A method and system for offshore wind power prediction

Abstract

The application relates to the offshore wind power prediction technical field and discloses an offshore wind power prediction method and system. Based on the wind speed change rate dynamic adjustment of the abnormal detection threshold, the historical data are subjected to abnormal value cleaning to obtain pretreatment data; a limit learning machine model is established, and the network structure parameters of the limit learning machine model are initialized; the input layer weight and the hidden layer threshold in the network structure parameters are optimized by using an improved electric eel foraging optimization algorithm to generate an optimized limit learning machine model; the pretreatment data are input into the optimized limit learning machine model for training to construct an offshore wind power prediction model; and the real-time offshore wind power of a target port is predicted. The generalization performance of the model under complex sea conditions is effectively improved, more accurate power prediction results are provided for power grid dispatching, and the engineering timeliness requirement of offshore wind farm real-time prediction is met.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of offshore wind power prediction technology, in particular to an offshore wind power prediction method and system. BACKGROUND

[0002] With the transformation of energy structure to clean, offshore wind power has become an important direction of renewable energy development due to its resource endowment and location advantage. Compared with onshore wind power, the wind speed of marine environment is higher and more stable, which can significantly improve the power generation efficiency. At the same time, the construction of offshore wind farm does not need to occupy land resources, which is especially suitable for coastal economically developed areas with scarce land. In addition, as the center of population and industrial load, the nearby consumption of offshore wind power can effectively reduce the power transmission loss and improve the energy utilization efficiency. However, the large-scale grid connection of offshore wind power also faces severe challenges: its output is affected by marine weather, equipment state and grid connection conditions, and has significant randomness and volatility. If the power prediction accuracy is insufficient, it will lead to an increase in power grid frequency modulation pressure and an increase in standby capacity demand, and even threaten the safe and stable operation of the power system. Therefore, high-precision offshore wind power prediction technology is the key foundation for realizing efficient consumption of offshore wind power and reliable regulation of power grid.

[0003] The core goal of wind power prediction is to analyze historical operation data and real-time environmental parameters, establish the mapping relationship between wind farm output and influencing factors, and then realize the accurate estimation of future power generation. This technology has multiple values for power system operation: at the power grid dispatching level, accurate prediction results can optimize power generation planning, reduce the frequency modulation burden of traditional power sources such as thermal power, thereby reducing system operation cost; at the equipment operation and maintenance level, power prediction can be linked with wind turbine state monitoring system to identify potential faults through output anomaly, and provide decision basis for preventive maintenance; at the power market level, prediction accuracy directly affects the economic benefits of wind power participating in spot trading and ancillary services, and high-precision prediction helps to improve the market competitiveness of wind power.

[0004] Therefore, there is an urgent need for an offshore wind power prediction method that can efficiently and accurately predict offshore wind power. SUMMARY

[0005] In order to solve the above technical problems, the present application provides an offshore wind power prediction method, which comprises the following steps:

[0006] S1: obtaining historical data of offshore wind power;

[0007] S2: dynamically adjusting the abnormal value detection threshold based on the wind speed change rate, performing abnormal value cleaning on the historical data to obtain preprocessed data;

[0008] S3: Establish an extreme learning machine model, initialize the network structure parameters of the extreme learning machine model;

[0009] S4: Optimize the input layer weight and hidden layer threshold in the network structure parameters using the improved electric eel foraging optimization algorithm, to generate an optimized extreme learning machine model;

[0010] S5: Input the preprocessed data into the optimized extreme learning machine model for training, to build a sea wind power prediction model;

[0011] S6: Use the sea wind power prediction model to predict the real-time sea wind power of the target port.

[0012] Preferably, the calculation method of dynamically adjusting the abnormality detection threshold based on the wind speed change rate is:

[0013] ;

[0014] In the formula, Uppert represents the upper limit of the abnormal value judgment, Lowert represents the lower limit of the abnormal value judgment, Q 75 represents the upper quartile, Q 25 represents the lower quartile, IQR t represents the interquartile range, represents the wind speed change rate, and λ represents the turbulence adjustment coefficient.

[0015] Preferably, the wind speed change rate is calculated using the following formula:

[0016] ;

[0017] In the formula, V t represents the wind speed value at time t, V t-1 represents the wind speed value at time t-1, and V rated represents the rated wind speed value.

[0018] Preferably, the improved electric eel foraging optimization algorithm is implemented by the following strategies: logistic chaotic mapping is used to generate an initial population; a golden sine strategy is introduced to determine the population movement path; and an adaptive random mutation strategy is combined to adjust the individual position of the newly generated individual in each iteration.

[0019] Preferably, the logistic chaotic mapping is used to generate an initial population, specifically as follows:

[0020] A vector is randomly generated, the dimension of the vector is Dim, and the value of the vector i in each dimension is a random number between 0 and 1;

[0021] The logistic chaotic mapping formula is used to generate a sequence [s1, …, si ,…,s N ], the logistic chaotic mapping formula is as follows:

[0022] ;

[0023] In the formula, s i represents the i th vector, s i-1 represents the i-1 th vector, and f represents a control parameter of the chaotic mapping.

