Wind power plant energy management method and system based on edge calculation

Abstract

The invention discloses a wind power plant energy management method and system based on edge calculation, and relates to the technical field of wind power, and the method comprises the steps: collecting the data of a wind turbine generator in real time, and carrying out the preprocessing of the collected data; calculating a power generation power prediction value based on the preprocessed data; calculating a power grid load demand predicted value based on the generated power predicted value; calculating a power grid load matching deviation based on the power grid load demand predicted value; and generating a wind turbine generator power output adjustment scheme based on the power grid load matching deviation. According to the wind power plant energy management method based on edge computing, the real-time performance of data processing is improved through a distributed architecture of edge computing nodes, and especially under the conditions that the wind power plant operation environment is complex and the data size is large, the dynamic change load requirement can be quickly responded; based on the prediction technology of hierarchical modeling, not only is the precision of wind power prediction and power grid load demand prediction improved, but also the adaptability of the model to complex meteorological conditions and non-linear relationships is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of wind power technology, and specifically to a wind farm energy management method and system based on edge computing. Background Art

[0002] At present, with the widespread application of wind power generation technology, the proportion of wind farms in the power system has gradually increased. However, due to the significant intermittent and volatile characteristics of wind power generation, its grid-connected operation poses a severe challenge to the stability and reliability of the power grid. The energy management system of traditional wind farms mainly relies on a centralized cloud processing mode, which usually includes data collection, processing and optimization processes. Due to the large-scale data transmission and central computing involved, the system's real-time and dynamic response capabilities are insufficient, making it difficult to meet the rapidly changing power grid needs.

[0003] Existing wind power forecasting methods are usually based on static mathematical models or simple time series analysis. These methods are difficult to fully consider the dynamic changes in meteorological conditions and the complexity of the wind farm operating environment. Especially when wind speed and wind direction change frequently, the error of the forecast results increases significantly. In addition, traditional load demand forecasting methods fail to fully consider the actual impact of wind power output on grid load demand, resulting in insufficient forecast accuracy and difficulty in providing reliable support for optimization decisions.

[0004] In terms of power adjustment and scheduling optimization, existing methods mostly focus on simplified linear optimization models, ignoring the response speed limit and power output constraints of wind turbines. The generated scheduling schemes may not be implemented in actual applications because they do not meet the physical limitations of the equipment. At the same time, traditional optimization methods do not take into account the dynamic interaction between wind farms and power grids, and fail to provide sufficient adaptability when load fluctuations and wind power fluctuations are large.

[0005] It can be seen that traditional wind farm energy management technology has the following shortcomings:

[0006] The efficiency of data collection and processing is low, making it difficult to quickly respond to dynamic changes in load demand.

[0007] Existing power forecasting and load demand forecasting models perform poorly under complex meteorological conditions.

[0008] The operating limitations of wind turbines and the characteristics of grid demand are not fully considered, and the optimization results are not applicable in actual operation.

[0009] In summary, in response to the above technical problems, it is necessary to propose an innovative wind farm energy management method to meet the needs of efficient operation of modern wind farms and grid stability by improving data processing efficiency, improving prediction accuracy and the applicability of optimization algorithms. Summary of the invention

[0010] In view of the above-mentioned problems, the present invention is proposed.

[0011] Therefore, the technical problem solved by the present invention is: the existing wind farm energy management method has the problems of insufficient real-time performance, low prediction accuracy and poor feasibility of optimization schemes, especially in the scenario of rapid changes in dynamic load demand, it is difficult to effectively deal with the matching problem of wind power fluctuations and grid load demand, as well as how to optimize the power output adjustment of wind turbines in a complex operating environment.

[0012] In order to solve the above technical problems, the present invention provides the following technical solutions: a wind farm energy management method based on edge computing, comprising:

[0013] Collect wind turbine data in real time and pre-process the collected data;

[0014] Calculate the power generation prediction value based on the preprocessed data;

[0015] Calculate the power grid load demand forecast value based on the power generation forecast value;

[0016] Calculating the grid load matching deviation based on the grid load demand forecast value;

[0017] Generate a wind turbine power output adjustment plan based on the grid load matching deviation.

[0018] As a preferred solution of the wind farm energy management method based on edge computing described in the present invention, wherein: the wind turbine data includes power generation data, grid load data and meteorological data;

[0019] The preprocessing of the collected data includes normalizing the power generation data, which is expressed as:

[0020]

[0021] Among them, P raw Represents the original wind turbine power generation data, P min Represents the minimum value in the data set, P max Represents the maximum value in the data set, P norm Represents the normalized wind turbine power generation data;

[0022] The fluctuation amplitude of the power grid load data is calculated and expressed as:

[0023]

[0024] Among them, P grid,i represents the i-th data point of the power grid load data, represents the mean value of the grid load, N represents the total number of grid load data, ΔPgrid Indicates the fluctuation amplitude of the power grid load;

[0025] The meteorological data is decomposed into wind speed and wind direction to generate vectorized meteorological characteristic data, which is expressed as:

[0026]

[0027] Among them, v represents wind speed, θ represents wind direction, A two-dimensional vector representing wind speed and direction.

