical applications, the maximum information coefficient MIC method can be used to analyze the correlation between the unit state parameters (such as speed, torque, pitch angle, power generation, blade root load, drive shaft load, etc.) and radar wind measurement data in the historical operation data of the wind farm. Taking the MIC correlation of speed-radar wind measurement data as an example, the data points are distributed in a two-dimensional plane, and the two-dimensional plane is refined according to the grid of i rows and j columns to calculate the maximum mutual correlation information MIC[x,y]:
[0148]
[0149] Among them, I(x,y) is the joint probability between variables x and y, a and b are the number of grids divided in the horizontal direction and vertical direction respectively, B is a preset variable, which is usually set to the 0.6th power of the data volume; p(x,y) is the joint probability distribution function of variables x and y, p(x) is the marginal probability density function of variable x, and p(y) is the marginal probability density function of variable y. Normalize the maximum mutual information, and calculate the maximum mutual information value of the two types of data when i and j take different values as the MIC value. Replace the original data of the unit state parameter, and select the unit state parameter with the largest MIC value, which is the parameter with the strongest correlation with the radar wind measurement data.
[0150] The data with the highest correlation are selected as input data, and the radar wind measurement data are used as output data. The LSTM time series dynamic deep learning neural network is used to estimate the front x 3 The wind speed at a distance of 200 meters (the measurement distance of the laser radar, a common value is 200 meters). LSTM contains a network layer composed of a series of gated units, each of which can selectively control the flow of information and learn whether this information needs to be remembered or forgotten. The cells in the LSTM network are the core units of LSTM. It is a special neural network unit used to process long-term dependencies in sequence data. Each LSTM cell consists of a cell state and three gates, namely the input gate, the forget gate, and the output gate. These three gates determine the degree of transmission of input information, historical information, and output information at the current moment through certain operations and activation functions, thereby realizing the memory of historical information and the processing of current information.
[0151] The maximum number of training rounds, learning rate strategy, initial learning rate, learning rate reduction factor, learning rate reduction cycle, gradient threshold and other parameters are set to train the neural network load model. During the model operation, the input wind speed at 200 meters in front of the unit is predicted based on the unit's operating status information as input.
[0152] The neutral atmospheric boundary layer without wind turbines is generated by large eddy simulation method, and the wind turbine is placed in the neutral atmospheric boundary layer. The interaction between wind turbine and atmospheric boundary layer is studied by LES coupling actuation line model method. The turbulence is divided into large-scale vortices and small-scale vortices by filtering method. The large-scale vortices are directly solved by NS equations, and the small-scale vortices are simulated by sub-grid model. The surface stress and temperature flux of the atmospheric boundary layer are simulated according to the surface roughness height.
[0153] The turbulent momentum equation is:
[0154]
[0155] In the formula, t is time, x is i1 ,y j1 are the spatial coordinates in the i1 and j1 directions respectively, u i1 、u j1 They are the velocity components in the i1 and j1 directions respectively, i1 represents the horizontal direction, j1 represents the vertical direction, p is the dynamic pressure, z is the surface roughness, and f e is the fan body force, Ω j is the rotation rate tensor, ρ 0 is the air density, p 0 is the initial value of dynamic pressure; ε jk If the permutation is odd, the value is -1; if the permutation is even, the value is +1; otherwise, it is 0;k is the velocity component in the k direction, τ ij is the sub-grid stress, g is the gravitational acceleration, p b is the buoyancy density of air.
