Method and system for reducing capacitive reactive power of offshore wind power generation system
Through multi-sensor data acquisition and deep learning model, high-precision prediction of the operating status and reactive power of the wind turbine unit, intelligent control algorithm is designed, which solves the problem that traditional compensation devices are difficult to cope with rapid changes, and realizes high-precision reactive power control, which enhances the Grid stability and system response speed.
Patent Information
- Application Number
- CN202410884525.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-03
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional fixed compensation devices are difficult to cope with the rapid changes in the operating state of wind turbines, resulting in excessive or insufficient reactive power compensation. The existing feedback control method is limited in control accuracy, making it difficult to meet the high-precision control requirements of offshore wind power generation systems.
By installing multiple sensors to collect data, we establish a relationship model for predicting the operating status of the wind turbine and the reactive power, use the multimodal neural network model to perform high-precision prediction, and design an intelligent control algorithm based on the prediction results to generate and perform filtering in real time The device control strategy is to dynamically adjust the reactive power threshold and fluctuation threshold to optimize reactive power control.
It significantly improves the accuracy of reactive power control, enhances the stability of the power grid, improves the system response speed, optimizes the operating efficiency of wind turbines, reduces maintenance costs, and improves the system's adaptability.
Smart Images

Figure CN119010231A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of optimization control of smart grids and wind power generation systems, and in particular to a method and system for reducing capacitive reactive power of an offshore wind power generation system. Background Art
[0002] As the global demand for renewable energy continues to grow, offshore wind power generation, as a clean and renewable energy form, is gradually emerging and becoming an important part of the global energy supply. However, in offshore wind power generation systems, especially in large-scale, multi-phase offshore wind farm projects, there are a series of technical and economic challenges, among which the length of submarine cables and capacitive reactive power are particularly prominent. These problems not only affect the efficiency of power transmission and energy utilization, but are also directly related to the assessment of the monthly power grid power regulation fee.
[0003] 1. Impact of submarine cable length
[0004] Offshore wind power generation systems usually involve the laying of a large number of submarine cables. In particular, in subsequent projects such as the second and third phases, a design of sharing a single outgoing line is often adopted to save costs and simplify the structure. This design significantly increases the total capacitive reactive power on the outgoing line, reducing the power transmission efficiency.
[0005] 2. The problem of capacitive reactive power
[0006] Capacitive reactive power is an important parameter in the power system, which mainly comes from the charging and discharging process of cables, capacitors and other equipment. In offshore wind power generation systems, due to the length of submarine cables and the design of shared outgoing lines, the capacitive reactive power of the entire station is often very high. This brings the following problems:
[0007] The increase in capacitive reactive power leads to a decrease in the system power factor, which reduces the power supply capacity of the power grid and increases the risk of voltage fluctuation and flicker in the power grid.
[0008] High capacitive reactive power not only affects the stability and power quality of the power grid, but also increases the assessment pressure of the monthly power regulation fee of the power grid. The power regulation fee is a reward and punishment measure introduced to ensure the power quality of the power grid, mainly for industrial users. In offshore wind power generation systems, high capacitive reactive power may cause the power supply bureau to fine the station, increasing operating costs. Summary of the invention
[0009] In view of the above-mentioned problems, the present invention is proposed.
[0010] Therefore, the technical problem solved by the present invention is that the traditional fixed compensation device is difficult to cope with the rapid changes in the operating state of the wind turbine, which often leads to excessive or insufficient reactive power compensation. The existing feedback control method has limited control accuracy and slow response speed, which makes it difficult to meet the high-precision control requirements of offshore wind power generation systems. Traditional methods perform poorly when dealing with dynamically changing operating states, and the control strategy lacks the ability to adjust and optimize in real time.
[0011] In order to solve the above technical problems, the present invention provides the following technical solutions: A method for reducing capacitive reactive power of an offshore wind power generation system, comprising:
[0012] Collect relevant data through various sensors installed on wind turbines, and analyze and process the collected data;
[0013] Using the processed data, a prediction model for the relationship between the operating status of wind turbines and reactive power is established;
[0014] An intelligent control algorithm is designed based on the prediction results of the relational prediction model, and the filter control strategy is generated and executed in real time.
