Method and system for predicting concentrated power of wind power plant based on big data

Abstract

The invention relates to a wind power plant concentrated power prediction method based on big data, and compared with the prior art, the method carries out the hierarchical training through a general control platform based on the data of all candidate electric fields, obtains a plurality of data sets with labels, carries out the training based on each data set, and obtains a plurality of models. Therefore, the general control platform can provide models of electric fields with other similar conditions for each wind power plant for power prediction; and optimization is continued according to actual data of the electric field in the operation and development process of the electric field, and a sustainable optimization model independently belonging to each electric field is established. And the universality and the resource cost of centralized power prediction are greatly improved.

Description

Technical Field

[0001] The present application relates to the field of power equipment operation and maintenance management, and in particular to a wind farm centralized power prediction method and system based on big data, computer equipment, and computer-readable storage medium. Background Art

[0002] With the coming of the electricity parity era and the acceleration of the market-oriented electricity trading process, cost reduction and efficiency improvement have become the core goals of the operation and management of new energy companies. In addition, the centralized management of new energy wind farms is becoming more and more centralized. The need for new energy companies to conduct unified forecasting and management of new energy wind farms on the group side has become very urgent.

[0003] The centralized power forecasting solution can maximize the comprehensive utilization benefits in group power generation management, power trading, and other aspects, realize remote optimization management of each power station, help new energy group enterprises achieve unified power forecasting management, reduce comprehensive operation and maintenance costs, and provide auxiliary support for group power trading; help the power grid coordinate the planning of power generation plans, unify management and scheduling, improve work efficiency, and reduce management costs.

[0004] Currently, major wind farms use their own historical power generation data as training data to train intelligent prediction models when predicting concentrated power. However, this method requires the computing equipment within the power plant to have considerable processing power. For some early-established wind farms, additional computing equipment needs to be purchased, which is costly.

[0005] In addition, for small or newly established wind farms, there is not enough historical data. For example, the external environment in each quarter varies greatly, and the adjacent technical data is not of reference value at all. If the data from the previous year is not used for training, the final prediction results will have a large deviation. Even within one quarter, the lack of sufficient data will lead to a very low prediction accuracy. Summary of the Invention

[0006] The embodiments of the present application provide a method, system, computer device and computer-readable storage medium for centralized power prediction of wind farms based on big data, so as to at least solve the problem in the related art of poor accuracy of wind farm power prediction models due to a lack of sufficient historical data.

[0007] In a first aspect, embodiments of the present application provide a method for predicting centralized power of wind farms based on big data, which implements centralized power prediction through a master control platform and distributed control platforms located in each wind farm. The method includes:

[0008] The master control platform obtains power operation and maintenance data of candidate wind farms and uses big data technology to filter and preprocess the power operation and maintenance data to obtain power data sets at multiple levels;

[0009] Based on the data sets of the multiple levels, blank models are trained respectively to obtain multiple finished product power prediction models corresponding to different levels and different types;

[0010] In response to an interactive instruction with any sub-control platform, based on the specific conditions of the wind farm contained in the interactive instruction, a specific finished product power prediction model matching the wind farm is returned;

[0011] The sub-control platform deploys the specific finished product power prediction model in combination with local business logic, and performs power prediction based on local power operation and maintenance data through the specific finished product power prediction model.

[0012] In some embodiments, big data technology is used to filter and preprocess the power operation and maintenance data to obtain power data sets at multiple levels, including:

[0013] Using big data technology to filter the power operation and maintenance data to obtain data related to power generation, and building a power database based on the data related to power generation;

[0014] According to different types of power impact parameters, the data in the power database is divided into multiple levels of power data sets, wherein the item classification of the low-level power data sets is more detailed than that of the high-level power data sets;

[0015] A classification label is set for each level of power data set and sent to the sub-control platform. The classification label includes various parameters used for classification and corresponding parameter ranges.