[0156] Based on the blade element momentum theory (BEM), the entire blade is divided into several aerodynamic characteristic regions according to the airfoil characteristic parameters such as blade chord length, thickness, pitch angle, etc., so as to perform finite element analysis and modeling. A microelement with a length of dr at the length r of the blade along the axial direction is called a blade element, whose chord length is l and pitch angle is β; the blade is divided into different microsegments along the radius of the wind wheel, and the aerodynamic lift and drag generated by each microsegment under the action of wind speed are determined by its airfoil. Assuming that the air flow between each microelement does not interfere with each other, and the forces and moments acting on all microelements can be synthesized without loss into the resultant force and total moment borne by the blade, the overall moment borne by the blade can be obtained by superimposing the forces of all microelements of the blade. Assuming that the blade element is acted upon by an incoming wind with a relative velocity of v, an aerodynamic force dF perpendicular to the chord is generated. The aerodynamic force is decomposed along the direction of the incoming wind and the direction perpendicular to it, and the lift dL and drag dD of the blade element under the incoming wind are obtained:
[0157]
[0158] Among them, ρ 0 is the air density, C l , C d are the lift coefficient and drag coefficient of the blade element, respectively, obtained by analyzing the aerodynamic performance through its airfoil information, r is the axial length of the blade, l is the chord length, and v is the relative speed of the incoming wind. The aerodynamic force dF is decomposed along the direction parallel and perpendicular to the impeller rotation plane to obtain the local impeller thrust dF a and torque dF u , expressed in terms of lift dL and drag dD:
[0159]
[0160] in, is the angle between the incoming wind and the impeller rotation plane. The force of each micro segment is calculated by aerodynamic methods, and the aerodynamic torque of the wind rotor blade as a whole can be obtained by superimposing the forces of each micro segment, and its expression is:
[0161]
[0162] In the formula, F r is the total impeller thrust, Q r is the total torque of the impeller, R is the impeller radius, R hub is the wheel hub radius, V totalis the wind speed at the blade. The structural dynamic response of the fan with N degrees of freedom is described by the assumed modal method, and its dynamic equation is established using the Kane dynamic equation:
[0163] F(r)+F R (r) = 0
[0164] Among them, F(r) and F R (r) is the generalized main force and generalized inertia force of the wind turbine system, the inertia force of the tower and blades F 1 According to the following formula:
[0165]
[0166] Where ρ(l) and a(l) are the line density and linear acceleration vector along the tower centerline and the blade axis, respectively, and L is the total length of the tower or blade.
[0167] Inertia force F of nacelle and wheel hub 2 According to the following formula:
[0168] F 2 =v 2 (-m 2 a 2 )+w 2 (-H)
[0169] Among them, v 2 、w 2 and a 2 are the linear velocity, angular velocity and linear acceleration of the nacelle and hub in the inertial system, m 2 is the mass of the nacelle and the hub, and H is the differential of the angular momentum of the nacelle and the hub about the center of mass in the inertial system.
[0170] The generalized main force F(r) on the fan system is composed of the aerodynamic force F acting on the fan. aero (r), gravity F G (r), elastic force F elastic (r) etc.
[0171] F(r)=F aero (r)+F G (r)+F elastic (r)
[0172] Among them, F aero (r) is calculated by the following formula:
[0173] F aero (r) = v 2 F 3 +w 2 M
[0174] Among them, F 3 and M are the aerodynamic force and pitching moment acting on the blade in the aerodynamic module, respectively.
[0175] The blades and tower are simplified into flexible cantilever beams with two degrees of freedom. The blades are connected to the rigid hub, and the tower is connected to the ground. A coordinate system is established based on orthogonal unit vectors to define the rigid structural reference system. The coordinate systems of each component can be transformed by the coordinate transfer matrix. According to the distributed characteristic parameters of the tower (such as diameter, density, and stiffness), the center of the tower is measured from bottom to top using n 1 The deformation U(z 3 ,t) can be written as:
[0176]
[0177] Among them, z 3 is the position along the cantilever direction, t is the time, φ i (z 3 ) is the vibration mode of the cantilever, is the generalized coordinate of the tower top or blade root; m ij , k ij represent generalized mass and generalized stiffness respectively; c j (t) is about the mode shape φ i (z 3 ), c j The first derivative of (t), c j The second-order derivative of (t), the force and displacement at any point on the wind turbine can be expressed by the above formula.
[0178] In the DWM model, small turbulent eddies with eddy scales less than twice the diameter of the wind rotor are used to calculate the evolution of the wake loss, and large turbulent eddies with eddy scales greater than twice the diameter of the wind rotor are used to calculate the wake meandering (increasing the turbulence intensity and wind turbine load). The wind rotor wind speed is low-pass filtered, and the evolution of the wake loss under quasi-steady-state conditions is simulated by the thin shear layer approximation. The thin shear layer approximation of the Reynolds-averaged NS equations in axisymmetric coordinates is used to simulate the evolution of the wake loss, and the eddy viscosity formula is used to capture the turbulent closure. The thin shear layer approximation discards the pressure term and assumes that the radial velocity gradient is much larger than the radial velocity gradient. The analytical expressions for conservation of momentum and conservation of mass are as follows:
[0179]
[0180] In the formula, v T is the eddy viscosity, V x With V r are the axial and radial velocities in the axisymmetric coordinate system, r0 is the diameter of the wind wheel, and r is the axial length of the blade:
[0181]
[0182] In the formula, and is the axial and radial wake loss.