[0015] As a preferred solution of the method for reducing capacitive reactive power of an offshore wind power generation system described in the present invention, wherein: collecting relevant data through a variety of sensors installed on the wind turbine generator set includes collecting wind turbine generator set operation data through power sensors, voltage and current sensors, power factor sensors and status sensors;
[0016] Collect meteorological data through anemometers, humidity sensors, air pressure sensors, and rainfall sensors;
[0017] Collect grid data through frequency sensors, voltage sensors, load sensors and fault sensors;
[0018] Collect environmental data through marine environment sensors, sunshine sensors and salt spray sensors;
[0019] Collect and record historical data.
[0020] As a preferred solution of the method for reducing capacitive reactive power of an offshore wind power generation system described in the present invention, wherein: analyzing and processing the collected data includes preprocessing the collected data, calculating the accuracy, completeness, real-time and consistency indicators of each data source using historical data and calculating the initial credibility indicator;
[0021] During the operation of the system, the data from each data source is detected in real time, the accuracy, real-time and consistency indicators are updated, and the credibility indicators are updated;
[0022] The data consistency calculation and normalization technology are used to process the data of each data source. The formula of the initial credibility index is expressed as:
[0023]
[0024] Among them, α i represents the initial credibility index of the i-th data source, A i represents the accuracy index of the i-th data source, C i represents the integrity index of the i-th data source, T i represents the real-time index of the i-th data source, S i represents the consistency index of the i-th data source, n represents the total number of data sources, A k represents the accuracy index of the kth data source, C k represents the integrity index of the kth data source, T k represents the real-time index of the kth data source, S k represents the consistency index of the kth data source;
[0025] α i The value range is from 0 to 1, α i =0 means that the data source is completely unreliable and all indicators are at the lowest value; i =1 indicates that the data source is completely reliable and all indicators are at the highest value;
[0026] According to the latest credibility index, the weight of each data source is dynamically adjusted, and the data of each data source is integrated using the latest weight;
[0027] The formula for dynamically adjusting weights and data fusion using the comprehensive credibility index calculation formula is expressed as:
[0028]
[0029] Among them, W i represents the weight of the i-th data source, A i represents the accuracy index of the i-th data source, C i represents the integrity index of the i-th data source, T i represents the real-time index of the i-th data source, S i represents the consistency index of the i-th data source, and n represents the total number of data sources;
[0030] W i The value range is 0 to 1, W i =0 means that the weight of the data source is the lowest value, and all indicators are the lowest values; W i =1 indicates that the weight of the data source is the highest and all indicators are the highest;
[0031] Construct a multimodal neural network model, conduct joint learning and feature extraction on multi-source data, and use a joint training strategy to optimize model parameters;
[0032] The loss function of joint training is expressed as:
[0033]
[0034] in, represents the loss function of joint training, n represents the total number of data sources, T represents the total time length, and X i (t) represents the data value of the i-th data source at time t, μ i represents the mean of the i-th data source, σ i represents the standard deviation of the ith data source, represents the predicted value of the i-th data source, and λ represents the regularization coefficient;
[0035] The range of is (0,∞), When , it means that the predicted value of the model completely matches the actual value and the loss is the minimum;
[0036] When it is close to ∞, it means that the prediction error of the model is large and the loss is maximum.
[0037] As a preferred solution of the method for reducing capacitive reactive power of an offshore wind power generation system described in the present invention, wherein: using a trained multimodal deep neural network model to fuse and predict real-time data to generate high-precision fused data;
[0038] The formula is:
[0039]
[0040] in, represents the predicted fusion data at time t, T represents the total time length, n represents the total number of data sources, and X i (t) represents the data value of the i-th data source at time t, μ i represents the mean of the i-th data source; σ i represents the standard deviation of the i-th data source;
[0041] The value range of is from negative infinity to positive infinity. Negative values indicate that the predicted data is lower than the actual data, and positive values indicate that the predicted data is higher than the actual data.
