Offshore wind power medium-and-long-term electric quantity evaluation method and system considering fluctuation and extreme weather influence

Abstract

The invention relates to the technical field of power system planning, in particular to an offshore wind power medium-and-long-term electric quantity evaluation method and system considering fluctuation and extreme weather influence. According to the method, the fluctuation of the medium-and-long-term generating capacity of the offshore wind power and the influence of the extreme weather are considered, the influence of the fluctuation of the medium-and-long-term generating capacity of the offshore wind power and the influence of the extreme weather on the medium-and-long-term generating capacity of the offshore wind power are effectively quantified, and the evaluation precision of the model is improved. According to the method, the alpha evolutionary algorithm is introduced for the first time to carry out hyper-parameter optimization on the deep learning model, and the GRU hyper-parameter accurate tuning is realized based on the alpha algorithm, so that the evaluation accuracy of the model is further improved.

Description

Technical Field

[0001] This invention relates to the field of power system planning technology, and in particular to a method and system for assessing the medium- and long-term power output of offshore wind power, taking into account volatility and the impact of extreme weather. Background Technology

[0002] In recent years, offshore wind power has become an important support for my country's energy transition due to its advantages such as being pollution-free, renewable, and located near load centers. Current research on offshore wind farm power assessment both domestically and internationally mainly focuses on the following aspects: first, long-term assessment methods based on reanalysis data, such as using JRA-55 reanalysis data to assess wind energy resources; second, assessment methods based on numerical weather prediction models, such as using the WRF model to analyze wind energy resource characteristics; and third, assessment methods based on machine learning, such as using support vector machines and artificial neural networks for short-term power prediction.

[0003] However, existing assessment methods have the following shortcomings: 1) They do not adequately consider the complexity of the marine environment, especially the impact of marine meteorological factors such as waves and currents; 2) They do not adequately consider the impact of extreme weather (such as strong winds and waves caused by typhoons), resulting in low reliability of assessment results under extreme weather conditions; 3) They lack effective medium- and long-term assessment models, especially for medium- and long-term power generation. Existing methods mainly focus on short-term (within 24 hours) power forecasts, which are difficult to provide long-term support for project planning and operation.

[0004] While deep learning methods have achieved significant results in short-term wind power forecasting, their application in medium- and long-term power assessment is relatively rare. Among these methods, Gated Recurrent Module (GRU) networks have demonstrated good performance in time series forecasting due to their advantages of small parameter count and high training efficiency. However, there are currently no studies applying GRU networks to the medium- and long-term power assessment of offshore wind farms that consider the impacts of volatility and extreme weather.

[0005] Therefore, there is an urgent need to propose a medium- to long-term power generation assessment method that can fully consider the volatility and extreme weather effects of offshore wind farms, in order to meet the practical needs of offshore wind farm planning, construction, operation, and management. This invention addresses the aforementioned technical problems by proposing a method and system for assessing the medium- to long-term power generation of offshore wind power that takes into account the volatility and extreme weather effects. Summary of the Invention

[0006] To address the problems in existing technologies, this invention provides a method and system for assessing the medium- and long-term power output of offshore wind power, taking into account volatility and the impact of extreme weather. The specific technical solution is as follows: A method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather, includes the following steps: Step S1: Obtain multi-dimensional marine meteorological data and calculate wind power resource potential indicators, wind power output volatility indicators, and extreme weather impact indicators; Step S2: Perform feature filtering on the multidimensional meteorological data obtained in step S1 and the three calculated indicators to obtain a multidimensional input feature set; Step S3: Input the multi-dimensional input features into the GRU model to learn the nonlinear relationship between the multi-dimensional input feature set and medium- and long-term wind power output; Step S4: Use the Alpha Evolutionary Algorithm to optimize the hyperparameters of the GRU model, and use the optimized GRU model to perform medium- and long-term offshore wind power power assessment.

[0007] Preferably, the wind power resource potential index in step S1 is specifically wind energy density, and its calculation formula is as follows: ; In the formula: P Wind energy density; ρ Represents air density, V The wind speed at the hub height of the offshore wind turbine.

[0008] Preferably, the wind power output volatility index in step S1 specifically includes the coefficient of variation of wind energy resources and the monthly variation coefficient of wind energy. The formula for calculating the coefficient of variation of wind energy resources is as follows: ; In the formula: CV Represents the coefficient of variation. σ P The monthly standard deviation representing wind energy resources. This represents the monthly average of wind energy resources. The formula for calculating the monthly variation coefficient of wind energy is as follows: ; In the formula: MVI Represents the monthly variation coefficient of wind energy; P i max , P i min Representing the first i The highest daily average wind energy density within the month, the first i The lowest daily average wind energy density within a month.

[0009] Preferably, the extreme values ​​of wind speed and ocean waves, the extreme weather impact indicators in step S1, are represented as follows: ; ; In the formula: v extreme , h extreme These represent the extreme values ​​of wind speed and wave height, respectively. v daily ( i ) indicates the first i The wind speed of the day, h daily ( i ) Show the first i The waves are high in the sky; Percentile 99 (·) represents the 99th percentile of the values ​​within n days.