[0024] According to the sequence, the initial population is obtained by distribution through the following formula:

[0025] ;

[0026] In the formula, x i represents the position of the i th electric eel corresponding to the i th vector, Ub represents the numerical upper limit of the initial population, and Lb represents the numerical lower limit of the initial population.

[0027] Preferably, the golden sine strategy is introduced to determine the population movement path, specifically, when the electric eel moves from the current position to the target position, the golden sine strategy is used to determine the population movement path, and the population position is updated, and the expression of the golden sine strategy is as follows:

[0028] ;

[0029] In the formula, x i (t) represents the position of the i th electric eel at time t, x i (t+1) represents the position of the i th electric eel at time t+1, T1 is a random parameter with a value range of [0, 2π], T2 is a random parameter with a value range of [0, π], b1 and b2 respectively represent phase parameters, represents the golden section number, X prey represents the global optimal position up to the current iteration.

[0030] Preferably, the self-adaptive random mutation strategy is combined to adjust the position of the newly generated individual in each iteration, specifically:

[0031] ;

[0032] In the formula, and represent the position of the newly generated individual after mutation in the n th dimension and the original position before mutation, m represents the current iteration number, M represents the maximum iteration number, rand represents a random number between 0 and 1, represents the upper boundary of the solution in each iteration, represents the lower boundary of the solution in each iteration.

[0033] Preferably, the fitness value in the improved electric eel foraging optimization algorithm is calculated using the following formula:

[0034] ;

[0035] Where, represents the fitness value, J represents the number of training samples, represents the true value of wind power in the jth training sample, Represents the predicted value of wind power in the jth training sample.

[0036] Preferably, the historical data includes wind speed, wind direction and offshore power generation.

[0037] On the other hand, the present application further provides an offshore wind power prediction system for executing any of the above-mentioned offshore wind power prediction methods, comprising: a data module, a model building module, and an output module;

[0038] The data module is used to obtain historical data of offshore wind power; dynamically adjust the abnormality detection threshold based on the wind speed change rate, clean the historical data of abnormal values, and obtain pre-processed data;

[0039] The model construction module is connected to the data module and is used to establish an extreme learning machine model and initialize the network structure parameters of the extreme learning machine model; use an improved electric eel foraging optimization algorithm to optimize the input layer weights and hidden layer thresholds in the network structure parameters to generate an optimized extreme learning machine model; input the preprocessed data into the optimized extreme learning machine model for training to construct an offshore wind power prediction model;

[0040] The output module is connected to the model building module and is used to predict the real-time offshore wind power of the target port by using the offshore wind power prediction model.

[0041] The embodiments of the present application have the following technical effects:

[0042] The offshore wind power prediction method provided by the application effectively improves the accuracy and robustness of offshore wind power prediction by fusing dynamic threshold cleaning and intelligent optimization algorithm. First, based on the wind speed change rate, the abnormal detection threshold is dynamically adjusted, which can adapt to the turbulence characteristics of the marine environment, accurately remove abnormal data while retaining the sudden change characteristics of the wind speed, and solve the problem of over-cleaning or insufficient cleaning of the traditional fixed threshold method in complex sea conditions, providing high-quality input for model training. Second, the improved electric eel foraging optimization algorithm is used to globally optimize the key parameters of the extreme learning machine, breaking through the limitation of traditional random initialization which is easy to fall into local optimum, and through optimization of the parameter combination of input layer weight and hidden layer threshold, the model's representation ability for the nonlinear power characteristics of offshore wind power is enhanced. Compared with the conventional prediction method, the technical scheme keeps the advantages of fast training of the extreme learning machine, and through the synergistic innovation of data cleaning and parameter optimization, the generalization performance of the model in complex sea conditions is effectively improved, providing more accurate power prediction results for power grid dispatching, and meeting the engineering timeliness requirements of real-time prediction of offshore wind farms. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the specific embodiments or prior art of the present application, the drawings needed in the description of the specific embodiments or prior art will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0044] Figure 1 is a step schematic diagram of an offshore wind power prediction method provided by an embodiment of the present application;

[0045] Figure 2 is a model framework schematic diagram of an offshore wind power prediction model provided by an embodiment of the present application;

[0046] Figure 3 is a flowchart of an offshore wind power prediction method provided by an embodiment of the present application;

[0047] Figure 4 is a framework diagram of the offshore wind power prediction model provided by an embodiment of the present application;

[0048] Figure 5 is a prediction result schematic diagram of the offshore wind power prediction model provided by an embodiment of the present application;

[0049] Figure 6 is a prediction result error comparison diagram of different prediction models provided by an embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the purposes, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be clearly and completely described below. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present application.

[0051] It plays an important role for offshore wind power prediction, but the particularity of offshore wind power makes the traditional prediction method face the problem of insufficient adaptability. First, the complexity of the marine environment is far beyond that of the land, and the coupling of meteorological and hydrological conditions such as typhoon, salt fog, and tide leads to stronger spatiotemporal volatility of wind speed and direction, and more significant nonlinear characteristics of power curve. Second, offshore wind farms are usually far from the land, and the deployment of sensors and data transmission face high cost and high delay problems, and historical data may have noise interference or local missing, which puts higher requirements on the robustness of the model. In addition, offshore wind turbines are in a high-humidity and high-salt corrosion environment for a long time, and the equipment aging rate accelerates, and its power characteristics are easy to drift with the running state, so the prediction model needs to have dynamic correction capability.