[0042] As a preferred solution of the method for reducing capacitive reactive power of an offshore wind power generation system described in the present invention, wherein: using the processed data to establish a relationship prediction model for predicting the operating status of a wind turbine and reactive power includes using a multi-layer neural network to construct a relationship prediction model, and the formula is expressed as follows:
[0043] Q(t)=σ(W2·tanh(+b1)+b2)
[0044] Where Q(t) represents the reactive power prediction value at time t, T represents the total time length, X(t) represents the processed data vector at time t, W1 and W2 represent the weight matrices of the neural network, b1 and b2 represent the bias vectors, and σ represents the Sigmoid activation function;
[0045] The value range of Q(t) is from 0 to 1. The closer the value is to 0, the lower the reactive power is, and the closer the value is to 1, the higher the reactive power is.
[0046] As a preferred solution of the method for reducing capacitive reactive power of an offshore wind power generation system described in the present invention, an intelligent control algorithm is designed based on the prediction results of the relationship prediction model, and the filter control strategy is generated and executed in real time, including dynamic adjustment according to real-time data, historical data and predicted data, and a dynamically adjusted reactive power threshold is designed, which is expressed as follows:
[0047]
[0048] Among them, Q thresh (t) represents the reactive power threshold dynamically adjusted at time t, μ history (t) represents the mean value of historical reactive power, σ history (t) represents the standard deviation of historical reactive power, k represents the adjustment coefficient of historical data, and α represents the weight coefficient of reactive power change rate;
[0049] When Q(t)>Q thresh When , the filter is added to reduce reactive power;
[0050] If Q(t)≤Q thresh When , the access of filters is reduced, the current state is maintained or further optimized.
[0051] As a preferred solution of the method for reducing capacitive reactive power of an offshore wind power generation system described in the present invention, wherein: an intelligent control algorithm is designed based on the prediction result of the relationship prediction model, and the filter control strategy is generated and executed in real time, and also includes designing a function to dynamically adjust the fluctuation threshold in combination with real-time data, historical data and environmental factors when the predicted reactive power fluctuates significantly. The formula is expressed as follows:
[0052]
[0053] Among them, ΔQ thresh (t) represents the reactive power fluctuation threshold dynamically adjusted at time t, E(t) is the environmental factor influence coefficient at time t, α, β, γ are adjustment coefficients, and k is the adjustment coefficient affected by historical data;
[0054] When the predicted reactive power fluctuation amplitude ΔQ pred (t)≥ΔQ thresh When , increase the amount of filter access;
[0055] When the predicted reactive power Q pred (t)<ΔQ thresh When the filter is connected, reduce the amount of filter access.
[0056] A system for reducing capacitive reactive power of an offshore wind power generation system, wherein:
[0057] The data acquisition module collects relevant data through various sensors installed on the wind turbine and analyzes and processes the collected data;
[0058] The power prediction module uses the processed data to establish a prediction model for the relationship between the operating status of the wind turbine and the reactive power;
[0059] The strategy module designs intelligent control algorithms based on the prediction results of the relational prediction model, and generates and executes filter control strategies in real time.
[0060] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein the processor implements the steps of any one of the methods of the present invention when executing the computer program.
[0061] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of any one of the methods of the present invention.
[0062] Beneficial effects of the invention: The method for reducing the capacitive reactive power of an offshore wind power generation system provided by the invention performs high-precision prediction of the relationship between the operating status of a wind turbine and reactive power through multi-sensor data acquisition and a deep learning model, thereby significantly improving the accuracy of reactive power control, enhancing grid stability, increasing system response speed, optimizing wind turbine operating efficiency, reducing maintenance costs, and improving the adaptability of the system. 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 method for reducing capacitive reactive power of an offshore wind power generation system provided by the first embodiment of the present invention;
[0065] Figure 2 A method for reducing capacitive reactive power of an offshore wind power generation system provided in accordance with a second embodiment of the present invention is a diagram of a wind turbine set in standby state. 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] Example 1
[0068] Reference Figure 1 , as an embodiment of the present invention, provides a method for reducing capacitive reactive power of an offshore wind power generation system, comprising:
[0069] S1: Collect relevant data through various sensors installed on wind turbines, and analyze and process the collected data.