[0010] Preferably, step S2 is as follows: In a two-dimensional plane X and Y The data scatter plot is divided into a × b 1 grid, of which a represent X Number of intervals in the direction, b represent Y The number of intervals to divide the direction; For each feature in the feature set, calculate its MIC value with the medium- and long-term power generation. The MIC value ranges from [0,1]. Set a threshold to remove features with MIC values ​​below the threshold, and retain the corresponding features with MIC values ​​greater than or equal to the threshold to form the final multidimensional input features for GRU model training. The formula for calculating MIC is: ; ; In the formula: I ( X , Y ) as a feature X With power generation Y The maximum mutual information coefficient between them; M ( C ) a , b In the dataset C The following is adopted a × b Normalized mutual information value during mesh generation; I *( C , a , b )for X , Y In the dataset CMaximum mutual information; N This represents the total number of samples. C ( N ) is a function for the number of samples.

[0011] Preferably, step S3 specifically includes the following steps: (1) Constructing the training sample set: Let the multidimensional input feature vector be... x t ,in t Indicates the first t Monthly; simultaneously, historical actual power generation for the corresponding month is collected as a tag value. y t , constitute training sample pairs ( x t , y t ); (2) Initialize the GRU model structure: The GRU model includes a reset gate. r t Update door z t Candidate hidden states h t and hidden state s t Among them, the hidden state s t Used to capture and transmit timing information, initial hidden state s 0 is set as the zero vector; (3) Forward propagation calculation: For the first t Input feature vectors for one month x t First, calculate the reset door. r t and the update gate z t Used to control the degree of retention and forgetting of historical information, reset the door. r t and the update gate z t The calculation formulas are as follows: ; ; In the formula: σ It is the sigmoid function; s t-1 Represents the hidden state vector of the previous month; U z , U r These are the weight parameter matrices for the input information in the update gate and the reset gate, respectively. Wz , W r These are the weight parameter matrices of the hidden states in the previous time step for the update gate and the reset gate, respectively. (4) Subsequently, the candidate hidden states are calculated. h t It is used to integrate the current month's feature information with historical memory, and is reset by the gate. r t Compared to the previous state of stealth. s t-1 The result of multiplying the elements: ; in: Uh and Wh Represents the weight parameter matrix; · represents element-wise multiplication; reset gate. rt Hidden state of history st-1 Selective resets can be performed to enable the model to respond flexibly to extreme weather events; (5) Further calculate the current hidden state s t It comprehensively retains both historical and current information; s t It is the update gate of the current time step. z t The hidden state at the previous time step s t-1 and the candidate hidden states at the current time step h t The candidate hidden state is calculated based on the convex combination of elements, and the formula for calculating the hidden state is as follows: ; Where: · represents element-wise product; update gate z t Controlling historical hidden states s t-1 and candidate hidden states h t The fusion ratio is used to capture the seasonal fluctuations in power generation. (6) Finally, the hidden state is hidden through a fully connected layer. s t Mapped to the first t Monthly power generation forecast ŷ t : ; in, W o This is the output layer weight matrix; b o Represents the bias vector; (7) Parameter training and optimization: Calculate the predicted values ŷ t Compared with actual power generation y t The mean squared error loss function is used to update the weight matrix using backpropagation and gradient descent. U z , U r , U h , W z , W r , W h , W o and bias vector b o By iteratively training to minimize the loss function, the GRU learns the nonlinear mapping relationship between multidimensional input features and medium- and long-term power generation.

[0012] Preferably, step S4, which uses the Alpha Evolutionary Algorithm to optimize the hyperparameters of the GRU model, specifically includes the following steps: (1) The hyperparameter optimization problem of the GRU model is formalized into a continuous optimization problem, and its expression is as follows: ; in: X Represents the decision variable vector, i.e. d A vector of hyperparameters to be optimized, including the number of hidden layer neurons, the number of network layers, the learning rate, the batch size, and the dropout rate; f ( X () represents the objective function, used to evaluate the combination of hyperparameters. X Performance; l_val Represents the loss function on the validation set; GRU( X ) represents using hyperparameters X The configured GRU model; Ω is the feasible region defined by hyperparameter constraints, Ω∈ Rd ; 2) An initial solution is generated based on the optimization space of the hyperparameters to be optimized. The calculation formula is as follows: ; in: Xi Indicates the first i There are 10 candidate solutions; lb and ub These are the lower and upper bound vectors for each hyperparameter; rand(0,1,[1, D ]), i =1,2,…, N ; N It refers to population size.D Represents the hyperparameter dimension; Discrete parameters need to be rounded down: ; in: This represents the rounding down operation; Xi , j Indicates the first i The th candidate solution vector j The values ​​of the hyperparameters; n A set of indices representing discrete parameters, including the number of hidden layer neurons, the number of network layers, and the batch size; (3) Construct an evolutionary matrix to maintain the continuity of the optimization process; use a two-path mechanism to update the basis vectors, the expression of which is: ; in: Pat and Pbt They represent Two evolutionary paths at any given moment; and It is a dynamic learning rate, and both are equal to , This indicates the number of times the function has been evaluated so far. Indicates the maximum allowed number of evaluations; A is constructed using sampling with replacement. D A square matrix of order B; B is constructed through sampling without replacement. K × D matrix; ω It is a weight vector calculated based on the objective function value; (4) Implementing solution updates: The update of each solution is achieved through the following formula: ; in: P These are basis vectors, providing the reference points for the search; θ It is a control parameter, and can be a scalar or a vector; and All solutions are samples taken from the population and satisfy... f ( Wi )≤ f ( Ei )≤ f ( Li ); α It is the decay factor, which controls the global exploration intensity, and its expression is: ; Δ r i It is a random step size, generated in the following way: ; Where R1 and R2 are random matrices in the interval [0,1], and S is a 0 / 1 random integer matrix; (5) Perform boundary constraint and selection operations. The boundary constraint adopts the symmetric reflection mechanism, and its expression is: ; In the formula: Ei,j Evolutionary matrix E The Middle i The first individual j The value of the nth dimension, i.e., the nth i The th hyperparameter vector in the th hyperparameter vector j The values ​​of the hyperparameters; The selection operation is based on a greedy strategy, and its expression is: ; in, Eit +1 indicates an updated candidate solution; Xkt Indicates the current solution; f (·) represents the objective function value; (6) Based on the above (1) to (5), the hyperparameters of the GRU model are optimized, and based on the obtained optimal hyperparameter combination, the GRU model is trained using multidimensional input feature data to achieve accurate evaluation of the medium and long-term power generation of offshore wind power.