[0028] As a preferred solution of the wind farm energy management method based on edge computing described in the present invention, the power generation prediction value is expressed as:

[0029] P base =α 1 ·v 3 +α 2 ·v 2 +α 3 v+α 4

[0030] P weather =P base +β 1 ·v·cos(θ)+β 2 ·v·sin(θ)

[0031] P wind,pred =P weather +β 3 ·P norm

[0032] Among them, P base represents the estimated value of generated power, α 1 ,α 2 ,α 3 ,α 4 represents the wind turbine power curve fitting coefficient, β 1 ,β 2 Represents the correction factor of wind speed component, P wind,pred represents the final predicted power generation, β 3 Represents the correction factor of historical power generation;

[0033] The calculation coefficient of the power generation prediction value is optimized by minimizing the mean square error method, which is expressed as:

[0034]

[0035] The optimized model output P wind,pred is the predicted value of generated power.

[0036] As a preferred solution of the wind farm energy management method based on edge computing described in the present invention, the grid load demand forecast value is expressed as:

[0037] P grid,base =γ 1 ΔP grid +γ 2 ·P grid,hist

[0038] P grid,corrected =P grid,base +γ 3 ·P wind,pred

[0039] P grid,pred =P grid,corrected +γ 4 ·(ΔP grid ) 2 +γ 5 ·(ΔP grid ·P wind,pred )

[0040] Among them, γ 1 ,γ 2 Represents the regression coefficient of fluctuation amplitude and historical load, P grid,hist Representation History 2 Grid load data, γ 3 The correction factor that represents the impact of wind power on grid load demand, (ΔP grid ) 2 The square term of the load fluctuation amplitude is used to capture the nonlinear effect of load fluctuation, ΔP grid ·P wind,pred represents the interaction term between load fluctuation and wind power, γ 4 ,γ 5 represents the regression coefficient of nonlinear supplementation;

[0041] The calculation coefficient of the power grid load demand forecast value is optimized by minimizing the mean square error method, which is expressed as:

[0042]

[0043] The optimized P grid,pred is the predicted value of grid load demand.

[0044] As a preferred solution of the wind farm energy management method based on edge computing described in the present invention, the grid load matching deviation is expressed as:

[0045] ΔP grid =P grid,actual -P grid,pred

[0046] Among them, P grid,actual Represented as actual grid load;

[0047] Generating a wind turbine power output adjustment plan includes defining an optimization objective function and setting constraint conditions;

[0048] The particle swarm optimization algorithm is used to iteratively solve the objective function, generate adjustment plans, and send the optimization results to the wind turbines.

[0049] As a preferred solution of the wind farm energy management method based on edge computing described in the present invention, the objective function is expressed as:

[0050]

[0051] Among them, J represents the optimization objective function, λ 1 ,λ 2 Represents the weight coefficient, P wind,adjusted,i represents the adjusted power output of the i-th wind turbine, P wind,current,i Indicates the actual output power of the i-th wind turbine.

[0052] As a preferred solution of the wind farm energy management method based on edge computing described in the present invention, wherein: the constraint conditions include power output limitation, response speed limitation and total power constraint;

[0053] The constraint condition is expressed as,

[0054] P wind,min,i ≤P wind,adjusted,i ≤P wind,,max,i

[0055] |P wind，adjusted，i -P wind，current，i |≤R maxi ·T

[0056]

[0057] Among them, P wind,min,i、 P wind,max,i represents the minimum and maximum power limits of the i-th wind turbine, R max,i It represents the maximum adjustment rate of the i-th wind turbine, and T represents the adjustment time period.

[0058] Another object of the present invention is to provide a wind farm energy management system based on edge computing, which can solve the problems of low data processing efficiency, insufficient power prediction accuracy, and lack of constraint adaptability of optimization algorithms in existing wind farms by constructing a wind farm energy management system based on edge computing, thereby effectively improving the real-time, accuracy and applicability of wind farm energy management.

[0059] To solve the above technical problems, the present invention provides the following technical solutions: a wind farm energy management system based on edge computing, comprising: a data acquisition module, an energy prediction module, a status monitoring module, an energy efficiency evaluation module and a power adjustment module; the data acquisition module is used to collect wind turbine data in real time and preprocess the collected data; the energy prediction module is used to calculate the power generation prediction value based on the preprocessed data; the status monitoring module is used to calculate the grid load demand prediction value based on the power generation prediction value; the energy efficiency evaluation module is used to calculate the grid load matching deviation based on the grid load demand prediction value; the power adjustment module is used to generate a wind turbine power output adjustment plan based on the grid load matching deviation.

[0060] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the wind farm energy management method based on edge computing as described above are implemented.

[0061] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the wind farm energy management method based on edge computing as described above.