[0070] Collect wind turbine operation data through power sensors, voltage and current sensors, power factor sensors and status sensors;
[0071] Collect meteorological data through anemometers, humidity sensors, air pressure sensors, and rainfall sensors;
[0072] Collect grid data through frequency sensors, voltage sensors, load sensors and fault sensors;
[0073] Collect environmental data through marine environment sensors, sunshine sensors and salt spray sensors;
[0074] Collect and record historical data.
[0075] Preprocess the collected data, use historical data to calculate the accuracy, completeness, real-time and consistency indicators of each data source and calculate the initial credibility indicators;
[0076] During the operation of the system, the data from each data source is detected in real time, the accuracy, real-time and consistency indicators are updated, and the credibility indicators are updated;
[0077]
[0078]
[0079] Among them, A i represents the accuracy index of the i-th data source, t represents the time point, T represents the total time length, and X i (t) represents the data value of the i-th data source at time t, X actual (t) represents the actual value, X actual,max Indicates the maximum value of the actual value, T i represents the real-time index of the i-th data source, exp represents the exponential function, T receive,i (t) represents the time when the i-th data source receives data at time t, T send,i (t) represents the time when the i-th data source sends data at time t, T represents the total time length, S i represents the consistency index of the i-th data source, X i (t) represents the data value of the i-th data source at time t, X j (t) represents the data value of the jth data source at time t, and n represents the total number of data sources.
[0080] The data consistency calculation and normalization technology are used to process the data of each data source. The formula of the initial credibility index is expressed as:
[0081]
[0082] Among them, α i represents the initial credibility index of the i-th data source, A i represents the accuracy index of the i-th data source, C i represents the integrity index of the i-th data source, T i represents the real-time index of the i-th data source, S i represents the consistency index of the i-th data source, n represents the total number of data sources, A k represents the accuracy index of the kth data source, C k represents the integrity index of the kth data source, T k represents the real-time index of the kth data source, S k Represents the consistency indicator of the kth data source.
[0083] α i The value range is from 0 to 1, α i=0 means that the data source is completely unreliable and all indicators are at the lowest value; i =1 indicates that the data source is completely reliable and all indicators are at the highest value.
[0084] According to the latest credibility index, the weight of each data source is dynamically adjusted, and the data of each data source is integrated using the latest weight.
[0085] The formula for dynamically adjusting weights and data fusion using the comprehensive credibility index calculation formula is expressed as:
[0086]
[0087] Among them, W i represents the weight of the i-th data source, A i represents the accuracy index of the i-th data source, C i represents the integrity index of the i-th data source, T i represents the real-time index of the i-th data source, S i represents the consistency index of the i-th data source, and n represents the total number of data sources.
[0088] W i The value range is 0 to 1, W i =0 means that the weight of the data source is the lowest value, and all indicators are the lowest values; W i =1 indicates that the weight of the data source is the highest and all indicators are the highest;
[0089] Construct a multimodal neural network model, perform joint learning and feature extraction on multi-source data, and use a joint training strategy to optimize model parameters.
[0090] The loss function of joint training is expressed as:
[0091]
[0092] in, represents the loss function of joint training, n represents the total number of data sources, T represents the total time length, and X i (t) represents the data value of the i-th data source at time t, μ i represents the mean of the i-th data source, σ i represents the standard deviation of the ith data source, represents the predicted value of the i-th data source, and λ represents the regularization coefficient.
[0093] The range of is (0,∞), When , it means that the predicted value of the model completely matches the actual value and the loss is the minimum.
[0094] When it is close to ∞, it means that the prediction error of the model is large and the loss is maximum.
[0095] The trained multimodal deep neural network model is used to fuse and predict real-time data to generate high-precision fused data.
[0096] The formula is:
[0097]
[0098] in, represents the predicted fusion data at time t, T represents the total time length, n represents the total number of data sources, and X i (t) represents the data value of the i-th data source at time t, μ i represents the mean of the i-th data source; σ i Represents the standard deviation of the ith data source.
[0099] The value range of is from negative infinity to positive infinity. Negative values indicate that the predicted data is lower than the actual data, and positive values indicate that the predicted data is higher than the actual data.