[0183]
[0184] In the formula, C nearwake It is a custom parameter that determines the extent of radial expansion of the wake in the near-wake region; is the wind speed perpendicular to the wind rotor after low-pass filtering; C t (r) is the average thrust coefficient, which is derived from the blade element momentum theory, V x (x,r) is the axial velocity, is the axial wake loss, V r (x,r) is the radial velocity, is the radial wake loss.
[0185] Wake losses are superimposed in the axial direction based on the root-squared-sum (RSS) method. The transverse components (radial wake losses) are superimposed by a vector sum. The RSS method assumes that the local kinetic energy of the axial losses in the combined wake is equal to the sum of the local energies of the axial losses of each wake at a given wind data point. The RSS method is applicable only to scalar arrays. This method is applicable to axial losses because overlapping wakes are likely to have similar axial directions. Since any given radial direction depends on the azimuth in an axisymmetric coordinate system, a vector sum is applied to the transverse components (radial wake losses).
[0186] This application constructs a digital twin model of a wind farm based on a digital twin model of the wind speed inflow of the wind farm, a digital twin model of the wind speed in front of the machine, a digital twin model of a single machine, a digital twin model of the wake characteristics, and a twin model of the wind farm-level energy management. Then, an unsupervised deep learning neural network is used to generate typical scenarios. A two-step optimization scheduling method of pre-scheduling and real-time optimization scheduling is adopted to determine the optimal yaw control strategy for different typical scenarios, and the yaw angle of the wind turbine in each scenario is controlled based on the optimal yaw control strategy to reduce the unit load and power loss. The established digital twin model of the wind farm avoids the defect of low accuracy of engineering wake modeling, reduces the optimization complexity by dividing the dominant wind direction and generating typical scenarios, and realizes effective optimization of the wind farm output power and unit load through active yaw control, thereby improving energy utilization and helping to reduce the operating cost of the wind farm.
[0187] Embodiment 2
[0188] See also Figure 4As shown, Figure 4 A schematic diagram of the structure of a wind farm multi-objective collaborative optimization scheduling device based on digital twins provided in the second embodiment of the present invention is shown, wherein the device comprises:
[0189] The wind turbine group grouping module 401 is used to divide the dominant wind direction of the wind farm according to the wind direction data of the wind farm; and to group the wind turbines in the wind farm according to the wake characteristics under the dominant wind direction to obtain a plurality of wind turbine group groups;
[0190] The typical scene division module 402 is used to generate a number of wind power scenes of the wind farm according to the wind speed data of the wind farm; and perform scene reduction on the generated several wind power scenes to obtain a typical scene of the wind farm;
[0191] The pre-dispatching instruction determination module 403 is used to determine, for each typical scenario, based on the reinforcement learning algorithm, a pre-dispatching instruction that can make the output power and unit load of the wind farm digital twin model in the typical scenario meet the objective function;
[0192] A control parameter optimization module 404 is used to use the pre-scheduling instruction as the initial value of the real-time optimization scheduling and optimize the control parameters using a distributed predictive control algorithm;
[0193] The wind turbine control module 405 is used to control the operation of each wind turbine based on the optimized control parameters in this typical scenario.
[0194] In an optional implementation, the dominant wind direction of the wind farm is divided according to the wind direction data of the wind farm, and the wind turbines in the wind farm are grouped according to the wake characteristics under the dominant wind direction to obtain a plurality of wind turbine groups, including:
[0195] Using a hierarchical clustering method to divide the dominant wind direction from the wind direction data;
[0196] Under the dominant wind direction, combined with the specific arrangement of the wind farm, the wind turbines are regarded as nodes of a directed graph, and the wake effects of the upstream turbines on the downstream turbines are regarded as edges of the directed graph;
[0197] Based on the directed graph established above, a community discovery algorithm is used to describe the community structure of the wind farm.
[0198] In an optional implementation manner, generating a plurality of wind power scenarios of the wind farm according to the wind speed data of the wind farm includes:
[0199] A randomly generated noise signal is used as an input random vector, and the measured wind speed data of the wind farm is used as sample data. A generative adversarial neural network is used to generate scenarios to obtain several wind power scenarios.