[0052] Current wind power prediction methods are mainly divided into physical models, statistical models and machine learning. The physical model is based on the modeling of fluid mechanics equations and wind turbine aerodynamic characteristics, and its advantage is that it can analyze the physical process of wind speed propagation, but it depends on high-precision meteorological prediction data and wind turbine parameters, and the complex weather conditions and dynamic changes of equipment parameters on the sea often lead to cumulative model errors. Statistical models (such as ARIMA, Kalman filter, etc.) construct linear mapping relationships through time series analysis, but they lack generalization ability when dealing with nonlinear relationships between wind speed and power, especially in adapting to the multivariate coupling characteristics of offshore wind power. Although machine learning methods represented by neural networks have strong nonlinear fitting ability, their performance is limited by the quality of training data: on the one hand, the high noise and sparsity of offshore wind power data can easily lead to model overfitting; on the other hand, the structure parameters (such as weights and thresholds) of traditional neural networks are usually randomly initialized, and when the data distribution is uneven, it is easy to fall into local optimum and difficult to capture the dynamic change rule of offshore wind power. In addition, the existing methods generally have high computational complexity, and offshore wind power prediction needs to consider real-time requirements, which poses challenges to the lightweight design and optimization efficiency of the model.

[0053] In summary, the existing prediction methods have obvious limitations in dealing with the special needs of offshore wind power. Physical models and statistical methods are difficult to handle high-dimensional nonlinear relationships, and traditional machine learning methods are subject to data quality and parameter sensitivity, resulting in insufficient prediction accuracy and stability. At the same time, the computational efficiency of complex models is difficult to meet the engineering needs of real-time prediction of offshore wind farms. Therefore, an offshore wind power prediction method that takes into account accuracy, robustness and computational efficiency is urgently needed to support the large-scale grid connection of offshore wind farms and the safe and economic operation of power grids.

[0054] Based on this, the present application provides an offshore wind power prediction method, as shown in Figure 1 The method comprises the following steps:

[0055] S1: obtaining historical data of offshore wind power;

[0056] In some embodiments, the historical data includes wind speed, wind direction and offshore power generation. Wherein the wind speed and wind direction are used as the core input features, i.e. the input parameters of the offshore wind power prediction model, and the offshore power generation is used as the target variable of supervised learning, which is also the output parameter of the offshore wind power prediction model.

[0057] S2: dynamically adjusting the abnormal value detection threshold based on the wind speed change rate, and performing abnormal value cleaning on the historical data to obtain preprocessed data;

[0058] In some embodiments, the calculation method of dynamically adjusting the abnormal value detection threshold based on the wind speed change rate is as follows:

[0059] ;

[0060] In the formula, Uppert represents the upper limit of abnormal value judgment, Lowert represents the lower limit of abnormal value judgment, Q 75 represents the upper quartile, Q 25 represents the lower quartile, IQR t represents the interquartile range, represents the wind speed change rate, and λ represents the turbulence adjustment coefficient.

[0061] In some embodiments, the wind speed change rate is calculated using the following formula:

[0062] ;

[0063] In the formula, V t represents the wind speed value at time t, V t-1 represents the wind speed value at time t-1, and V rated represents the rated wind speed value, i.e. the wind speed at which the wind turbine reaches the rated power output, which can be determined according to the technical specifications of the actual wind turbine model of the target port.

[0064] The above outlier cleaning is a preprocessing operation on all categories of data in historical data. The preprocessed wind speed historical data, wind direction historical data and offshore power generation historical data are obtained.

[0065] The dynamic threshold calculation method has the innovation point that the wind speed change rate is introduced as an adjustment factor. The upper quartile Q 75 and the lower quartile Q 25 The reference range of the data distribution is constituted, and the interquartile range IQRt reflects the data dispersion degree. The turbulence adjustment coefficient λ is set according to the characteristics of the sea area, and is used to enlarge or reduce the threshold adjustment range. When the wind speed change rate ΔVt increases, the threshold interval expands with the (1+λΔVt) term, which not only retains the real wind speed fluctuation of extreme weather such as typhoon passage, but also effectively eliminates abnormal jump points. Compared with the fixed threshold method, the dynamic mechanism fully considers the characteristics of large wind speed change scale and easy wind power fluctuation in the sea, clearly reflects the influence mechanism of turbulence intensity on power fluctuation in fluid mechanics, solves the problem of over-cleaning or insufficient cleaning of the traditional fixed threshold method in complex sea conditions, and provides high-quality input for model training and data basis for providing offshore wind power prediction accuracy.

[0066] The offshore wind farm is usually far away from the land, the cost of sensor deployment and real-time data collection is high, and the historical data may be missing. The preprocessed historical data is replaced by the median value in the sliding window to eliminate the abnormal data and missing values.