[0013] An assessment system for medium- and long-term offshore wind power capacity taking into account volatility and the impact of extreme weather, comprising the method described herein: The data processing module is used to acquire multi-dimensional marine meteorological data and calculate wind power resource potential indicators, wind power output volatility indicators, and extreme weather impact indicators. The feature processing module is used to perform feature filtering on the acquired multidimensional meteorological data and the three calculated indicators to obtain a multidimensional input feature set. The model training module is used to input multi-dimensional input features into the GRU model to learn the nonlinear relationship between the multi-dimensional input feature set and medium- and long-term wind power output. The model optimization module is used to optimize the hyperparameters of the GRU model using the Alpha Evolutionary Algorithm, and to evaluate the power generation of offshore wind power in the medium and long term based on the GRU model optimized by the Alpha Evolutionary Algorithm.

[0014] A computer-readable storage medium, characterized in that the computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device where the computer-readable storage medium is located to execute a method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather.

[0015] A processor for running a program, wherein the program, when running, executes the aforementioned method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention takes into account the volatility of medium- and long-term offshore wind power generation and the impact of extreme weather, effectively quantifying the volatility of medium- and long-term offshore wind power generation and the impact of extreme weather on medium- and long-term offshore wind power generation, thereby improving the evaluation accuracy of the model.

[0017] This invention introduces deep learning models into the field of medium- and long-term power generation assessment of offshore wind power. In the context of the application of a high proportion of power electronic equipment, it can make full use of the large-scale datasets of offshore wind farms to achieve accurate assessment of medium- and long-term power generation of offshore wind power.

[0018] This invention introduces the Alpha Evolutionary Algorithm (AEL) for the first time to optimize hyperparameters in deep learning models. The AEL possesses excellent balancing ability between global exploration and local fine-tuning, while remaining invariant to translations and scaling of the search space. This allows it to effectively handle hyperparameters at different scales and also exhibits dimensionality adaptability, performing exceptionally well in high-dimensional optimization problems and making it suitable for the joint optimization of multiple hyperparameters. Based on the AEL, precise tuning of GRU hyperparameters is achieved, further improving the model's evaluation accuracy. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0020] Figure 1 This is a flowchart of the method of the present invention.

[0021] Figure 2 This is a structural diagram of the GRU model.

[0022] Figure 3 The impact of derived features on the evaluation model should be considered.

[0023] Figure 4 To evaluate the effectiveness of different models.

[0024] Figure 5 This is a system schematic diagram of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0027] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0028] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0029] Example 1: like Figure 1 As shown, this embodiment provides a method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather, including the following steps: Step S1: Obtain multidimensional marine meteorological data from the ERA5 system and calculate wind power resource potential index, wind power output volatility index, and extreme weather impact index.

[0030] The specific indicator of wind power resource potential is wind energy density. This indicator can effectively assess the volume of wind energy resources, and its calculation formula is as follows: ; In the formula: P Wind energy density; ρ The air density is represented by 1.225 kg / m³ in this invention. 3 ), V The wind speed at the hub height of the offshore wind turbine.

[0031] The volatility indicators for wind power production specifically include the coefficient of variation of wind energy resources and the monthly variation coefficient of wind energy. The high volatility and strong randomness of wind power production are also significant factors affecting the accuracy of medium- and long-term wind power assessments. The coefficient of variation is used to reflect monthly fluctuations in wind energy resources, and the formula for calculating the coefficient of variation of wind energy resources is as follows: ; In the formula: CV Represents the coefficient of variation. σ P The monthly standard deviation representing wind energy resources. Represents the monthly average of wind energy resources; due to CV It is a negative indicator, low. CV The value indicates higher power generation stability.

[0032] The formula for calculating the monthly variation coefficient of wind energy is as follows: ; In the formula: MVI Represents the monthly variation coefficient of wind energy; P i max , P i min Representing the first i The highest daily average wind energy density within the month, the first i The lowest daily average wind energy density within a month.