[0100] The data sources are comprehensively evaluated by three indicators: accuracy, real-time and consistency. These indicators can reflect the reliability and timeliness of the data and provide a basis for the subsequent calculation of comprehensive credibility. The comprehensive credibility α is calculated by weighted average method based on the accuracy, completeness, real-time and consistency of each data source. i , thereby quantifying the credibility of each data source. This method can balance the importance of different indicators and ensure that each indicator has an appropriate impact on the final result. i , dynamically weight each data source according to the comprehensive credibility, ensuring that the weight of the data source in different time periods can be adaptively adjusted to reflect the changes in the reliability of the current data source. Use the loss function in deep learning to train the multimodal neural network model, combine the characteristics of multiple data sources, optimize the model parameters, and ensure that the model fits the actual data with high precision. 5. Prediction formula design: Through the multimodal deep neural network model, real-time data is fused and predicted to generate high-precision fused data. The formula combines methods such as standardization, feature extraction, and dynamic feature extraction to ensure the accuracy and stability of the prediction results.
[0101] By comprehensively utilizing multi-source data and using deep learning models to make high-precision predictions of reactive power, the operating efficiency of wind turbines is improved. By generating and executing filter control strategies in real time, the fluctuation amplitude of reactive power is reduced, and the stability of the power grid is enhanced. By using dynamic weighting and real-time adjustment strategies, the system can dynamically adjust the control strategy according to the current operating status and environmental conditions, improving the system's adaptive ability. By jointly training the loss function to optimize the model parameters, the effectiveness and optimization of the control strategy are ensured, energy loss is reduced, and operating costs are reduced. The use of deep learning models and intelligent control algorithms enables the system to quickly respond to changes in the operating status of wind turbines, achieve real-time control, and improve the response speed of the system.
[0102] S2: Use the processed data to establish a prediction model for the relationship between the operating status of the wind turbine and the reactive power.
[0103] A multi-layer neural network is used to build a relationship prediction model, and the formula is expressed as:
[0104] Q(t)=σ(W2·tanh(+b1)+b2)
[0105] Among them, Q(t) represents the reactive power prediction value at time t, T represents the total time length, X(t) represents the processed data vector at time t, W1 and W2 represent the weight matrices of the neural network, b1 and b2 represent the bias vectors, and σ represents the Sigmoid activation function.
[0106] The value range of Q(t) is from 0 to 1. The closer the value is to 0, the lower the reactive power is, and the closer the value is to 1, the higher the reactive power is.
[0107] The multi-layer neural network (MLP) model, through multiple hidden layers and nonlinear activation functions, can effectively capture the complex nonlinear relationship of the input data and improve the model's expressiveness and prediction accuracy. The hyperbolic tangent activation function can standardize the input data to the range of -1 to 1, which helps to alleviate the gradient vanishing problem. The Sigmoid activation function maps the output to the range of 0 to 1, which facilitates the understanding and application of reactive power prediction results. By introducing weight matrices W1 and W2, as well as bias vectors b1 and b2, the input data can be linearly transformed and translated, further enhancing the model's expressiveness.
[0108] Through high-precision reactive power prediction, filter control strategies can be generated and executed in real time, reactive power control of wind turbines can be optimized, and the operating efficiency of wind farms can be improved. Accurate control of reactive power helps reduce grid voltage fluctuations, enhance grid stability, and ensure safe operation of the power system. Multi-layer neural network models and intelligent control algorithms can be used to achieve intelligent management of wind turbines, reduce manual intervention, and improve the level of management automation and scientific decision-making.
[0109] S3: Design an intelligent control algorithm based on the prediction results of the relational prediction model to generate and execute the filter control strategy in real time.
[0110] Dynamically adjust according to real-time data, historical data and predicted data, and design a dynamically adjusted reactive power threshold. The formula is expressed as:
[0111]
[0112] Among them, Q thresh (t) represents the reactive power threshold dynamically adjusted at time t, μ history (t) represents the mean value of historical reactive power, σ history (t) represents the standard deviation of historical reactive power, k represents the adjustment coefficient of historical data, and α represents the weight coefficient of reactive power change rate;
[0113] When Q(t)>Q thresh When the reactive power is too high, the filter is connected to reduce the reactive power.
[0114] If Q(t)≤Q thresh When , the access of filters is reduced, the current state is maintained or further optimized.