[0067] S3: establishing an extreme learning machine model, initializing network structure parameters of the extreme learning machine model;

[0068] The offshore wind power is affected by marine weather conditions (such as typhoon, salt spray, tide) and wave dynamics, and the spatial and temporal fluctuations of wind speed and wind direction are higher, resulting in more significant nonlinear characteristics of power output. The present application selects an extreme learning machine (ELM) as a new type of single-layer feedforward neural network designed for high-dimensional nonlinear models to predict offshore wind power. The network structure parameters of ELM are initialized, including input layer weight, hidden layer threshold, hidden layer node and activation function, etc. However, the input layer weight in ELM is randomly generated, usually a random number in uniform distribution or normal distribution. The hidden layer threshold, that is, the bias of each hidden layer neuron. In the activation function, the bias is added after the weighted sum of the input signal, and then processed by the activation function. For example, for a hidden layer node, its output may be the result of the activation function (such as sigmoid or ReLU) acting on the weighted sum of the input data plus the bias. The hidden layer threshold is also randomly set during initialization and remains unchanged during the training process. Randomly setting the input layer weight and the hidden layer threshold will make the prediction result of the power unable to meet the accuracy requirement. Therefore, the improved electric eel foraging optimization algorithm is used to optimize the input layer weight and the hidden layer threshold in the network structure parameters to generate an optimized extreme learning machine model.

[0069] S4: The improved electric eel foraging optimization algorithm is used to optimize the input layer weight and the hidden layer threshold in the network structure parameters to generate an optimized extreme learning machine model.

[0070] The offshore wind power environment is complex, the wind farm is far from the land, the cost of the deployed data collection equipment is high and there may be data missing, the number of input samples is large and the structure is complex, which leads to the ill-conditioned of ELM calculation matrix and the serious decline of network performance, thereby affecting the prediction accuracy. Therefore, the improved electric eel foraging optimization algorithm (IEEFO) is used to optimize the input layer weight and the hidden layer threshold to improve the prediction performance of ELM, and an offshore wind power prediction model based on IEEFO-ELM is established. In the offshore wind power prediction model based on IEEFO-ELM, the input parameters of IEEFO are wind speed and wind direction, the output parameters are the optimized input layer weight and hidden layer threshold of ELM, and the offshore power generation power is output by ELM based on the optimized input layer weight and hidden layer threshold. This method has stronger global search ability and can better adapt to complex data structure and sample distribution.

[0071] In some embodiments, the improved electric eel foraging optimization algorithm is improved by the following strategy: S4A: generating an initial population by logistic chaotic mapping; in some embodiments, the logistic chaotic mapping is used to generate an initial population, specifically:

[0072] Randomly generate a vector, where the dimension of the vector is Dim, and the value of the vector i in each dimension is a random number between 0 and 1;

[0073] Use the logistic chaotic mapping formula to generate a sequence of vectors N [s1,…,s i ,…,s N ], N is also the number of initial solutions generated in the chaotic map generation electric eel foraging optimization algorithm. The logistic chaotic map formula is as follows:

[0074] ;

[0075] Where s i represents the i-th vector, s i-1 represents the i-1th vector, and f represents the control parameter of the chaotic mapping;

[0076] According to the sequence, the initial population is obtained by performing distribution using the following formula:

[0077] ;

[0078] Where x i Represents the position of the i-th electric eel corresponding to the i-th vector, that is, the initial solution corresponding to the electric eel, Ub represents the numerical upper limit of the initial population, and Lb represents the numerical lower limit of the initial population.

[0079] When the Logistic chaotic map is implemented, s i The dimension Dim is the number of variables. The electric eel foraging optimization algorithm optimizes two parameters of the extreme learning machine: the input layer weights and the hidden layer thresholds, so the value of Dim is 2. Dimensionality is a key parameter of the electric eel optimization algorithm. Each individual electric eel in the algorithm is equivalent to a possible solution in the solution space, which contains all the variables that require optimal values. During the chaotic sequence generation process, f is the control parameter of the chaotic map, ranging from 0 to 4. The larger the value of f, the more uniform the map distribution, and the better the initialization effect of the electric eel foraging optimization algorithm. This method effectively utilizes the characteristics of chaotic maps to provide a more random and exploratory initial population for the EEFO algorithm, thereby improving the algorithm's performance.

[0080] The historical data of offshore wind power may be unevenly distributed due to sensor failure or transmission delay, and the initial population randomly initialized may easily cause overfitting or underfitting of the prediction model. The initial population generated by the Logistic chaotic mapping has ergodicity and uniformity, can cover a wider solution space, and avoid local convergence caused by sparse data. This application improves the adaptability of offshore wind power in unstable data quality scenarios, enhances the robustness of the model to noise, provides a better search starting point for subsequent iterations, and improves the optimization effect of input layer weight and hidden layer threshold.

[0081] S4B: introducing a golden sine strategy to determine the population moving path;

[0082] In some embodiments, the golden sine strategy is introduced to determine the population moving path, specifically, when the electric eel moves from the current position to the target position, the golden sine strategy is used to determine the population moving path, and the population position is updated. The expression of the golden sine strategy is:

[0083]

[0084] In the formula, x i (t) represents the position of the i-th electric eel at time t, x i (t+1) represents the position of the i-th electric eel at time t+1, T1 is a random parameter with a value range of [0, 2π], T2 is a random parameter with a value range of [0, π], b1 and b2 represent phase parameters, represents the golden section number, X prey represents the global optimal position up to the current iteration.