[0062] Beneficial effects of the present invention: The wind farm energy management method based on edge computing provided by the present invention significantly improves the real-time performance of data processing through the distributed architecture of edge computing nodes, especially when the wind farm operating environment is complex and the amount of data is huge, and can quickly respond to dynamically changing load demands; the prediction technology based on hierarchical modeling not only improves the accuracy of wind power prediction and grid load demand prediction, but also enhances the model's adaptability to complex meteorological conditions and nonlinear relationships; the optimization scheme achieves the best balance between load demand and wind turbine operating cost through a dynamic weight adjustment mechanism and a balanced objective function, while combining constraints to ensure the feasibility of the optimization results. Overall, the present invention provides efficient and flexible technical support for the intelligent operation of wind farms and the stability of power grids, and is particularly suitable for scenarios with severe load fluctuations or complex wind farm operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0064] Figure 1 An overall flow chart of a wind farm energy management method based on edge computing provided for one embodiment of the present invention.

[0065] Figure 2 An overall structural diagram of a wind farm energy management system based on edge computing provided for the second embodiment of the present invention. DETAILED DESCRIPTION

[0066] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0067] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0068] Example 1

[0069] Reference Figure 1 , as an embodiment of the present invention, provides a wind farm energy management method based on edge computing, comprising:

[0070] Collect wind turbine data in real time and pre-process the collected data;

[0071] Calculate the power generation prediction value based on the preprocessed data;

[0072] Calculate the power grid load demand forecast value based on the power generation forecast value;

[0073] Calculating the grid load matching deviation based on the grid load demand forecast value;

[0074] Generate a wind turbine power output adjustment plan based on the grid load matching deviation.

[0075] The wind turbine data includes power generation data, grid load data and meteorological data;

[0076] The preprocessing of the collected data includes normalizing the power generation data, which is expressed as:

[0077]

[0078] Among them, P raw Represents the original wind turbine power generation data, P min Represents the minimum value in the data set, P max Represents the maximum value in the data set, P normRepresents the normalized wind turbine power generation data;

[0079] The fluctuation amplitude of the power grid load data is calculated and expressed as:

[0080]

[0081] Among them, P grid,i represents the i-th data point of the power grid load data, represents the mean value of the grid load, N represents the total number of grid load data, ΔP grid Indicates the fluctuation amplitude of the power grid load;

[0082] The meteorological data is decomposed into wind speed and wind direction to generate vectorized meteorological characteristic data, which is expressed as:

[0083]

[0084] Where v is the wind speed in m / s, θ is the wind direction in degrees (measured clockwise from due north), A two-dimensional vector representing wind speed and direction, containing the wind speed components in the east direction (x-axis) and north direction (y-axis).

[0085] It should be noted that the preprocessing of the collected data is not only the formatting of the original data, but also improves the quality of the subsequent model input data and the applicability of the model by extracting and standardizing the features of different types of data. The normalization processing method for power generation data eliminates the possible impact of inconsistent data distribution between different wind turbines by standardizing the data range, providing a unified input for subsequent model training and prediction.

[0086] Furthermore, the fluctuation amplitude calculation of the power grid load data can dynamically capture the characteristics of the power grid load changes by calculating the mean square error of the fluctuation data. Especially in the scenario of high-frequency load fluctuations, this calculation method can provide a clear fluctuation amplitude representation and enhance the adaptability of the load forecasting model to sudden demands.

[0087] Furthermore, vectorized meteorological feature data reduces the nonlinear relationship of the original data by decomposing wind speed and wind direction data into two-dimensional components, making the feature dimension of the model input more suitable for wind power prediction scenarios. This processing method can significantly reduce the training complexity of the model and show higher prediction accuracy in scenarios with rapid changes in wind speed or multiple wind directions.

[0088] Specifically, the data preprocessing process specifically targets the sampling frequency and noise characteristics of different data sources, and performs real-time denoising and interpolation processing through edge computing nodes to ensure the real-time and continuity of input data. This approach is particularly suitable for distributed data collection needs of remote wind farms.

[0089] The predicted value of generated power is expressed as:

[0090] P base =α 1 ·v 3 +α 2 ·v 2 +α 3 ·v+α 4

[0091] P weather =P base +β 1 ·v·cos(θ)+β 2 ·v·sin(θ)

[0092] P wind,pred =P weather +β 3 ·P norm

[0093] Among them, P base represents the estimated value of generated power, α 1 ,α 2 ,α 3 ,α 4 represents the wind turbine power curve fitting coefficient, β 1 ,β 2 Represents the correction factor of wind speed component, P wind,pred represents the final predicted power generation, β 3 Represents the correction factor of historical power generation;

[0094] The calculation coefficient of the power generation prediction value is optimized by minimizing the mean square error method, which is expressed as:

[0095]

[0096] The optimized model output P wind,pred is the predicted value of generated power.

[0097] It should be noted that the power generation prediction model takes the wind turbine power curve as its basis and captures the nonlinear relationship between wind speed and power generation through physical modeling, thereby achieving a direct response of wind speed changes to power output. The fitting parameters of the model are trained in combination with historical wind turbine operation data, so that the model can maintain high accuracy when adapting to various wind turbine characteristics.

[0098] Furthermore, the meteorological characteristic correction part of the model dynamically adjusts the basic power generation estimate by introducing wind speed and wind direction components, which can more accurately reflect the power output change trend under complex meteorological conditions (such as frequent changes in wind direction or sudden increases in wind speed). This correction method solves the problem of insufficient sensitivity to wind direction changes in traditional power prediction methods by strengthening the model's adaptability to the dynamic characteristics of meteorological data.