[0016] In some embodiments, big data technology is used to filter the power operation and maintenance data to obtain data related to power generation, including:

[0017] Acquiring data related to generated power from the power operation and maintenance data to obtain initial power-related data, wherein the initial power-related data includes: wind farm geographic data, climate data, power generation-related equipment data, and historical generated power;

[0018] Based on parameter integrity rules, the initial power-related data are screened to obtain a plurality of power data with data integrity, and the power database is constructed based on the plurality of power data.

[0019] In some embodiments, after screening the initial power-related data based on parameter integrity rules, the method further includes:

[0020] Obtain data with missing parameters, and package the missing data corresponding to each wind farm to generate data to be supplemented;

[0021] The data to be supplemented is sent to the corresponding wind farm sub-control platform to instruct the sub-control platform to fill in the missing data to obtain complete supplemented power data, and return the supplemented power data.

[0022] In some embodiments, the method further comprises:

[0023] When the amount of historical power operation and maintenance data of any target wind farm is greater than a preset threshold, a dedicated database is set up for the target wind farm, and the dedicated database is divided into multiple levels of power data sets according to the operation and maintenance information of the target wind farm.

[0024] In some embodiments, the method further comprises:

[0025] By performing similarity calculation or graphical analysis on all power data sets under the target level, it is determined whether there are abnormal data in the power data sets that do not conform to the overall data rules.

[0026] If yes, in the power database, the target power data set with abnormal data and the upper power data set of the target power data set are deleted.

[0027] Furthermore, a data source of the abnormal data is obtained, and the abnormal data is sent to a sub-control platform of a corresponding wind farm according to the data source.

[0028] In some embodiments, after filtering and preprocessing the power operation and maintenance data using big data technology to obtain power data sets at multiple levels, the method further includes:

[0029] The sub-control platform obtains a specific power data set from the master control platform based on its own equipment parameters and local climate parameters;

[0030] Based on the specific power data set and local power related data, the blank model is trained to obtain a local end power prediction model, and the local end power prediction model is deployed in the actual power prediction task of the wind farm.

[0031] In a second aspect, an embodiment of the present application provides a wind farm centralized power prediction system based on big data, which implements centralized power prediction through a master control platform and a sub-control platform located in each wind farm. The system includes a master control platform and a sub-control platform, wherein:

[0032] The master control platform is used to obtain power operation and maintenance data of candidate wind farms, and use big data technology to filter and preprocess the power operation and maintenance data to obtain power data sets at multiple levels;

[0033] Furthermore, based on the data sets of the multiple levels, blank models are trained respectively to obtain multiple finished product power prediction models corresponding to different levels and types, and in response to an interactive instruction with any sub-control platform, a specific finished product power prediction model matching the wind farm is returned according to the specific conditions of the wind farm contained in the interactive instruction;

[0034] The sub-control platform is used to deploy the specific finished product power prediction model in combination with local business logic, and perform power prediction based on local power operation and maintenance data through the specific finished product power prediction model.

[0035] In a third aspect, an embodiment of the present application provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first aspect above when executing the computer program.

[0036] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect above.

[0037] Compared to related technologies, the embodiments of this application provide a method for centralized wind farm power prediction based on big data. Compared to existing technologies, this solution uses a master control platform to perform hierarchical training based on the data of all candidate electric fields to obtain multiple labeled data sets. Subsequently, training is performed on each data set to obtain multiple models. Therefore, the master control platform can provide each wind farm with models of other electric fields with similar conditions for power prediction; and during the operation and development of the electric field, it can continue to optimize based on the actual electric field data to establish a sustainable optimization model unique to each electric field. This greatly improves the versatility and resource cost of centralized power prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0039] Figure 1 This is a flow chart of a method for predicting concentrated power of a wind farm based on big data according to an embodiment of the present application;

[0040] Figure 2 This is a structural block diagram of a wind farm centralized power prediction system based on big data according to an embodiment of the present application;

[0041] Figure 3 Schematic diagram of the internal structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is described and illustrated below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely used to explain this application and are not intended to limit this application. Based on the embodiments provided in this application, all other embodiments obtained by those of ordinary skill in the art without making any creative efforts are within the scope of protection of this application.