[0085] The traditional straight-line marching strategy is easy to miss potential optimal solutions in complex nonlinear space, and the correlation between offshore wind power and wind speed has a high degree of nonlinearity (such as turbulence effect). The golden sine strategy is used to adjust the search step and direction by introducing the golden section coefficient, so that the electric eel individuals expand the search range in the curved path and enhance the global exploration ability. Specifically, the phase parameters b1 and b2 are calculated by the golden section number τ=0.618. This design makes the population moving path distributed in the unit circle according to the golden ratio, improving the diversity of search direction. T1 is a random parameter with a value range of [0, 2π], T2 is a random parameter with a value range of [0, π], which gives the algorithm random disturbance ability. The global optimal position X prey ​As an attractor guides the movement direction of the population, while maintaining the non-negativity of the movement step through the absolute value operation of the sine function, it avoids invalid oscillation. The electric eel individuals expand the search range in the curved path, which can not only deepen the local optimization, but also jump out of the current area to explore new solution space. For example, in the wind speed mutation scene, the golden sine strategy captures the inflection point of power drop through random phase disturbance, enhancing the adaptability of the model to extreme events. Further improve the optimization effect of input layer weight and hidden layer threshold.

[0086] S4C: Adjust the position of the newly generated individual in each iteration by combining the adaptive random mutation strategy.

[0087] In some embodiments, the position of the newly generated individual in each iteration is adjusted by combining the adaptive random mutation strategy, specifically:

[0088] ;

[0089] In the formula, and indicate the position of the newly generated individual after mutation in the nth dimension and the original position before mutation, m indicates the current iteration number, M indicates the maximum iteration number, rand indicates a random number between 0 and 1, indicates the upper boundary of the solution in each iteration, indicates the lower boundary of the solution in each iteration, and the upper and lower boundaries are set according to the order of magnitude of the parameters to be optimized.

[0090] The offshore wind speed is affected by many factors such as turbulence, tides, and sea waves, and the power data presents severe fluctuations and non-stationarity; the wind direction has a complex nonlinear relationship with the power. The traditional Electric Eel Foraging Optimization (EEFO) algorithm relies on the current optimal solution as a "lighthouse",

[0091] However, the strong nonlinearity and high volatility of offshore wind power data make it difficult for traditional optimization algorithms to adapt to the dynamic changes of such data, resulting in the inability of model parameters to effectively capture complex data patterns. And its strong dependence on the current optimal solution as a guide, it is easy to fall into local optimization in the early iteration, and cannot further optimize the model parameters, thereby limiting the prediction performance. Each iteration forces individuals to converge to the current optimal solution, and the population diversity quickly decays, limiting the global search capability. In a complex multi-peak solution space, the algorithm is prone to suboptimal solutions and is difficult to jump out of the local optimum, such as the power inflection point corresponding to sudden changes in wind speed, which cannot adapt to the characteristics of rapid and unpredictable changes in environmental parameters such as offshore wind speed and wind direction. Therefore, the algorithm needs to have stronger dynamic adaptability, and this guiding mechanism based on static optimal solution may not be able to respond to such changes in time. Based on this, the present application generates some new individuals based on the last eel individual at each iteration, and the generation method is not limited, any one of the conventional new individual generation methods can be used, such as crossover, mutation, etc.

[0092] After generating new individuals, mutation operation is performed, and the newly generated individuals are randomly disturbed to a certain extent to further increase diversity and possibly jump out of the current local optimal region. The disturbance amplitude of Gaussian mutation and Cauchy mutation is fixed and cannot be dynamically adjusted according to the iteration stage or environmental changes. However, in the offshore wind farm scenario, a large range of exploration is required in the early stage (such as sudden power drop caused by typhoon), but the fixed amplitude may not be sufficient; local optimization is required in the later stage (such as parameter tuning under stable wind speed), but the fixed amplitude may be too large, reducing the convergence efficiency. The response to sudden wind speed changes (ΔVt increases) is lagging, and the disturbance cannot match the actual power fluctuation amplitude; Gaussian mutation focuses on local fine search, and Cauchy mutation focuses on global jump, but using a single mutation strategy is difficult to balance exploration and development, such as Gaussian mutation easily falling into local optimum, for example, suboptimal parameter combination under stable wind speed; Cauchy mutation produces too much invalid disturbance in the later stage, prolonging the convergence time.

[0093] The present application uses the above-mentioned adaptive random mutation strategy to adjust the position of the newly generated individual in each iteration, in order for the eel to obtain a better position, the fitness value of the eel population is judged after each mutation, and the optimal position is updated only when the fitness value of the individual is better, otherwise the original position is retained.