[0099] Furthermore, combined with the correction part of the historical power generation data, by taking the normalized historical power generation power as the input factor, the model can compensate for the uncontrollable errors in the meteorological data (such as sensor drift or sampling noise), thereby improving the stability and robustness of the overall prediction results.

[0100] Specifically, by minimizing the mean square error to optimize the model coefficient training method, an iterative optimization strategy is used to dynamically adjust each coefficient to ensure that the fitting error gradually decreases with the increase in the amount of training data. This training method is particularly suitable for the scenario of coordinated power generation of multiple wind turbines and can optimize the output differences between different units.

[0101] The grid load demand forecast value is expressed as:

[0102] P grid,base =γ 1 ΔP grid +γ 2 ·P grid,hist

[0103] P grid,corrected =P grid,base +γ 3 ·P wind,pred

[0104] P grid,pred =P grid,corrected +γ 4 ·(ΔP grid ) 2 +γ 5 ·(ΔP grid ·P wind,pred )

[0105] Among them, γ 1 ,γ 2 Represents the regression coefficient of fluctuation amplitude and historical load, P grid,hist Representation History 2 Grid load data, γ 3 The correction factor that represents the impact of wind power on grid load demand, (ΔP grid ) 2 The square term of the load fluctuation amplitude is used to capture the nonlinear effect of load fluctuation, ΔP grid ·P wind,predrepresents the interaction term between load fluctuation and wind power, γ 4 ,γ 5 represents the regression coefficient of nonlinear supplementation;

[0106] The calculation coefficient of the power grid load demand forecast value is optimized by minimizing the mean square error method, which is expressed as:

[0107]

[0108] The optimized P grid,pred is the predicted value of grid load demand.

[0109] It should be noted that the grid load demand forecasting model is divided into three parts: basic calculation, correction calculation and nonlinear supplement. The basic calculation part captures the main change trend of grid demand through regression analysis of load fluctuation amplitude and historical load data, providing the model with a high-correlation basic forecast value.

[0110] Furthermore, the correction calculation part combines the dynamic impact of wind power prediction value on grid load demand and effectively improves the model's sensitivity to actual wind power changes by introducing wind power prediction value into the load demand calculation formula. Especially in grids with a high proportion of wind power access, this correction term can significantly improve the match between the prediction results and the actual values.

[0111] Furthermore, the nonlinear supplementary part can dynamically capture the complex nonlinear relationship in the grid load demand by introducing square terms and cross terms. These nonlinear characteristics not only come from the load fluctuation amplitude itself, but also interact with the fluctuation characteristics of wind power output. The model's supplementary terms can accurately fit these complex relationships.

[0112] Specifically, the training method of load prediction model parameters is optimized through mean square error, and a multi-level stepwise optimization strategy is adopted. After the initial regression calculation, the weights of the supplementary items are further adjusted through cross-validation to ensure the robustness of the prediction model under highly fluctuating load scenarios.

[0113] The grid load matching deviation is expressed as:

[0114] ΔP grid =P grid,actual -P grid,pred

[0115] Among them, P grid,actual Represented as actual grid load;

[0116] Generating a wind turbine power output adjustment plan includes defining an optimization objective function and setting constraint conditions;

[0117] The particle swarm optimization algorithm is used to iteratively solve the objective function, generate adjustment plans, and send the optimization results to the wind turbines.

[0118] It should be noted that the calculation of grid load matching deviation directly reflects the difference between the actual grid load and the predicted load, which provides a clear optimization direction and quantitative standard for the subsequent optimization objective function.

[0119] Furthermore, the matching deviation calculation results can dynamically update the parameters of the power grid load demand prediction model in real-time feedback, so that the load matching deviation continues to converge to an optimal range, thereby achieving continuous optimization of the model. This dynamic update method is particularly suitable for scenarios where load demand changes rapidly.

[0120] Furthermore, in the process of generating the power output adjustment plan for the wind turbine, by introducing the grid load matching deviation into the objective function, it is ensured that the optimization plan takes minimizing the load deviation as the core goal, while taking into account the operating limitations of the wind turbine, thereby improving the feasibility of the adjustment plan.

[0121] Specifically, the calculation results of the matching deviation can also be used to evaluate the overall operating efficiency of the wind farm. By comparing with the historical matching deviation data, it can provide key indicators of the operating status of the wind farm and provide guidance for subsequent scheduling optimization.

[0122] The objective function is expressed as,

[0123]

[0124] Among them, J represents the optimization objective function, λ 1 ,λ 2 Represents the weight coefficient, which is used to balance load matching and adjust the cost; P wind,adjusted,i represents the adjusted power output of the i-th wind turbine, P wind,current,i Indicates the actual output power of the i-th wind turbine.

[0125] It should be noted that the optimization objective function dynamically adjusts the execution direction of the optimization scheme by balancing the weight relationship between the grid load matching deviation and the wind turbine adjustment cost, thereby achieving the optimal decision under different operating scenarios.