[0033] Another important factor affecting offshore wind power output is extreme weather. For offshore wind farms, extreme conditions include strong winds and large waves, often caused by typhoons. The extreme values ​​of wind speed and sea waves are used as characteristics of the impact of extreme weather on wind farm output, as shown below: ; ; In the formula: v extreme , h extreme These represent the extreme values ​​of wind speed and wave height, respectively. v daily ( i ) indicates the first i The wind speed of the day, h daily ( i ) Show the first i The waves are high in the sky; Percentile 99 (·) represents the 99th percentile of the values ​​within n days.

[0034] Step S2 involves performing feature filtering on the multidimensional meteorological data obtained in step S1 and the three calculated indicators to obtain a multidimensional input feature set.

[0035] The meteorological features provided by the ERA5 system itself include wind speed at 10m, wind direction at 10m, wind speed at 100m, and wind direction at 100m. With the introduction of derived features, the multidimensional feature set has a high dimensionality, which can negatively impact the accuracy of model evaluation and increase computational costs. Therefore, feature screening of the multidimensional feature set is necessary. This invention uses the maximum mutual information coefficient (MIC) for feature screening, that is, by setting a certain threshold, features with low correlation to medium- and long-term wind power production are eliminated.

[0036] The maximum mutual information coefficient (MIC) is calculated between multidimensional features within a feature set and wind power output, thus measuring the correlation between multidimensional features and wind power output. MIC, through grid partitioning, effectively uncovers nonlinear relationships between variables in a dataset. The MIC value ranges from [0,1]; the closer the MIC is to 1, the stronger the correlation between two variables, and vice versa. The core idea of ​​MIC is: in a dataset... C There exists a set of two-dimensional variables ( X , Y ),exist X direction and Y Directions are divided separately a and b If there are intervals, then in a two-dimensional plane, they form ab Each grid. Then calculate... X and Y Maximum mutual information coefficient I ( X , Y The maximum mutual information coefficient between multidimensional features and medium- to long-term wind power output is analyzed, and a certain threshold is set to filter out the features corresponding to the maximum mutual information coefficient below this threshold, thus obtaining the final input feature set used to train the model.

[0037] Step S2 in this embodiment is as follows: For characteristic variables X (e.g., a characteristic such as wind energy density, coefficient of variation, etc.) and the target variable Y The MIC calculation between (medium- and long-term wind power production) is first performed on a two-dimensional plane. X and Y The data scatter plot is divided into a × b 1 grid, of which a represent X Number of intervals in the direction, b represent YThe number of directional intervals; finding a partitioning scheme that maximizes mutual information by traversing different grid partitioning methods; For each feature in the feature set, calculate its MIC value with the medium- and long-term power generation. The MIC value ranges from [0,1]. Set a threshold (0.2 in this invention), remove features with MIC values ​​lower than the threshold, and retain the corresponding features with MIC values ​​greater than or equal to the threshold to form the final multidimensional input features for GRU model training; The formula for calculating MIC is: ; ; In the formula: I ( X , Y ) as a feature X With power generation Y The maximum mutual information coefficient between them; M ( C ) a , b In the dataset C The following is adopted a × b Normalized mutual information value during mesh generation; I *( C , a , b )for X , Y In the dataset C Maximum mutual information; N This represents the total number of samples, i.e., the number of historical months. C ( N ) is a function for the number of samples, used to limit the complexity of grid partitioning, typically taken as C( N )= N 0.6. a and b The search scope is affected C ( N Constraints, i.e., requirements ab < C ( N ).

[0038] Step S3: Input the multidimensional input features into the GRU model to learn the nonlinear relationship between the multidimensional input feature set and medium- and long-term wind power output.