[0094] (1) Dynamic disturbance amplitude adjustment is realized

[0095] Iteration stage adaptation: the disturbance amplitude coefficient (1-0.5×m / M) decreases linearly with the increase of iteration number m, realizing "early large-scale exploration and later fine tuning". Early stage enhances global search capability (such as fast response in typhoon scenario), and later stage avoids excessive disturbance (such as efficient convergence under stable wind speed).

[0096] Range-adaptive: the perturbation amplitude is proportional to the variable upper and lower bounds (Ub−Lb), avoiding unreasonable perturbation caused by different characteristic dimension differences.

[0097] (2) Random direction perturbation enhances diversity

[0098] Random direction perturbation between −1 and 1 is generated by (−1+2×rand), breaking the population convergence. In a complex solution space (such as multiple potential optimal parameters corresponding to multiple wind speed intervals), random direction perturbation helps individuals explore different regions, reducing the risk of local optimization.

[0099] (3) Realistic scenario adaptability

[0100] Wind speed mutation response: when the wind speed change rate ΔVt is high, the perturbation amplitude coefficient 0.1×(1−0.5×m / M) can still maintain a large value (in the early iteration), ensuring sufficient perturbation to capture power drops.

[0101] Smooth period efficient convergence: the perturbation amplitude decreases in the later iteration, avoiding invalid perturbation and improving the efficiency of model parameter optimization (such as the RMSE of IEEFO-ELM in Table 1, which is reduced by 41.9% compared to traditional ELM).

[0102] This formula is an adaptive mutation strategy used to mutate the newly generated individuals, which is equivalent to extracting some solutions and updating their positions again. For example, extracting the positions of 10% of the individuals with the lowest fitness (worst positions) in the current iteration and updating them can improve the diversity of the population. This improves the global search ability of the algorithm and maintains the diversity of the population to prevent the eel foraging optimization algorithm from falling into a local optimal solution.

[0103] The improved eel foraging optimization algorithm IEEFO optimizes the input layer weights and hidden layer thresholds of ELM, uses logistic chaotic mapping to generate the initial population, and introduces the golden sine strategy to determine the population movement path. Combined with the adaptive random mutation strategy, the position of the newly generated individual in each iteration is adjusted:

[0104] Logistic chaotic mapping provides high-quality initial solutions, and the golden sine strategy performs global-local balanced search based on this. The adaptive mutation strategy then performs secondary perturbation on the preliminary optimization results, forming a closed-loop optimization process of "extensive coverage → deep development → dynamic correction". This forms a linkage effect between initialization and search, effectively improving the model's generalization performance in high volatility and strong nonlinearity scenarios of offshore wind power.

[0105] Dynamic adaptation to complex environments: the golden sine strategy responds to wind direction changes through random phase, and the adaptive mutation strategy matches the wind speed fluctuation intensity through dynamic amplitude. The combination of the two ensures that the algorithm adjusts its search behavior flexibly in complex sea conditions, providing more accurate power prediction results.

[0106] The chaotic mapping reduces invalid search paths, the golden sine strategy accelerates the positioning of high-quality areas, and the adaptive mutation strategy prevents local convergence. The three work together to significantly shorten the convergence time, balance efficiency and accuracy, and meet the engineering timeliness requirements of real-time prediction of offshore wind farms. The RMSE of IEEFO-ELM in Table 1 is reduced by 41.9% compared to traditional ELM, and the model generalization ability is improved (R²=0.9849).

[0107] In some embodiments, the fitness value in the improved eel foraging optimization algorithm is calculated using the following formula:

[0108] ;

[0109] In the formula, the fitness value is represented by J, the number of training samples, the real value of the wind power in the jth training sample, and the predicted value of the wind power in the jth training sample.

[0110] The fitness value calculation function directly reflects the prediction accuracy of the model, guiding the optimization algorithm to find the parameter combination that minimizes the training error. In an example, in specific calculations, to avoid overfitting, k-fold cross-validation can also be used to calculate the fitness value. By minimizing the RMSE, the optimized extreme learning machine model exhibits stronger generalization ability on unknown data.

[0111] S5: input the preprocessed data into the optimized extreme learning machine model for training, and construct an offshore wind power prediction model, namely an IEEFO-ELM model. The model framework of the offshore wind power prediction model constructed in the present application is as shown in Figure 2 , and the framework diagram for prediction by the model is as shown in Figure 4 .

[0112] In an example, the preprocessed data is divided into training data and test data, and the data allocation ratio is 9:1.

[0113] S6: use the offshore wind power prediction model to predict the real-time offshore wind power of the target port. The specific prediction process can be seen from Figure 3 .

[0114] The wind power data set of November 23, 2011 of Tianjin Port Pacific International Container Terminal is used to predict the wind power by the offshore wind power prediction model constructed in the present application. In terms of parameter settings, the population size of the IEEFO algorithm is set to 50, and the maximum number of iterations is 400 to ensure that the algorithm can fully search for the optimal solution. The number of nodes in the hidden layer of the ELM model is set to 30 to construct a suitable model structure.