[0126] Furthermore, the matching deviation term in the objective function is expressed by the weight coefficient λ 1 Dynamically adjust the priority of minimizing deviations, which is suitable for scenarios where the grid load fluctuates drastically. For example, when the grid load demand increases significantly, this item can quickly drive the wind turbine to adjust the output power to meet the demand.

[0127] Furthermore, the design of the adjustment cost term reflects the energy consumption and operation cost of the adjustment scheme by introducing the square difference between the actual output of the wind turbine and the adjusted power. 2 , a balance can be found between stabilizing grid operation and reducing dispatch costs.

[0128] Specifically, the overall design of the objective function improves the economy and flexibility of wind turbines through multi-objective optimization while ensuring that the grid needs are met, providing support for the coordinated operation of the grid and wind farms. The constraints include power output limit, response speed limit and total power limit;

[0129] The constraints include:

[0130] Power output limit:

[0131] P wind,min,i ≤P wind,adjusted,i ≤P wind,max,i

[0132] Among them, P wind,min,i、 P wind,max,i represents the minimum and maximum power limits of the i-th wind turbine.

[0133] Total power constraint:

[0134]

[0135] Ensure that the adjusted total power of the wind turbines meets the grid load requirements.

[0136] Response speed limit:

[0137] |P wind,adjusted,i -P wind,current,i |≤R max,i ·T

[0138] Among them, R max,i It represents the maximum adjustment rate of the i-th wind turbine, and T represents the adjustment time period.

[0139] It should be noted that the design of the constraint conditions ensures the feasibility and safety of the adjustment scheme in technical implementation by combining power output limitation, response speed limitation and total power constraint, while meeting the actual operation requirements of the wind farm.

[0140] Furthermore, power output limitation protects the physical equipment safety of wind turbines by constraining the adjustment range of each wind turbine to avoid exceeding the equipment operating capacity, especially under extreme conditions of high or low wind speeds.

[0141] Furthermore, the response speed limit combines the maximum response rate of the device and the adjustment time period to ensure that the adjustment scheme can respond promptly to rapidly changing load demands while preventing power fluctuations caused by overly fast adjustments.

[0142] Specifically, the total power constraint ensures the overall stability of the grid operation by dynamically balancing the relationship between the total power output of wind turbines and grid demand, while also improving the power output efficiency of the wind farm.

[0143] It should also be noted that the particle swarm algorithm uses particle encoding, and each particle represents a wind turbine power adjustment scheme, which is expressed as,

[0144] P wind,adjusted =[P wind,adjusted,1 ,P wind,adjusted,2 ,…,P wind,adjusted,N ]

[0145] The dimension of each particle is the number N of wind turbines.

[0146] Particle initialization: Randomly initialize the position of each particle (power output scheme), expressed as,

[0147] P wind,adjusted,i ∈[P wind,min,i ,P wind,max,i ]

[0148] Initialize the velocity of each particle, expressed as,

[0149] v i,t = rand·(P wind,max,i -P wind,min,i )

[0150] Make sure the initial particles meet the power constraints and the total power constraints.

[0151] Speed ​​update (combined with constraints),When updating the particle's speed, it is corrected in combination with the total power constraint, expressed as,

[0152] v i,t+1 =w·v i,t +c 1 ·rand 1 ·(p best,i -x i,t )+c 2 ·rand 2 ·(g best -x i,t )

[0153] The updated speed is limited, expressed as,

[0154] v i,t+1=min(max(v i,t+1 ,-R max,i ·T),R max,i ·T)

[0155] Position update (combined with power constraints), updates the position of the particle (power output scheme), expressed as,

[0156] x i,t+1 =x i,t +v i,t+1

[0157] The updated position is clipped to ensure that the power output limit is met, expressed as,

[0158] P wind,min,i ≤x i,t+1 ≤P wind,max,i

[0159] Fitness function calculation, calculate the fitness of the current particle, expressed as,

[0160]

[0161] For particles that do not meet the total power constraint, an additional penalty term is added, expressed as,

[0162]

[0163] Update the best position and update the historical best position p of each particle best,i ; Update the global best position g best .

[0164] Termination condition: when the objective function J converges or reaches the maximum number of iterations, the iteration is stopped and the global optimal position g is output. best .

[0165] Optimal solution g best That is the wind turbine power output adjustment scheme, expressed as,

[0166] P wind,adjusted =[P wind,adjusted,1 ,P wind,adjusted,2 ,…,P wind,adjusted,N ]

[0167] Adjustment instructions are sent to each wind turbine through edge computing nodes to implement the optimization plan.

[0168] Monitor the adjusted output power P wind,actual and load data P grid,actual , and repeat the optimization based on the new deviation.