[0039] Based on the GRU model, which learns and fits the nonlinear relationship between multidimensional input features and medium- to long-term wind power output, the GRU model has unique advantages in offshore wind farm power assessment: its gating mechanism can adaptively handle long-term dependencies, making it particularly suitable for capturing seasonal variations in wind farm power generation and dynamic characteristics of weather systems; its state update mechanism can effectively balance the retention of historical information and the integration of new information, and has good responsiveness to the impact of extreme weather. Specifically, it includes the following steps: (1) Constructing a training sample set: Collect ERA5 meteorological data for historical periods, calculate derived features such as wind energy density, coefficient of variation, monthly variation coefficient, extreme wind speed, and extreme wave height according to claim 2, combine them with the original meteorological features, and after feature filtering, form a multidimensional input feature vector. x t ,in t Indicates the first t Monthly; simultaneously, historical actual power generation for the corresponding month is collected as a tag value. y t , constitute training sample pairs ( x t , y t ); (2) Initialize the GRU model structure: The GRU model includes a reset gate. r t Update door z t Candidate hidden states h t and hidden state s t Among them, the door update z t This is responsible for controlling the degree to which historical information is retained, enabling the model to capture seasonal fluctuations in wind farm power generation and maintain a memory of long-term climate characteristics; while the reset gate... r t It is responsible for selectively adjusting historical information, enabling the model to quickly respond to the impact of extreme weather events and adjust its assessment of abnormal meteorological conditions in a timely manner. The hidden state... s t Used to capture and transmit timing information, initial hidden state s 0 is set as the zero vector; (3) Forward propagation calculation: For the first t Input feature vectors for one month x t First, calculate the reset door. r t and the update gate z t Used to control the degree of retention and forgetting of historical information, reset the door.r t and the update gate z t The calculation formulas are as follows: ; ; In the formula: σ It is the sigmoid function; s t-1 Represents the hidden state vector of the previous month; U z , U r These are the weight parameter matrices for the input information in the update gate and the reset gate, respectively. W z , W r These are the weight parameter matrices of the hidden states in the previous time step for the update gate and the reset gate, respectively. (4) During the operation of offshore wind farms, their power generation is affected by a variety of complex factors. The candidate hidden state and its generation mechanism can facilitate the comprehensive processing of the interaction of multi-source meteorological data while considering the dynamic changes of environmental factors. Subsequently, the candidate hidden states are calculated. h t It is used to integrate the current month's feature information with historical memory, and is reset by the gate. r t Compared to the previous state of stealth. s t-1 The result of multiplying the elements: ; in: Uh and Wh Represents the weight parameter matrix; · represents element-wise multiplication; reset gate. rt Hidden state of history st-1 Selective resets can be performed to enable the model to respond flexibly to extreme weather events; (5) Further calculate the current hidden state s t It comprehensively retains both historical and current information; s t It is the update gate of the current time step. z t The hidden state at the previous time step s t-1 and the candidate hidden states at the current time step h t The candidate hidden state is calculated based on the convex combination of elements, and the formula for calculating the hidden state is as follows: ; Where: · represents element-wise product; update gate z t Controlling historical hidden states s t-1 and candidate hidden states h t The fusion ratio is used to capture the seasonal fluctuations in power generation. (6) Finally, the hidden state is hidden through a fully connected layer. s t Mapped to the first t Monthly power generation forecast ŷ t : ; in, W o This is the output layer weight matrix; b o Represents the bias vector; (7) Parameter training and optimization: Calculate the predicted values ŷ t Compared with actual power generation y t The mean squared error loss function is used to update the weight matrix using backpropagation and gradient descent. U z , U r , U h , W z , W r , W h , W o and bias vector b o By iteratively training to minimize the loss function, the GRU learns the nonlinear mapping relationship between multidimensional input features and medium- and long-term power generation.

[0040] Step S4: The GRU model is optimized for hyperparameters using the Alpha Evolutionary Algorithm to avoid errors in model prediction accuracy caused by subjective parameter settings. The power generation of offshore wind power in the medium and long term is then assessed based on the GRU model optimized by the Alpha Evolutionary Algorithm.

[0041] The hyperparameter optimization of the GRU model using the Alpha Evolutionary Algorithm specifically includes the following steps: (1) The hyperparameter optimization problem of the GRU model is formalized into a continuous optimization problem, and its expression is as follows: ; in:X Represents the decision variable vector, i.e. d A vector of hyperparameters to be optimized, including the number of hidden layer neurons, the number of network layers, the learning rate, the batch size, and the dropout rate; f ( X () represents the objective function, used to evaluate the combination of hyperparameters. X Performance; l_val Represents the loss function on the validation set; GRU( X ) represents using hyperparameters X The configured GRU model; Ω is the feasible region defined by hyperparameter constraints, Ω∈ Rd ; 2) An initial solution is generated based on the optimization space of the hyperparameters to be optimized. The calculation formula is as follows: ; in: Xi Indicates the first i There are 10 candidate solutions; lb and ub These are the lower and upper bound vectors for each hyperparameter; rand(0,1,[1, D ]), i =1,2,…, N ; N This refers to the population size, which is typically set to 40. D Represents the hyperparameter dimension; Discrete parameters need to be rounded down: ; in: This represents the rounding down operation; Xi , j Indicates the first i The th candidate solution vector j The values ​​of the hyperparameters; n A set of indices representing discrete parameters, including the number of hidden layer neurons, the number of network layers, and the batch size; (3) Construct an evolutionary matrix to maintain the continuity of the optimization process; use a two-path mechanism to update the basis vectors, the expression of which is: ; in: Pat and Pbt They represent Two evolutionary paths at any given moment; and It is a dynamic learning rate, and both are equal to , This indicates the number of times the function has been evaluated so far. Indicates the maximum allowed number of evaluations; A is constructed using sampling with replacement.D A square matrix of order B; B is constructed through sampling without replacement. K × D matrix; ω It is a weight vector calculated based on the objective function value; (4) Implementing solution updates: The update of each solution is achieved through the following formula: ; in: P These are basis vectors, providing the reference points for the search; θ It is a control parameter, and can be a scalar or a vector; and All solutions are samples taken from the population and satisfy... f ( Wi )≤ f ( Ei )≤ f ( Li ); α It is the decay factor, which controls the global exploration intensity, and its expression is: ; Δ r i It is a random step size, generated in the following way: ; Where R1 and R2 are random matrices in the interval [0,1], and S is a 0 / 1 random integer matrix; (5) Perform boundary constraint and selection operations. Boundary constraints can ensure the feasibility of the solution. Here, a symmetric reflection mechanism is adopted. This approach not only ensures the feasibility of the solution but also maintains the continuity and smoothness of the search. Its expression is: ; In the formula: Ei,j Evolutionary matrix E The Middle i The first individual j The value of the nth dimension, i.e., the nth i The th hyperparameter vector in the th hyperparameter vector j The values ​​of the hyperparameters; The selection operation is based on a greedy strategy to ensure monotonic improvement of the population and preservation of elite solutions. It also forms the basis for the convergence proof. The expression for the selection operation is: ; in, Eit +1 indicates an updated candidate solution; Xkt Indicates the current solution; f (·) represents the objective function value; (6) Based on the above (1) to (5), the hyperparameters of the GRU model are optimized, and based on the obtained optimal hyperparameter combination, the GRU model is trained using multidimensional input feature data to achieve accurate evaluation of the medium and long-term power generation of offshore wind power.