[0115] In order to comprehensively evaluate the performance of the IEEFO-ELM model in wind power prediction, the present application compares it with another three prediction models. In order to ensure the stability and reliability of the evaluation results, 30 experiments are carried out for each prediction model. After the experiment is completed, the evaluation results of the four models are calculated and analyzed by using the error index formula. Through a series of experiments and evaluations, the performance of various models in wind power prediction can be more accurately understood, which provides strong support for practical application. The prediction results of the IEEFO-ELM model for wind power are as shown in Figure 5 Figure 5 The prediction value curve obtained by the IEEFO-ELM model in wind power prediction is highly fitted with the actual value curve, which fully proves that the model has excellent performance and accuracy in the field of wind power prediction.

[0116] The comparison chart of the prediction result error of different prediction models is as shown in Figure 6 Figure 6 In the comparison chart, the power fluctuation of the IEEFO-ELM model is represented by a blue line, the fluctuation of the EEFO-ELM model is represented by a green line, the fluctuation of the PSO-ELM model is represented by a pink line, and the fluctuation of the GWO-ELM model is represented by a cyan line. The horizontal coordinate represents 96 samples, each sample represents the wind power and his prediction value of a time period, the wind power is predicted for 24 hours, and the interval is 15 minutes, that is, 24x4=96 sample points. From the chart, it can be observed that the wind power prediction curve fluctuation of the IEEFO-ELM model is almost always below the curves of EEFO-ELM, PSO-ELM and GWO-ELM, and the power fluctuation of the four models is also smaller than that of the traditional ELM prediction model. Compared with PSO-ELM and GWO-ELM models, although the overall prediction effect of EEFO-ELM is good, compared with IEEFO-ELM, there are still several sample points with large fluctuation. In summary, from the perspective of wind power prediction fluctuation, among many prediction models, the IEEFO-ELM model performs better in wind power prediction.

[0117] The present application also compares the prediction results of the above-mentioned several prediction models from the dimensions of mean absolute percentage error (MAPE) and root mean square error (RMSE) and determination coefficient R 2 The comparison results are shown in Table 1.

[0118] Table 1 Comparison of evaluation indexes of PSO-ELM, GWO-ELM, EEFO-ELM and IEEFO-ELM

[0119]

[0120] ​​In order to comprehensively evaluate the performance of the wind power prediction model, multiple error indicators are adopted in the present application, including the mean absolute percentage error, the root mean square error and the determination coefficient, so as to ensure accurate and comprehensive evaluation of the prediction ability of the model.

[0121] Since there is no clear interval range for the two evaluation indicators of RMSE and MAPE when measuring the prediction performance, in order to more comprehensively evaluate the prediction ability of the model, the present application defines the determination coefficient (R 2 ), R 2 ∈[0,1], and the calculation formula is as follows:

[0122] ;

[0123] In the formula, K represents the total number of test samples, represents the actual value of the wind power in the test sample, represents the predicted value of the wind power in the test sample, and k represents the kth test sample.

[0124] According to the data in Table 1, the IEEFO-ELM model proposed in the present chapter exhibits excellent performance in terms of prediction error. The MAPE index of the IEEFO-ELM model is 45.78, which is slightly larger than that of other models. Compared with EEFO-ELM, PSO-ELM and GWO-ELM, the RMSE index of the IEEFO-ELM model reaches 7.163, which is significantly better than that of EEFO-ELM model (10.50), PSO-ELM model (12.06), GWO-ELM model (11.74) and the original ELM model (12.3), which is reduced by 31.8% compared with EEFO-ELM model, 40.6% compared with PSO-ELM model, 38.9% compared with GWO-ELM model, and 41.9% compared with the traditional ELM model. Moreover, the IEEFO-ELM has significantly improved in terms of power fluctuation and R 2 The R 2 index is 0.9849, which greatly improves the fitting degree compared with other wind power prediction models, further confirming the effectiveness of the model in wind power prediction.

[0125] Exemplarily, the IEEFO-ELM model proposed in the present application can also be integrated into the port new energy management and control system. The IEEFO-ELM model proposed in the port new energy management and control system has better stability and higher prediction accuracy in wind power prediction, providing a more reliable method for the wind power prediction module of the port new energy management and control system, and improving the optimization control ability of the system for new energy and the overall intelligent level of the port.

[0126] The application further provides a method for predicting offshore wind power, and the method comprises the following steps: a data module, a model construction module and an output module are used to obtain historical data of offshore wind power; an abnormal value cleaning is performed on the historical data based on a wind speed change rate to obtain preprocessed data; a limit learning machine model is established, and network structure parameters of the limit learning machine model are initialized; input layer weight values and hidden layer threshold values in the network structure parameters are optimized by using an improved electric eel foraging optimization algorithm to generate an optimized limit learning machine model; the preprocessed data is input into the optimized limit learning machine model for training to construct an offshore wind power prediction model; and real-time offshore wind power of a target port is predicted by using the offshore wind power prediction model.

[0127] The data module is configured to obtain historical data of offshore wind power; perform abnormal value cleaning on the historical data based on a wind speed change rate to obtain preprocessed data; and dynamically adjust an abnormality detection threshold.

[0128] The model construction module is connected to the data module and is configured to establish a limit learning machine model, initialize network structure parameters of the limit learning machine model, optimize input layer weight values and hidden layer threshold values in the network structure parameters by using an improved electric eel foraging optimization algorithm to generate an optimized limit learning machine model, and input the preprocessed data into the optimized limit learning machine model for training to construct an offshore wind power prediction model.