[0169] Specifically, the application of the particle swarm optimization algorithm in the present invention is customized to meet the needs of wind turbine power adjustment. First, in the particle initialization stage, a position is assigned to each particle to represent the power adjustment scheme of the wind turbine. During initialization, the position range of the particle is determined by the minimum and maximum power output range of each wind turbine, and the matching conditions of the total power output of the wind farm and the grid load demand must be met. If the initial position of the particle does not meet the total power constraint, the position ratio of the particle is adjusted by normalization to meet the overall power balance. Next, in the speed update process, the speed of the particle is not only based on the gradient change of the objective function, but also dynamically adjusted in combination with the current grid load deviation. If the load deviation is large, the particle is preferentially guided to move in the direction of the larger power adjustment gap to accelerate the optimization process. In addition, during the iteration process, the historical optimal position and the global optimal position of the particle are regularly updated, and the dynamic convergence strategy is used to determine whether the optimization termination condition is reached to ensure efficient convergence of the optimization process.

[0170] In the process of power generation prediction and grid load demand prediction, the present invention adopts dynamic parameter optimization and weight adjustment mechanism to ensure the adaptability and prediction accuracy of the model. For the power generation prediction model, its parameters are continuously updated through training of historical data, so that the model can adapt to the characteristics and meteorological conditions of different wind turbines. The training process adopts a step-by-step optimization strategy, and each iteration adjusts the weight of the parameter according to the prediction error, thereby continuously narrowing the gap between the actual power generation and the predicted value. In the grid load demand prediction, the weight coefficient of the nonlinear supplementary part is dynamically adjusted according to the real-time characteristics of the load fluctuation. For example, when the load fluctuation is significant, the weight of the coefficient related to the fluctuation is increased, so that the model can more accurately capture the impact of the fluctuation on the demand. At the same time, the weight of the load correction part will be adaptively updated as the wind power output changes, ensuring that the prediction results reflect the actual demand changes of the power grid in real time.

[0171] The present invention relies on the distributed architecture of edge computing nodes to achieve efficient processing and real-time optimization of wind farm data. The edge node uses multi-threading technology to process different types of data in parallel, such as power generation data, grid load data and meteorological data. Each type of data is collected, preprocessed and stored by an independent thread to improve the response speed of the system. In addition, the edge node can maintain the basic optimization function of the wind farm in the event of network disconnection or communication delay through local storage and caching technology. For large wind farms, multiple nodes operate collaboratively through a distributed computing framework, each node independently performs acquisition tasks, and the master node is responsible for integrating the optimization results and sending them to the wind turbines in a unified manner. In addition, the edge node also uses a real-time feedback mechanism to compare the actual operating status with the model prediction value, and dynamically adjust the model parameters and optimization targets to cope with complex meteorological conditions and changes in load demand. This architecture effectively improves the stability and real-time response capability of wind farm operation.

[0172] The constraints set in the present invention not only limit the search space of the optimization algorithm, but also ensure that the generated power adjustment scheme can be implemented in actual operation. The power output limit defines the adjustment range according to the operating capacity of the wind turbine to prevent the output power from exceeding the physical limit of the equipment. The response speed limit is based on the maximum adjustment rate of the wind turbine to ensure that the adjustment process will not cause mechanical damage to the equipment while meeting the time requirements of the change in grid demand. The total power constraint dynamically balances the total output of the wind farm with the actual demand of the grid to avoid insufficient or overloaded overall power output of the wind farm. In addition, these constraints take effect in real time in each iteration of the particle swarm optimization algorithm. Specifically, after the particle position and velocity are updated, the algorithm will detect whether any particle violates the constraint. If a constraint violation occurs, the algorithm will adjust the position and velocity of the particle by cutting or introducing a penalty mechanism to force it back to the feasible solution space. This dynamic constraint mechanism not only improves the practicality of the optimization scheme, but also significantly reduces the generation of infeasible solutions.

[0173] Example 2

[0174] Reference Figure 2 , as an embodiment of the present invention, provides a wind farm energy management system based on edge computing, including:

[0175] Data acquisition module 100, energy prediction module 200, status monitoring module 300, energy efficiency evaluation module 400 and power adjustment module 500;

[0176] The data acquisition module 100 is used to collect wind turbine data in real time and pre-process the collected data;

[0177] The energy prediction module 200 is used to calculate the power generation prediction value based on the preprocessed data;

[0178] The state monitoring module 300 is used to calculate the power grid load demand prediction value based on the power generation power prediction value;

[0179] The energy efficiency evaluation module 400 is used to calculate the grid load matching deviation based on the grid load demand prediction value;

[0180] The power adjustment module 500 is used to generate a wind turbine power output adjustment plan based on the grid load matching deviation.

[0181] Example 3

[0182] An embodiment of the present invention is different from the first two embodiments in that:

[0183] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0184] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0185] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0186] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0187] Example 4

[0188] As an embodiment of the present invention, a wind farm energy management method based on edge computing is provided. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0189] This experiment is based on the actual operating environment of the wind farm and simulates the performance of different energy management methods in a typical wind farm. The experimental scenario is a wind farm with 24 wind turbines, with complex terrain and changeable meteorological conditions, and the data collection frequency is once per second. The collected data includes the real-time power generation of wind turbines, grid load demand and its fluctuation characteristics, and meteorological data (wind speed and wind direction). The experimental cycle is 60 minutes, mainly targeting scenarios with large wind speed fluctuations and frequent changes in grid load demand to verify the real-time and dynamic response capabilities of the method of the present invention. The data acquisition devices are all actually deployed sensors. The data of each wind turbine is collected and stored in the local storage system through the edge computing node, and the grid load and meteorological data are used as global references. The traditional method compared in the experiment is a centralized cloud management mode, with a centralized server as the core to process data and generate scheduling plans.