[0042] This embodiment uses offshore wind power meteorological data from 1950 to 2019 in a certain region as an example to calculate the three derived characteristics of wind power resource potential, wind power output volatility, and extreme weather impact proposed in this invention. All data are normalized.

[0043] Based on the obtained three aspects of derived features, this embodiment combines the original feature set with feature filtering and sets the elimination threshold to 0.2. The resulting multidimensional input features are shown in Table 1.

[0044] Table 1 Multidimensional Input Features Based on the obtained multidimensional input features, a GRU model is used to fit the nonlinear relationship between the multidimensional input features and the medium- and long-term offshore wind power generation. Simultaneously, an alpha evolution algorithm is used to optimize multiple hyperparameters of the GRU model, and the final medium- and long-term power generation assessment is achieved based on the optimized hyperparameters. The boundary settings for hyperparameter optimization are shown in Table 2.

[0045] Table 2 Hyperparameter Optimization Boundaries The impact of the designed derived features on the accuracy of the evaluation model was analyzed. A GRU was trained based on the obtained optimal hyperparameters, and medium- to long-term electricity assessments were performed. The AE-GRU model was trained using features other than wind energy density, monthly fluctuation coefficient, coefficient of variation, extreme wind speed, and extreme wave height. The evaluation accuracy was compared with that of the AE-GRU trained using all input features. The evaluation index values ​​are shown in Table 3. Figure 3 As shown.

[0046] Depend on Figure 3It is evident that the AE-GRU model, which does not consider derived features, significantly overestimates medium- to long-term offshore wind power generation. In contrast, the assessment model proposed in this invention, which considers derived features, shows a better fit to actual offshore wind power generation and better reflects the trend of medium- to long-term wind power generation. Furthermore, as shown in Table 3, the AE-GRU model proposed in this invention, which considers derived features, outperforms the AE-GRU model without considering derived features in all three metrics: MAPE (reflecting overall accuracy), MAE (reflecting overall error level), and RMSE (reflecting the impact of extreme values). In conclusion, this demonstrates the effectiveness of the derived features introduced in this invention, which measure the impact of volatility and extreme weather, in improving the model's assessment accuracy.

[0047] Table 3 Performance metrics for evaluating the model The model of this invention is compared and analyzed with traditional models and currently more advanced models. The evaluation results of different methods are shown below. Figure 4 The evaluation metrics for the proposed model are calculated and shown in Table 4. As can be seen from Table 4, the AE-GRU model proposed in this invention achieves the best evaluation performance, followed by Copula-LSTM, then the Markov modified model, and the traditional SARIMA model performs the worst. One reason for this is that the proposed model uses the AE optimization algorithm to finely adjust the hyperparameters; another reason may be the relatively small size of the dataset used, with only about 400 samples in the training set. On such a small to medium-sized dataset, GRU, with its more refined gating structure, can achieve better evaluation results than LSTM. Finally, combining... Figure 4 It can be more intuitively observed that the evaluation model proposed in this invention can effectively capture the seasonal and interannual fluctuations in power generation, and can better fit the actual power generation curve. Copula-LSTM can also basically capture the medium- and long-term power generation trends, but the traditional model SARIMA and the model based on Markov corrections fail to capture this trend well, especially from August to October, where the evaluation results of these two models are very poor. In summary, the effectiveness of the evaluation model proposed in this invention for evaluating the medium- and long-term power generation of offshore wind power has been verified.

[0048] Table 4 Performance metrics for evaluating the model Example 2: like Figure 5 As shown, based on the same inventive concept as Embodiment 1, this embodiment provides an assessment system for the medium- and long-term power generation of offshore wind power that takes into account volatility and the impact of extreme weather. The method described includes: The data processing module is used to acquire multi-dimensional marine meteorological data and calculate wind power resource potential indicators, wind power output volatility indicators, and extreme weather impact indicators. The feature processing module is used to perform feature filtering on the acquired multidimensional meteorological data and the three calculated indicators to obtain a multidimensional input feature set. The model training module is used to input multi-dimensional input features into the GRU model to learn the nonlinear relationship between the multi-dimensional input feature set and medium- and long-term wind power output. The model optimization module is used to optimize the hyperparameters of the GRU model using the Alpha Evolutionary Algorithm, and to evaluate the power generation of offshore wind power in the medium and long term based on the GRU model optimized by the Alpha Evolutionary Algorithm.

[0049] Example 3: Based on the same inventive concept as Embodiment 1, this embodiment provides a computer-readable storage medium, characterized in that the computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device where the computer-readable storage medium is located to execute a method for evaluating the medium- and long-term power generation of offshore wind power that takes into account volatility and the impact of extreme weather.