[0129] The output module is connected to the model construction module and is configured to predict real-time offshore wind power of a target port by using the offshore wind power prediction model.

[0130] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the application, and not to limit them; although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments.

Claims

1. A method for predicting offshore wind power, characterized in that: The steps include: S1: Obtain historical data of offshore wind power; S2: Dynamically adjust the anomaly detection threshold based on the wind speed change rate, clean the historical data of outliers, and obtain preprocessed data; The calculation method for dynamically adjusting the abnormality detection threshold based on the wind speed change rate is as follows: ； In the formula, Uppert represents the upper limit of outlier judgment, Lowert represents the lower limit of outlier judgment, Q 75 represents the upper quartile, Q 25 Indicates the lower quartile, IQR t represents the interquartile range, represents the wind speed change rate, and λ represents the turbulence adjustment coefficient; S3: Establish an extreme learning machine model and initialize the network structure parameters of the extreme learning machine model; S4: optimizing the input layer weights and hidden layer thresholds in the network structure parameters using an improved electric eel foraging optimization algorithm to generate an optimized extreme learning machine model; S5: inputting the preprocessed data into the optimized extreme learning machine model for training to construct an offshore wind power prediction model; S6: Using the offshore wind power prediction model, predict the real-time offshore wind power of the target port.

2. The offshore wind power prediction method according to claim 1, characterized in that: The wind speed change rate is calculated using the following formula: ； Where V t Indicates the wind speed value at time t, V t-1 Indicates the wind speed value at time t-1, V rated Indicates the rated wind speed value.

3. The offshore wind power prediction method according to claim 1, characterized in that: The improved electric eel foraging optimization algorithm is improved by the following strategies: Use logistic chaotic mapping to generate the initial population; Introduce the golden sine strategy to determine the population movement path; Combined with the adaptive random mutation strategy, the position of the newly generated individuals in each iteration is adjusted.

4. The offshore wind power prediction method according to claim 3, characterized in that: The logistic chaotic map is used to generate the initial population, specifically: Randomly generate a vector, where the dimension of the vector is Dim, and the value of the vector i in each dimension is a random number between 0 and 1; Use the logistic chaotic mapping formula to generate a sequence of vectors N [s1,…,s i ,…,s N ], the logistic chaotic mapping formula is as follows: ； Where s i represents the i-th vector, s i-1 represents the i-1th vector, and f represents the control parameter of the chaotic mapping; According to the sequence, the initial population is obtained by performing distribution using the following formula: ； Where x i represents the position of the i-th electric eel corresponding to the i-th vector, Ub represents the upper limit of the initial population, and Lb represents the lower limit of the initial population.

5. The offshore wind power prediction method according to claim 3, characterized in that: The introduction of the golden sine strategy to determine the population movement path is as follows: when the electric eel moves from the current position to the target position, the population movement path is determined by the golden sine strategy to complete the update of the population position. The expression of the golden sine strategy is: ； Where x i (t) represents the position of the i-th electric eel at time t, x i (t+1) represents the position of the i-th electric eel at time t+1, T1 is a random parameter with a value range of [0,2π], T2 is a random parameter with a value range of [0,π], b1 and b2 represent phase parameters, represents the golden ratio, X prey Represents the global optimal position up to the current iteration.

6. The offshore wind power prediction method according to claim 3, characterized in that: The adaptive random mutation strategy is combined to adjust the position of the newly generated individuals in each iteration, specifically: ； Where, and Indicates the position of the newly generated individual after the n-th dimension position mutation and the original position before the mutation, m represents the current number of iterations, M represents the maximum number of iterations, and rand represents a random number between 0 and 1. represents the upper bound of the solution in each iteration, represents the lower bound of the solution at each iteration.

7. The offshore wind power prediction method according to claim 1, characterized in that: The fitness value in the improved electric eel foraging optimization algorithm is calculated using the following formula: ； Where, represents the fitness value, J represents the number of training samples, represents the true value of wind power in the jth training sample, Represents the predicted value of wind power in the jth training sample.

8. The offshore wind power prediction method according to claim 1, characterized in that: The historical data includes wind speed, wind direction and offshore power generation.

9. An offshore wind power prediction system, characterized in that: An offshore wind power prediction method according to any one of claims 1 to 8 is implemented, comprising: a data module, a model building module and an output module; The data module is used to obtain historical data of offshore wind power; dynamically adjust the abnormality detection threshold based on the wind speed change rate, clean the historical data of abnormal values, and obtain pre-processed data; The model construction module is connected to the data module and is used to establish an extreme learning machine model and initialize the network structure parameters of the extreme learning machine model; use an improved electric eel foraging optimization algorithm to optimize the input layer weights and hidden layer thresholds in the network structure parameters to generate an optimized extreme learning machine model; input the preprocessed data into the optimized extreme learning machine model for training to construct an offshore wind power prediction model; The output module is connected to the model building module and is used to predict the real-time offshore wind power of the target port by using the offshore wind power prediction model.

Images