[0190] The experimental process of the method of the present invention includes that in the wind farm, the edge computing node of each wind turbine is responsible for real-time collection of local power generation, grid load data and meteorological data. After the collection is completed, the edge node immediately normalizes the power generation data, removes the data dimension difference, and calculates the fluctuation amplitude of the grid load data to extract the load fluctuation characteristics. At the same time, the wind speed and wind direction data are vectorized into two-dimensional meteorological features to reduce the impact of nonlinear input on the prediction model. After the data preprocessing is completed, the power generation prediction and load demand prediction are completed through the local computing resources of the edge node. The prediction model uses historical power generation data, vectorized meteorological features and load fluctuation characteristics for real-time iterative optimization, and outputs the power generation prediction value and the load demand prediction value. Based on the prediction results and the real-time grid load deviation, the edge node generates a power adjustment plan through a customized particle swarm optimization algorithm, and considers the power output limit, response rate and total power constraint. The final optimization plan is directly sent by the edge node to the wind turbine for execution to adjust the power output of the wind turbine. The entire process is completed in the edge node without transmitting the data to the cloud, thereby ensuring the real-time processing and the rapid response capability of the adjustment plan.

[0191] The traditional method adopts a centralized management mode. The real-time power generation data collected by each wind turbine is transmitted to the cloud server through the network. The power grid load demand data and meteorological data are also uniformly transmitted to the cloud for centralized processing. After all data are normalized and basic features are extracted in the cloud, they enter the static power generation prediction model for prediction. Power prediction is based on a simple regression analysis of historical power generation data and wind speed data, and the predicted value is used as the input for load demand prediction. Load demand prediction uses a linear regression model to analyze the historical change characteristics of the grid load, ignoring the dynamic impact of real-time wind power changes on load demand. After completing the prediction, the cloud server uses a linear optimization algorithm to generate a power adjustment plan, but because the algorithm does not fully consider the physical operation limitations of the wind turbine, the generated plan needs to be manually reviewed and manually adjusted to meet the actual operation requirements. The final adjustment plan is sent from the cloud to the wind turbine for execution. The centralization of data transmission and cloud computing leads to long processing and response times, especially in scenarios with frequent load fluctuations, the applicability of the optimization plan is significantly limited. The experimental results are shown in Table 1.

[0192] Table 1 Experimental data comparison table

[0193] index Traditional methods Method of the present invention Data preprocessing time (seconds) 180 25 Power generation prediction mean square error (MSE) <![CDATA[6.5MW 2 ]]> <![CDATA[2.1MW 2 ]]> Load demand forecast mean square error (MSE) <![CDATA[3.5MW 2 ]]> <![CDATA[0.99MW 2 ]]> Adjustment plan generation time (seconds) 45 8 Adaptability to load fluctuations (deviation) ±6MW ±0.5MW

[0194] The method of the present invention significantly reduces data transmission delays in the data preprocessing stage through the distributed deployment of edge computing nodes. Traditional methods rely on centralized cloud computing models, and data needs to be transmitted to the cloud through the network for processing, resulting in a long preprocessing time. However, the present invention directly completes data standardization, feature extraction and vectorization processing at the edge nodes, and only requires global summary of key indicators, which greatly improves processing efficiency.

[0195] In the power generation prediction link, the method of the present invention adopts a dynamic nonlinear model, takes vectorized meteorological characteristics and historical power generation data as key input variables, and combines the real-time optimization parameter adjustment mechanism to significantly reduce the prediction error. The static mathematical model of the traditional method is not adaptable enough to complex meteorological conditions and fails to effectively capture the dynamic impact of wind speed fluctuations on power generation. The nonlinear modeling of the present invention solves this problem through a dynamic correction strategy.

[0196] In load demand forecasting, the present invention significantly improves the responsiveness of forecast results to dynamic changes in grid demand by extracting load fluctuation characteristics, correcting wind power, and introducing nonlinear supplementary terms. Traditional methods only perform historical trend analysis based on linear regression models, which cannot reflect the impact of real-time wind power fluctuations on load demand, causing forecast results to deviate from actual demand. The method of the present invention is more suitable for complex load fluctuation scenarios, and its forecast accuracy is much higher than that of traditional methods.

[0197] In the generation of power adjustment schemes, the present invention adopts a customized particle swarm optimization algorithm, combined with power output limits and response speed limits to optimize the wind turbine power adjustment scheme in real time. Compared with the traditional linear optimization algorithm, the method of the present invention is more adaptable to changes in dynamic constraints, and the generated adjustment scheme has stronger adaptability and stability to grid load fluctuations, while the optimization time is greatly shortened. This improvement is due to the parallel search characteristics of the particle swarm optimization algorithm and the real-time processing capabilities of edge nodes.