[0050] Example 4: Based on the same inventive concept as Embodiment 1, this embodiment provides a processor for running a program, wherein the program executes the aforementioned method for evaluating the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather.

[0051] Those skilled in the art will recognize that the modules of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0052] In the embodiments provided by this invention, it should be understood that the division of modules is only a logical functional division. In actual implementation, there may be other division methods, such as multiple modules can be combined into one module, one module can be split into multiple modules, or some features can be ignored.

[0053] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0054] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather, characterized in that... Includes the following steps: Step S1: Obtain multi-dimensional marine meteorological data and calculate wind power resource potential indicators, wind power output volatility indicators, and extreme weather impact indicators; Step S2: Perform feature filtering on the multidimensional meteorological data obtained in step S1 and the three calculated indicators to obtain a multidimensional input feature set; Step S3: Input the multi-dimensional input features into the GRU model to learn the nonlinear relationship between the multi-dimensional input feature set and medium- and long-term wind power output; Step S4: Use the Alpha Evolutionary Algorithm to optimize the hyperparameters of the GRU model, and use the optimized GRU model to perform medium- and long-term offshore wind power power assessment.

2. The method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather, as described in claim 1, is characterized in that... In step S1, the wind power resource potential index is specifically wind energy density, and its calculation formula is as follows: ； In the formula: P Wind energy density; ρ Represents air density, V The wind speed at the hub height of the offshore wind turbine.

3. The method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather, as described in claim 1, is characterized in that... The wind power output volatility indicators in step S1 specifically include the coefficient of variation of wind energy resources and the monthly variation coefficient of wind energy. The formula for calculating the coefficient of variation of wind energy resources is as follows: ； In the formula: CV Represents the coefficient of variation. σ P The monthly standard deviation representing wind energy resources. This represents the monthly average of wind energy resources. The formula for calculating the monthly variation coefficient of wind energy is as follows: ； In the formula: MVI Represents the monthly variation coefficient of wind energy; P i max , P i min Representing the first i The highest daily average wind energy density within the month, the first i The lowest daily average wind energy density within a month.

4. The method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather, as described in claim 1, is characterized in that... The extreme values ​​of wind speed and ocean waves, which are specific indicators of the impact of extreme weather in step S1, are expressed as follows: ； ； In the formula: v extreme , h extreme These represent the extreme values ​​of wind speed and wave height, respectively. v daily ( i ) indicates the first i The wind speed of the day, h daily ( i ) Show the first i The waves are high in the sky; Percentile 99 (·) represents the 99th percentile of the values ​​within n days.

5. The method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather, as described in claim 1, is characterized in that... Step S2 is as follows: In a two-dimensional plane X and Y The data scatter plot is divided into a × b 1 grid, of which a represent X Number of intervals in the direction, b represent Y The number of intervals to divide the direction; For each feature in the feature set, calculate its MIC value with the medium- and long-term power generation. The MIC value ranges from [0,1]. Set a threshold to remove features with MIC values ​​below the threshold, and retain the corresponding features with MIC values ​​greater than or equal to the threshold to form the final multidimensional input features for GRU model training. The formula for calculating MIC is: ； ； In the formula: I ( X , Y ) as a feature X With power generation Y The maximum mutual information coefficient between them; M ( C ) a , b In the dataset C The following is adopted a × b Normalized mutual information value during mesh generation; I *( C , a , b )for X , Y In the dataset C Maximum mutual information; N This represents the total number of samples. C ( N ) is a function for the number of samples.

6. The method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather, as described in claim 1, is characterized in that... Step S3 specifically includes the following steps: (1) Constructing the training sample set: Let the multidimensional input feature vector be... x t ,in t Indicates the first t Monthly; simultaneously, historical actual power generation for the corresponding month is collected as a tag value. y t , constitute training sample pairs ( x t , y t ); (2) Initialize the GRU model structure: The GRU model includes a reset gate. r t Update door z t Candidate hidden states h t and hidden state s t Among them, the hidden state s t Used to capture and transmit timing information, initial hidden state s 0 is set as the zero vector; (3) Forward propagation calculation: For the first t Input feature vectors for one month x t First, calculate the reset door. r t and the update gate z t Used to control the degree of retention and forgetting of historical information, reset the door. r t and the update gate z t The calculation formulas are as follows: ； ； In the formula: σ It is the sigmoid function; s t-1 Represents the hidden state vector of the previous month; U z , U r These are the weight parameter matrices for the input information in the update gate and the reset gate, respectively. W z , W r These are the weight parameter matrices of the hidden states in the previous time step for the update gate and the reset gate, respectively. (4) Subsequently, the candidate hidden states are calculated. h t It is used to integrate the current month's feature information with historical memory, and is reset by the gate. r t Compared to the previous state of stealth. s t-1 The result of multiplying the elements: ； in: Uh and Wh Represents the weight parameter matrix; · represents element-wise multiplication; reset gate. rt Hidden state of history st-1 Selective resets can be performed to enable the model to respond flexibly to extreme weather events; (5) Further calculate the current hidden state s t It comprehensively retains both historical and current information; s t It is the update gate of the current time step. z t The hidden state at the previous time step s t-1 and the candidate hidden states at the current time step h t The candidate hidden state is calculated based on the convex combination of elements, and the formula for calculating the hidden state is as follows: ； Where: · represents element-wise product; update gate z t Controlling historical hidden states s t-1 and candidate hidden states h t The fusion ratio is used to capture the seasonal fluctuations in power generation. (6) Finally, the hidden state is hidden through a fully connected layer. s t Mapped to the first t Monthly power generation forecast ŷ t : ； in, W o This is the output layer weight matrix; b o Represents the bias vector; (7) Parameter training and optimization: Calculate the predicted values ŷ t Compared with actual power generation y t The mean squared error loss function is used to update the weight matrix using backpropagation and gradient descent. U z , U r , U h , W z , W r , W h , W o and bias vector b o By iteratively training to minimize the loss function, the GRU learns the nonlinear mapping relationship between multidimensional input features and medium- and long-term power generation.