[0115] When the predicted reactive power fluctuates significantly, a function is designed to dynamically adjust the fluctuation threshold by combining real-time data, historical data and environmental factors. The formula is:
[0116]
[0117] Among them, ΔQ thresh (t) represents the reactive power fluctuation threshold dynamically adjusted at time t, E(t) is the environmental factor influence coefficient at time t, α, β, γ are adjustment coefficients, and k is the adjustment coefficient affected by historical data.
[0118] When the predicted reactive power fluctuation amplitude ΔQ pred (t)≥ΔQ thresh When the filter is connected, increase the amount of filter access.
[0119] When the predicted reactive power Q pred (t)<ΔQ thresh When the filter is connected, reduce the amount of filter access.
[0120] By dynamically adjusting the reactive power threshold and fluctuation threshold, the fluctuation range of reactive power can be accurately controlled. The shortcomings of the traditional fixed threshold method in dealing with complex operating environments are avoided, and the control accuracy and grid stability are improved. The reactive power change rate is calculated in real time, and the threshold is adjusted dynamically according to the change rate, which can quickly respond to the operation changes of wind turbines. Compared with the traditional method, this design can adjust the access status of the filter faster to ensure the stable operation of the system. The precise control of reactive power reduces the fluctuation of grid voltage, enhances the stability of the grid, and ensures the safe operation of the power system. This is especially important for large-scale offshore wind farms. By dynamically adjusting the reactive power threshold and fluctuation threshold, the operating efficiency of wind turbines can be optimized, unnecessary reactive power output can be reduced, energy loss can be reduced, and power generation efficiency can be improved. Combined with the multi-layer neural network model and intelligent control algorithm, intelligent management of wind turbines is realized. The system can learn and adjust autonomously, reducing manual intervention, improving the automation level of management and the scientific nature of decision-making.
[0121] On the other hand, this embodiment also provides a system for reducing capacitive reactive power of an offshore wind power generation system, which includes:
[0122] The data acquisition module collects relevant data through various sensors installed on the wind turbine and analyzes and processes the collected data.
[0123] The power prediction module uses the processed data to establish a prediction model for the relationship between the operating status of the wind turbine and the reactive power.
[0124] The strategy module designs intelligent control algorithms based on the prediction results of the relational prediction model, and generates and executes filter control strategies in real time.
[0125] If the above 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 method 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 and other media that can store program codes.
[0126] 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.
[0127] 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.
[0128] 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.
[0129] Example 2
[0130] The experiment was conducted on 10 wind turbines in an offshore wind farm. Wind speed sensors, wind direction sensors, temperature sensors, humidity sensors, voltage sensors, current sensors and reactive power sensors were installed on each wind turbine. After the sensors were installed, the operation data of the wind turbines were collected in real time. The data acquisition system recorded the data once a minute and transmitted the data to the central database for storage and processing. At the end of each day, the collected data was transmitted to the data processing center for processing. Data processing includes data cleaning, outlier processing and feature extraction. The processed data was used to train a multi-layer neural network model. The model input includes parameters such as wind speed, wind direction, temperature, humidity, voltage and current, and the output is the predicted value of reactive power. The model uses 80% of the data for training and 20% of the data for verification to ensure the accuracy and robustness of the model. Based on the results of the relational prediction model, an intelligent control algorithm was designed. The algorithm dynamically adjusts the access amount of the filter by real-time monitoring of the operating status and reactive power of the wind turbine to optimize the output of reactive power. Specifically, when the predicted reactive power is higher than the set threshold, the access amount of the filter is increased to reduce the reactive power; when the predicted reactive power is lower than the set threshold, the access amount of the filter is reduced to maintain the stability of the power grid. After the test, all test data are collected for summary and analysis. The effect of the invention is evaluated by comparing the changes in reactive power, voltage and current before and after intelligent control. The experimental results are shown in Table 1.