[0043] Obviously, the drawings described below are merely examples or embodiments of the present application. Those skilled in the art can, without inventive effort, apply the present application to other similar scenarios based on these drawings. Furthermore, it is also understood that, although the effort involved in such a development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, changes in design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as an insufficiency of the content disclosed in this application.

[0044] References to "embodiments" in this application mean that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it refer to independent or alternative embodiments that are mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described in this application may be combined with other embodiments unless there is a conflict.

[0045] Unless otherwise defined, the technical or scientific terms used in this application should have the ordinary meaning understood by a person of ordinary skill in the technical field to which this application belongs. The words "one", "a", "the" and the like used in this application do not indicate a limit on quantity and may indicate the singular or plural. The terms "include", "comprise", "have" and any variations thereof used in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units that are inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The word "multiple" used in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific order for the objects.

[0046] Existing technologies generally require human personnel to monitor and maintain renewable energy wind farms, resulting in high O&M costs. However, centralized power forecasting enables centralized O&M of renewable energy wind farms within a centralized control center within a group. This allows for real-time monitoring of wind farm data and operational status from multiple perspectives, including group, region, and plant level. This allows for cluster management of national or regional power plants, reducing the O&M costs of decentralized, multi-channel power plants. Furthermore, centralized management of O&M personnel allows for reduced or even unmanned operation at each wind farm, saving on overall O&M costs while improving efficiency. From the perspective of provincial and municipal grid dispatching, this clustered forecasting approach facilitates unified forecast management, reducing dispatching and management costs.

[0047] At the same time, detailed guidance can be provided for power plant safety assurance. Because the group's centralized control center monitors the status of each power plant, it can meticulously track the overall operational status of the power plants, monitor wind farm output and forecast information in real time, and directly access real-time data from each device, accurately controlling the operational details of the wind farm. In the event of an alarm, the problem can be quickly reported and addressed, significantly reducing assessment risks.

[0048] With the trend toward market-based electricity trading, new energy group companies will increasingly participate. Centralized power forecasting can capture wind farm power generation data in real time, providing accurate power and generation forecasts. This provides precise quotes and quantity reporting for medium- and long-term, cross-regional, intra-provincial, and spot transactions, ensuring new energy power generation companies' confidence in the power generation capacity of their wind farms and improving overall profitability. Furthermore, the solution can be integrated with power trading, assisting group companies in unified transaction planning.

[0049] The centralized power forecasting solution can maximize the comprehensive utilization benefits in group power generation management, power trading, and other aspects, realize remote optimization management of each power station, help new energy group enterprises achieve unified power forecasting management, reduce comprehensive operation and maintenance costs, and provide auxiliary support for group power trading; help the power grid coordinate the planning of power generation plans, unify management and scheduling, improve work efficiency, and reduce management costs.

[0050] In summary, centralized power forecasting can improve the management efficiency of new energy groups and reduce overall operating costs. It has great application value for the refinement and standardization of power generation operation management of new energy enterprises.

[0051] Currently, major wind farms use their own historical power generation data as training data to train intelligent prediction models when predicting concentrated power. However, this method requires the computing equipment within the power plant to have considerable processing power. For some early-established wind farms, additional computing equipment needs to be purchased, which is costly.

[0052] However, for small or newly established wind farms, there is not enough historical data. For example, the external environment in each quarter varies greatly, and the adjacent technical data is not of reference value at all. If the data from the previous year is not used for training, the final prediction results will have a large deviation. Even within one quarter, the lack of sufficient data will lead to a very low prediction accuracy.