7. The method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather, as described in claim 1, is characterized in that... Step S4, which uses the Alpha Evolutionary Algorithm to optimize the hyperparameters of the GRU model, specifically includes the following steps: (1) The hyperparameter optimization problem of the GRU model is formalized into a continuous optimization problem, and its expression is as follows: ； in: X Represents the decision variable vector, i.e. d A vector of hyperparameters to be optimized, including the number of hidden layer neurons, the number of network layers, the learning rate, the batch size, and the dropout rate; f ( X () represents the objective function, used to evaluate the combination of hyperparameters. X Performance; l_ val Represents the loss function on the validation set; GRU( X ) represents using hyperparameters X The configured GRU model; Ω is the feasible region defined by hyperparameter constraints, Ω∈ Rd ; 2) An initial solution is generated based on the optimization space of the hyperparameters to be optimized. The calculation formula is as follows: ； in: Xi Indicates the first i There are 10 candidate solutions; lb and ub These are the lower and upper bound vectors for each hyperparameter; rand(0,1,[1, D ]), i =1,2,…, N ; N It refers to population size. D Represents the hyperparameter dimension; Discrete parameters need to be rounded down: ； in: This represents the rounding down operation; Xi , j Indicates the first i The th candidate solution vector j The values ​​of the hyperparameters; n A set of indices representing discrete parameters, including the number of hidden layer neurons, the number of network layers, and the batch size; (3) Construct an evolutionary matrix to maintain the continuity of the optimization process; use a two-path mechanism to update the basis vectors, the expression of which is: ； in: Pat and Pbt They represent Two evolutionary paths at any given moment; and It is a dynamic learning rate, and both are equal to , This indicates the number of times the function has been evaluated so far. Indicates the maximum allowed number of evaluations; A is constructed using sampling with replacement. D A square matrix of order B; B is constructed through sampling without replacement. K × D matrix; ω It is a weight vector calculated based on the objective function value; (4) Implementing solution updates: The update of each solution is achieved through the following formula: ； in: P These are basis vectors, providing the reference points for the search; θ It is a control parameter, and can be a scalar or a vector; and All solutions are samples taken from the population and satisfy... f ( Wi )≤ f ( Ei )≤ f ( Li ); α It is the decay factor, which controls the global exploration intensity, and its expression is: ； Δ r i It is a random step size, generated in the following way: ； Where R1 and R2 are random matrices in the interval [0,1], and S is a 0 / 1 random integer matrix; (5) Perform boundary constraint and selection operations. The boundary constraint adopts the symmetric reflection mechanism, and its expression is: ； In the formula: Ei,j Evolutionary matrix E The Middle i The first individual j The value of the nth dimension, i.e., the nth i The th hyperparameter vector in the th hyperparameter vector j The values ​​of the hyperparameters; The selection operation is based on a greedy strategy, and its expression is: ； in, Eit +1 indicates an updated candidate solution; Xkt Indicates the current solution; f (·) represents the objective function value; (6) Based on the above (1) to (5), the hyperparameters of the GRU model are optimized, and based on the obtained optimal hyperparameter combination, the GRU model is trained using multidimensional input feature data to achieve accurate evaluation of the medium and long-term power generation of offshore wind power.

8. A system for assessing the medium- and long-term power output of offshore wind power, taking into account volatility and the impact of extreme weather, characterized in that... The method described by any one of claims 1 to 7 includes: The data processing module is used to acquire multi-dimensional marine meteorological data and calculate wind power resource potential indicators, wind power output volatility indicators, and extreme weather impact indicators. The feature processing module is used to perform feature filtering on the acquired multidimensional meteorological data and the three calculated indicators to obtain a multidimensional input feature set. The model training module is used to input multi-dimensional input features into the GRU model to learn the nonlinear relationship between the multi-dimensional input feature set and medium- and long-term wind power output. The model optimization module is used to optimize the hyperparameters of the GRU model using the Alpha Evolutionary Algorithm, and to evaluate the power generation of offshore wind power in the medium and long term based on the GRU model optimized by the Alpha Evolutionary Algorithm.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device containing the computer-readable storage medium to perform an assessment method for medium- and long-term offshore wind power taking into account volatility and the effects of extreme weather, as described in any one of claims 1 to 7.

10. A processor, characterized in that, The processor is used to run a program, wherein the program executes a method for assessing the medium- and long-term power generation of offshore wind power, taking into account volatility and the impact of extreme weather, as described in any one of claims 1 to 7.

Images