[0200] In an optional implementation manner, the reducing the generated wind power scenarios to obtain a typical scenario of the wind farm includes:
[0201] Using an unsupervised clustering method to reduce the generated wind power scenarios to obtain a typical scenario of the wind farm;
[0202] In an optional implementation scheme, for each typical scenario, determining the pre-dispatching instructions based on the reinforcement learning algorithm so that the output power and unit load of the wind farm digital twin model in the typical scenario meet the objective function includes:
[0203] With output power and unit load as the objective function, and the load of key components of the wind turbine and the limit angle of yaw control as the constraint conditions, an incentive mechanism is established;
[0204] According to the reward mechanism, pre-dispatching instructions for each wind turbine are obtained through self-learning, thereby realizing pre-dispatching of yaw control of the wind farm.
[0205] In an optional embodiment, the constraint conditions include the ultimate load limit and yaw angle limit of the wind turbine generator set;
[0206] The extreme value distribution theory is used to simulate the ultimate load limit, and the yaw angle limit condition is that the unit load does not exceed the ultimate load within the limited angle range.
[0207] In an optional implementation, the pre-scheduling instruction is used as the initial value of the real-time optimization scheduling, and the control parameters are optimized using a distributed predictive control algorithm, including:
[0208] When optimizing the control parameters, the constraint conditions are used as constraints in the optimization process, and the objective function is used as the optimization target. The overall optimization calculation of the wind farm is divided into parallel optimization calculations of multiple wind turbine subsystems, and the yaw scheduling instructions of each wind turbine are obtained in real time to achieve optimization of the control parameters.
[0209] In an optional embodiment, the device comprises a model building module, wherein the model building module is used to:
[0210] The input wind speed is soft-measured through the unit status, and verified with the lidar wind measurement data to build a digital twin model of the wind farm inflow wind speed;
[0211] The propagation law of the fluid in front of the computer is calculated, and a digital twin model of the wind speed in front of the machine is constructed according to the input wind speed of the wind farm;
[0212] The finite element modeling method is used to combine blade element theory and momentum theory to obtain the aerodynamic characteristics of the wind turbine. The multi-body dynamics method is used to calculate the structural response of the wind turbine to build a single-machine digital twin model.
[0213] Construct a digital twin model of wake characteristics based on the dynamic meandering principle;
[0214] The digital twin model of the wind farm is constructed by combining the wind farm inflow wind speed digital twin model, the front-machine wind speed digital twin model, the single-machine digital twin model, the wake characteristic digital twin model and the wind farm-level energy management twin model.
[0215] In an optional embodiment, the model building module is specifically used for:
[0216] The data with the highest correlation with the radar wind measurement data in the unit state parameters are selected as input data, the radar wind measurement data are used as output data, and a deep learning neural network method is used to construct the digital twin model of the inflow wind speed of the wind farm;
[0217] A neutral atmospheric boundary layer without a wind turbine is generated by using a large eddy simulation method, a wind turbine is placed in the neutral atmospheric boundary layer, and the interaction between the wind turbine and the atmospheric boundary layer is studied by using a LES coupled actuation line model method, thereby constructing the wind speed digital twin model in front of the wind turbine;
[0218] The blade element momentum theory is used to analyze the aerodynamic performance of the impeller, and the single-machine digital twin model is constructed based on the aerodynamic performance analysis results;
[0219] A passive tracer model of the wake is established based on a dynamic meandering wake model. Based on the evolution of the wake loss, the meandering of the wake, the increase in turbulence intensity caused by the wake, the drop in wind speed near the wake area and the radial expansion of the wake, the passive tracer model is calibrated to obtain the digital twin model of the wake characteristics.
[0220] The wind farm-level energy management twin model is used to dynamically adjust the power output of each wind turbine according to the real-time conditions of the wind farm and the grid demand, so as to optimize the energy distribution and operation efficiency of the entire wind farm.
[0221] Embodiment 3
[0222] Based on the same application concept, see Figure 5 As shown, Figure 5 FIG. 4 shows a schematic diagram of the structure of a computer device provided by Embodiment 3 of the present invention, wherein: Figure 5 As shown, a computer device 500 provided in Embodiment 3 of the present application includes:
[0223] A processor 501, a memory 502 and a bus 503, wherein the memory 502 stores machine-readable instructions executable by the processor 501. When the computer device 500 is running, the processor 501 communicates with the memory 502 through the bus 503. When the processor 501 is running, the machine-readable instructions execute the steps of the multi-objective collaborative optimization scheduling method for wind farms based on digital twins shown in the above-mentioned embodiment 1.