[0131] Table 1 Experimental results data table
[0132]
[0133] It can be seen from the table that after the intelligent control algorithm is adopted, the reactive power values of the 10 wind turbines fluctuate between 290 and 315VAR, with a small fluctuation range, and all remain within a reasonable range. Before the intelligent control algorithm is adopted, the reactive power value fluctuates greatly, with a maximum value of 320VAR and a minimum value of 280VAR, and the fluctuation range is large, which is easy to affect the stability of the power grid. The data in the table show that after the intelligent control algorithm is adopted, the voltage value of the wind turbine fluctuates between 687 and 695V, and the current value fluctuates between 838 and 875A, both of which remain within a stable range. Before the intelligent control algorithm is adopted, the voltage and current fluctuate greatly, with a maximum voltage value of 700V, a minimum voltage value of 685V, a maximum current value of 880A, and a minimum current value of 830A. Through the implementation of the intelligent control algorithm, the reactive power control accuracy of the wind turbine is significantly improved, the output of voltage and current is optimized, and the stability of the power grid is enhanced. Compared with the existing technology, the invention can not only monitor the operating status of wind turbines in real time, but also dynamically adjust the control strategy according to the prediction results, thus realizing the precise control of reactive power. This adaptive intelligent control method not only improves the operating efficiency of wind turbines, but also reduces energy loss, and has important engineering application value and economic benefits.
[0134] 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 method for reducing capacitive reactive power of an offshore wind power generation system, It is characterized in that include: Collect relevant data through various sensors installed on wind turbines, and analyze and process the collected data; Using the processed data, a prediction model for the relationship between the operating status of wind turbines and reactive power is established; An intelligent control algorithm is designed based on the prediction results of the relational prediction model, and the filter control strategy is generated and executed in real time.
2. A method for reducing capacitive reactive power of an offshore wind power generation system as claimed in claim 1, Features: Collect relevant data through various sensors installed on the wind turbine, including collecting wind turbine operation data through power sensors, voltage, current sensors, power factor sensors and status sensors; Collect meteorological data through anemometers, humidity sensors, air pressure sensors, and rainfall sensors; Collect grid data through frequency sensors, voltage sensors, load sensors and fault sensors; Collect environmental data through marine environment sensors, sunshine sensors and salt spray sensors; Collect and record historical data.
3. A method for reducing capacitive reactive power of an offshore wind power generation system as claimed in claim 2, Features: Analyzing and processing the collected data includes preprocessing the collected data, calculating the accuracy, completeness, real-time and consistency indicators of each data source using historical data and calculating the initial credibility indicators; During the operation of the system, the data from each data source is detected in real time, the accuracy, real-time and consistency indicators are updated, and the credibility indicators are updated; The data consistency calculation and normalization technology are used to process the data of each data source. The formula of the initial credibility index is expressed as: Among them, α i represents the initial credibility index of the i-th data source, A i represents the accuracy index of the i-th data source, C i represents the integrity index of the i-th data source, T i represents the real-time index of the i-th data source, S i represents the consistency index of the i-th data source, n represents the total number of data sources, A k represents the accuracy index of the kth data source, C k represents the integrity index of the kth data source, T k represents the real-time index of the kth data source, S k represents the consistency index of the kth data source; α i The value range is from 0 to 1, α i =0 means that the data source is completely unreliable and all indicators are at the lowest value; i =1 indicates that the data source is completely reliable and all indicators are at the highest value; According to the latest credibility index, the weight of each data source is dynamically adjusted, and the data of each data source is integrated using the latest weight; The formula for dynamically adjusting weights and data fusion using the comprehensive credibility index calculation formula is expressed as: Among them, W i represents the weight of the i-th data source, A i represents the accuracy index of the i-th data source, C i represents the integrity index of the i-th data source, T i represents the real-time index of the i-th data source, S i represents the consistency index of the i-th data source, and n represents the total number of data sources; W i The value range is from 0 to 1, W i =0 means that the weight of the data source is the lowest value, and all indicators are the lowest values; W i =1 indicates that the weight of the data source is the highest and all indicators are the highest; Construct a multimodal neural network model, conduct joint learning and feature extraction on multi-source data, and use a joint training strategy to optimize model parameters; The loss function of joint training is expressed as: in, represents the loss function of joint training, n represents the total number of data sources, T represents the total time length, and X i (t) represents the data value of the i-th data source at time t, μ i represents the mean of the i-th data source, σ i represents the standard deviation of the ith data source, represents the predicted value of the i-th data source, and λ represents the regularization coefficient; The range of is (0,∞), When , it means that the predicted value of the model completely matches the actual value and the loss is the minimum; When it is close to ∞, it means that the prediction error of the model is large and the loss is maximum.