[0198] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A wind farm energy management method based on edge computing, characterized in that: include: Collect wind turbine data in real time and pre-process the collected data; Calculate the power generation prediction value based on the preprocessed data; Calculate the power grid load demand forecast value based on the power generation forecast value; Calculating the grid load matching deviation based on the grid load demand forecast value; Generate a wind turbine power output adjustment plan based on the grid load matching deviation.

2. The wind farm energy management method based on edge computing according to claim 1, characterized in that: The wind turbine data includes power generation data, grid load data and meteorological data; The preprocessing of the collected data includes normalizing the power generation data, which is expressed as: Among them, P raw Represents the original wind turbine power generation data, P min Represents the minimum value in the data set, P max Represents the maximum value in the data set, P norm Represents the normalized wind turbine power generation data; The fluctuation amplitude of the power grid load data is calculated and expressed as: Among them, P grid,i represents the i-th data point of the power grid load data, represents the mean value of the grid load, N represents the total number of grid load data, ΔP grid Indicates the fluctuation amplitude of the power grid load; The meteorological data is decomposed into wind speed and wind direction to generate vectorized meteorological characteristic data, which is expressed as: Among them, v represents wind speed, θ represents wind direction, A two-dimensional vector representing wind speed and direction.

3. The wind farm energy management method based on edge computing according to claim 2, characterized in that: The predicted value of generated power is expressed as: P base =α1·v 3 +α2·v 2 +α3·v+α4 P weather =P base +β1·v·cos(θ)+β2·v·sin(θ) P wind,pred =P weather +β3·P norm Among them, P base represents the estimated power generation value, α1, α2, α3, α4 represent the wind turbine power curve fitting coefficients, β1, β2 represent the correction coefficients of the wind speed component, P wind,pred represents the final predicted power generation, β3 represents the correction factor of historical power generation; The calculation coefficient of the power generation prediction value is optimized by minimizing the mean square error method, which is expressed as: The optimized model output P wind,pred is the predicted value of generated power.

4. The wind farm energy management method based on edge computing according to claim 3, characterized in that: The grid load demand forecast value is expressed as: P grid,base =γ1·ΔP grid +γ2·P grid,hist P grid,corrected =P grid,base +γ3·P wind,pred P grid,pred =P grid,corrected +γ4·(ΔP grid ) 2 +γ5·(ΔP grid ·P wind,pred ) Among them, γ1, γ2 represent the regression coefficients of volatility and historical load, P grid,hist represents the historical grid load data, γ3 represents the correction coefficient of the impact of wind power on grid load demand, (ΔP grid ) 2 The square term of the load fluctuation amplitude is used to capture the nonlinear effect of load fluctuation, ΔP grid ·P wind,pred represents the interaction term between load fluctuation and wind power, γ4, γ5 represent the regression coefficients of nonlinear supplementation; The calculation coefficient of the power grid load demand forecast value is optimized by minimizing the mean square error method, which is expressed as: The optimized P grid,pred is the predicted value of grid load demand.

5. The wind farm energy management method based on edge computing according to claim 4, characterized in that: The grid load matching deviation is expressed as: ΔP grid =P grid,actual -P grid,pred Among them, P grid,actual Represented as actual grid load; Generating a wind turbine power output adjustment plan includes defining an optimization objective function and setting constraint conditions; The particle swarm optimization algorithm is used to iteratively solve the objective function, generate adjustment plans, and send the optimization results to the wind turbines.

6. The wind farm energy management method based on edge computing according to claim 5, characterized in that: The objective function is expressed as, Among them, J represents the optimization objective function, λ1,λ2 represent weight coefficients, P wind,adjusted,i represents the adjusted power output of the i-th wind turbine, P wind,current,i Indicates the actual output power of the i-th wind turbine.

7. The wind farm energy management method based on edge computing according to claim 6, characterized in that: The constraints include power output limit, response speed limit and total power limit; The constraint condition is expressed as, P wind,min,i ≤P wind,adjusted,i ≤P wind,,max,i |P wind,adjusted,i -P wind,current,i |≤R max,i ·T Among them, P wind,min,i、 P wind,max,i represents the minimum and maximum power limits of the i-th wind turbine, R max,i It represents the maximum adjustment rate of the i-th wind turbine, and T represents the adjustment time period.

8. A system using the wind farm energy management method based on edge computing as claimed in any one of claims 1 to 7, characterized in that: include: A data acquisition module (100), an energy prediction module (200), a state monitoring module (300), an energy efficiency evaluation module (400), and a power adjustment module (500); The data acquisition module (100) is used to collect wind turbine data in real time and pre-process the collected data; The energy prediction module (200) is used to calculate the power generation prediction value based on the preprocessed data; The state monitoring module (300) is used to calculate a power grid load demand prediction value based on the generated power prediction value; The energy efficiency evaluation module (400) is used to calculate the grid load matching deviation based on the grid load demand prediction value; The power adjustment module (500) is used to generate a wind turbine power output adjustment plan based on the grid load matching deviation.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the wind farm energy management method based on edge computing described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the wind farm energy management method based on edge computing described in any one of claims 1 to 7 are implemented.

Images