[0224] Embodiment 4
[0225] Based on the same application concept, an embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the multi-objective collaborative optimization scheduling method for wind farms based on digital twins described in any of the above embodiments are executed.
[0226] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.
[0227] The computer program product for performing multi-objective collaborative optimization scheduling of wind farms based on digital twins provided in an embodiment of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the previous method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.
[0228] The wind farm multi-objective collaborative optimization scheduling device based on digital twins provided in the embodiment of the present invention can be specific hardware on the device or software or firmware installed on the device. The device provided in the embodiment of the present invention has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference can be made to the corresponding contents in the aforementioned method embodiment. Technical personnel in the relevant field can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices and units described above can all refer to the corresponding processes in the aforementioned method embodiment, and will not be repeated here.
[0229] In the embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.
[0230] 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, that is, they 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.
[0231] In addition, each functional unit in the embodiment provided by the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0232] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.
[0233] It should be noted that similar numbers and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and are not to be understood as indicating or implying relative importance.
[0234] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can still modify the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or perform equivalent replacements on some of the technical features thereof; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. They should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims
1. A wind farm multi-objective collaborative optimization scheduling method based on digital twins, characterized in that: The method comprises: The dominant wind direction of the wind farm is divided according to the wind direction data of the wind farm, and the wind turbines in the wind farm are grouped according to the wake characteristics under the dominant wind direction to obtain a plurality of wind turbine groups; Generating a plurality of wind power scenarios of the wind farm according to the wind speed data of the wind farm; performing scenario reduction on the generated plurality of wind power scenarios to obtain a typical scenario of the wind farm; For each typical scenario, the pre-dispatch instructions are determined based on the reinforcement learning algorithm so that the output power and unit load of the wind farm digital twin model in the typical scenario meet the objective function; The pre-scheduling instruction is used as the initial value of the real-time optimization scheduling, and the control parameters are optimized using a distributed predictive control algorithm; In this typical scenario, the operation of each wind turbine is controlled based on the optimized control parameters.
2. The method according to claim 1, characterized in that The method of dividing the dominant wind direction of the wind farm according to the wind direction data of the wind farm, and grouping the wind turbines in the wind farm according to the wake characteristics under the dominant wind direction to obtain a plurality of wind turbine groups includes: Using a hierarchical clustering method to divide the dominant wind direction from the wind direction data; Under the dominant wind direction, combined with the specific arrangement of the wind farm, the wind turbines are regarded as nodes of a directed graph, and the wake effects of the upstream turbines on the downstream turbines are regarded as edges of the directed graph; Based on the directed graph established above, a community discovery algorithm is used to describe the community structure of the wind farm.
3. The method according to claim 1, characterized in that Generating a plurality of wind power scenarios of the wind farm according to the wind speed data of the wind farm comprises: A randomly generated noise signal is used as an input random vector, and the measured wind speed data of the wind farm is used as sample data. A generative adversarial neural network is used to generate scenarios to obtain several wind power scenarios.
4. The method according to claim 1, characterized in that: The step of reducing the generated wind power scenarios to obtain a typical scenario of the wind farm includes: An unsupervised clustering method is used to reduce the generated wind power scenarios to obtain a typical scenario of the wind farm.
5. The method according to claim 1, characterized in that For each typical scenario, the pre-dispatching instructions that can make the output power and unit load of the wind farm digital twin model in the typical scenario meet the objective function are determined based on the reinforcement learning algorithm, including: With output power and unit load as the objective function, and the load of key components of the wind turbine and the limit angle of yaw control as the constraint conditions, an incentive mechanism is established; According to the reward mechanism, pre-dispatching instructions for each wind turbine are obtained through self-learning, thereby realizing pre-dispatching of yaw control of the wind farm.
6. The method according to claim 5, characterized in that The constraint conditions include the ultimate load limit and yaw angle limit of the wind turbine; The extreme value distribution theory is used to simulate the limit load limit, and the yaw angle limit condition is that the unit load does not exceed the limit load within the limit angle range.