4. A method for reducing capacitive reactive power of an offshore wind power generation system as claimed in claim 3, Features: Use the trained multimodal deep neural network model to fuse and predict real-time data to generate high-precision fused data; The formula is: in, represents the predicted fusion data at time t, T represents the total time length, n represents the total number of data sources, and X i (t) represents the data value of the i-th data source at time t, μ i represents the mean of the i-th data source; σ i represents the standard deviation of the i-th data source; The value range of is from negative infinity to positive infinity. Negative values indicate that the predicted data is lower than the actual data, and positive values indicate that the predicted data is higher than the actual data.
5. A method for reducing capacitive reactive power of an offshore wind power generation system as claimed in claim 4, Features: Using the processed data to establish a relationship prediction model for predicting the operating status of wind turbines and reactive power includes using a multi-layer neural network to build a relationship prediction model, and the formula is expressed as: Q(t)=σ(W 2 ·tanh(+b 1 )+b 2 ) Where Q(t) represents the reactive power prediction value at time t, T represents the total time length, X(t) represents the processed data vector at time t, and W 1 and W 2 represents the weight matrix of the neural network, b 1 and b 2 represents the bias vector, σ represents the Sigmoid activation function; The value range of Q(t) is from 0 to 1. The closer the value is to 0, the lower the reactive power is, and the closer the value is to 1, the higher the reactive power is.
6. A method for reducing capacitive reactive power of an offshore wind power generation system as claimed in claim 5, Features: Based on the prediction results of the relational prediction model, an intelligent control algorithm is designed to generate and execute the filter control strategy in real time, including dynamic adjustment based on real-time data, historical data and predicted data, and design a dynamically adjusted reactive power threshold, which is expressed as: Among them, Q thresh (t) represents the reactive power threshold dynamically adjusted at time t, μ history (t) represents the mean value of historical reactive power, σ history (t) represents the standard deviation of historical reactive power, k represents the adjustment coefficient of historical data, and α represents the weight coefficient of reactive power change rate; When Q(t)>Q thresh When , the filter is added to reduce reactive power; If Q(t)≤Q thresh When , the access of filters is reduced, the current state is maintained or further optimized.
7. A method for reducing capacitive reactive power of an offshore wind power generation system as claimed in claim 6, Features: Designing an intelligent control algorithm based on the prediction results of the relational prediction model, generating and executing the filter control strategy in real time also includes designing a function to dynamically adjust the fluctuation threshold by combining real-time data, historical data and environmental factors when the predicted reactive power fluctuates significantly. The formula is expressed as: Among them, ΔQ thresh (t) represents the reactive power fluctuation threshold dynamically adjusted at time t, E(t) is the environmental factor influence coefficient at time t, α, β, γ are adjustment coefficients, and k is the adjustment coefficient affected by historical data; When the predicted reactive power fluctuation amplitude ΔQ pred (t)≥ΔQ thresh When , increase the amount of filter access; When the predicted reactive power Q pred (t)<ΔQ thresh When the filter is connected, reduce the amount of filter access.
8. A system for reducing capacitive reactive power of an offshore wind power generation system using the method according to any one of claims 1 to 7, Features: The data acquisition module collects relevant data through various sensors installed on the wind turbine and analyzes and processes the collected data; The power prediction module uses the processed data to establish a prediction model for the relationship between the operating status of the wind turbine and the reactive power; The strategy module designs intelligent control algorithms based on the prediction results of the relational prediction model, and generates and executes filter control strategies in real time.
9. A computer device, include: Memory and processor; The memory stores a computer program, characterized in that: when the processor executes the computer program, the processor implements the steps of a method for reducing capacitive reactive power of an offshore wind power generation system.
10. A computer-readable storage medium having a computer program stored thereon, Features: When the computer program is executed by a processor, steps of a method for reducing capacitive reactive power of an offshore wind power generation system are implemented.
Citation Information
Cited By
Reactive voltage control method and device for wind power plant
CN119864817A