7. The method according to claim 5, characterized in that The method of using the pre-scheduling instruction as the initial value of the real-time optimization scheduling and optimizing the control parameters by using a distributed predictive control algorithm includes: When optimizing the control parameters, the constraint conditions are used as constraints in the optimization process, and the objective function is used as the optimization target. The overall optimization calculation of the wind farm is divided into parallel optimization calculations of multiple wind turbine subsystems, and the yaw scheduling instructions of each wind turbine are obtained in real time to achieve optimization of the control parameters.
8. The method according to claim 1, characterized in that The method comprises: The input wind speed is soft-measured through the unit status, and verified with the lidar wind measurement data to build a digital twin model of the wind farm inflow wind speed; The propagation law of the fluid in front of the computer is calculated, and a digital twin model of the wind speed in front of the machine is constructed according to the input wind speed of the wind farm; The finite element modeling method is used to combine blade element theory and momentum theory to obtain the aerodynamic characteristics of the wind turbine. The multi-body dynamics method is used to calculate the structural response of the wind turbine to build a single-machine digital twin model. Construct a digital twin model of wake characteristics based on the dynamic meandering principle; The digital twin model of the wind farm is constructed by combining the wind farm inflow wind speed digital twin model, the front-machine wind speed digital twin model, the single-machine digital twin model, the wake characteristic digital twin model and the wind farm-level energy management twin model.
9. The method according to claim 8, characterized in that The method further comprises: The data with the highest correlation with the radar wind measurement data in the unit state parameters are selected as input data, the radar wind measurement data are used as output data, and a deep learning neural network method is used to construct the digital twin model of the inflow wind speed of the wind farm; A neutral atmospheric boundary layer without a wind turbine is generated by using a large eddy simulation method, a wind turbine is placed in the neutral atmospheric boundary layer, and the interaction between the wind turbine and the atmospheric boundary layer is studied by using a LES coupled actuation line model method, thereby constructing the wind speed digital twin model in front of the wind turbine; The blade element momentum theory is used to analyze the aerodynamic performance of the impeller, and the single-machine digital twin model is constructed based on the aerodynamic performance analysis results; A passive tracer model of the wake is established based on a dynamic meandering wake model. Based on the evolution of the wake loss, the meandering of the wake, the increase in turbulence intensity caused by the wake, the drop in wind speed near the wake area and the radial expansion of the wake, the passive tracer model is calibrated to obtain the digital twin model of the wake characteristics. The wind farm-level energy management twin model is used to dynamically adjust the power output of each wind turbine according to the real-time conditions of the wind farm and the grid demand, so as to optimize the energy distribution and operation efficiency of the entire wind farm.
10. A wind farm multi-objective collaborative optimization scheduling device based on digital twin, characterized in that: The device comprises: A wind turbine group grouping module is used to divide the dominant wind direction of the wind farm according to the wind direction data of the wind farm; and to group the wind turbines in the wind farm according to the wake characteristics under the dominant wind direction to obtain a number of wind turbine group groups; A typical scene division module is used to generate a number of wind power scenes of the wind farm according to the wind speed data of the wind farm; and perform scene reduction on the generated several wind power scenes to obtain a typical scene of the wind farm; A pre-dispatching instruction determination module is used to determine, for each typical scenario, based on a reinforcement learning algorithm, a pre-dispatching instruction that enables the output power and unit load of the wind farm digital twin model in the typical scenario to meet the objective function; A control parameter optimization module, used to use the pre-scheduling instruction as the initial value of the real-time optimization scheduling and optimize the control parameters using a distributed predictive control algorithm; The wind turbine control module is used to control the operation of each wind turbine based on the optimized control parameters in this typical scenario.
Citation Information
Patent Citations
Wind farm wind turbine combination process based on wake effect clustering
CN103886179A
Wind power plant flow field coupling characteristic evaluation and intelligent grouping method and system
CN111091298A
Distributed wind power plant power optimization method and device
CN111682592A
Yaw optimization control method, device equipment and medium by considering wake flow of wind power plant
CN112459965A
Wind power generation digital twin system
CN113236491A
Cited By
Wind turbine generator array wake flow management method and system for Doppler laser wind measurement
CN119532106A
Wind power plant yaw control method, system and device based on layered optimization and medium
CN119755011A
Wind farm yaw control method, system, device and medium based on hierarchical optimization
CN119755011B
Coordination control method and system for offshore wind plant and storage medium
CN119891361A
A coordinated control method, system and storage medium for an offshore wind farm
CN119891361B