[0053] In view of this, the present application provides a method for predicting centralized power of wind farms based on big data. The method realizes centralized power prediction through a master control platform and sub-control platforms located in each wind farm. Figure 1 This is a flow chart of a method for predicting concentrated power of a wind farm based on big data according to an embodiment of the present application, such as Figure 1 As shown, the method includes the following steps:

[0054] S101, the master control platform, obtains power operation and maintenance data of relevant wind farms and uses big data technology to filter and preprocess the power operation and maintenance data to obtain multi-level power data sets;

[0055] In this embodiment, a master control platform and a sub-control platform are provided. The master control platform is deployed in the power grid operation and maintenance center or cloud server and has large-scale data computing capabilities; the sub-control platforms are deployed in various distributed wind farms or centralized control departments.

[0056] It should be noted that in the technical solution of this embodiment, each customer entity can customize the sub-control platform according to its own needs based on the site environment and meteorological conditions. The sub-control platform can be used specifically to process local or specific regional data and does not need to carry the large amount of data of the power grid. This not only reduces the requirements for platform deployment equipment, but also better adapts to wind farms of different sizes and geographical distributions.

[0057] At the same time, by establishing a connection between the master control platform and the sub-control platforms, some data processing and prediction model training computing scenarios can be dispersed across the sub-control platforms, thereby reducing the computing pressure on the master control platform and improving the computing speed of the entire system.

[0058] Specifically, the candidate wind farms in this embodiment are not specifically limited and may be all wind farms in a certain area, or wind farms in different areas belonging to the same group company;

[0059] Furthermore, the master control platform obtains the power operation and maintenance data of the candidate wind farms and processes it using big data, including the following subdivision steps:

[0060] Step 1: In the data acquisition module of the master control data platform, data is obtained from each wind farm based on the existing data call permissions of each wind farm, and all data within the permission range is searched;

[0061] The data collection module first verifies its access rights to the data for each wind farm. This typically involves interacting with authentication systems (such as OAuth and LDAP) to confirm that the master control data platform or the relevant user / service has the legal authority to access the data for a specific wind farm. This authorization verification can include checking information such as API keys, tokens, user identities, and roles to ensure that the requester is legitimate and authorized to access the requested dataset.

[0062] Secondly, the master control platform needs to establish interfaces with the data management systems of each wind farm (such as SCADA and CMS). These interfaces may be based on protocols such as HTTP, HTTPS, and WebSocket, allowing data to be exchanged in formats such as JSON and XML. During the interface integration process, details such as the data exchange format, frequency, and security requirements must be clarified to ensure smooth and secure data transmission.

[0063] Finally, after the authority verification is passed, the data acquisition module will send a data query request to the data management system of each wind farm according to the preset query logic or user request.

[0064] In this embodiment, a query request for a wind farm might include information such as a time range, data type (e.g., wind speed, wind direction, power generation), and specific equipment or location, precisely specifying the dataset to be retrieved. Upon receiving the query request, the data management system retrieves and returns the corresponding data based on the request parameters and its internal data storage strategy.

[0065] S102, using big data technology to filter the power operation and maintenance data to obtain data related to power generation, and building a power database based on the data related to power generation;

[0066] In Step 2.1, in the initial screening state: all operation and maintenance data are screened and sorted to obtain data related to power generation. This data includes but is not limited to: geographical data of the power field, climate data, power generation-related equipment data, historical power generation data, and other power generation-related data;

[0067] The location information may include the latitude and longitude of the electric field, altitude, and terrain characteristics (such as plains, mountains, or near water); the surrounding environment may include the distribution of surrounding obstacles (such as mountains and buildings) and their impact on wind direction and speed;

[0068] Meteorological data can include temperature, humidity, air pressure, wind direction, wind speed, etc. This data directly affects the power generation efficiency of wind turbines. It is understandable that long-term accumulated meteorological data can be used to analyze climate patterns and trends, providing an important reference for power forecasting;

[0069] Power generation equipment data can include basic information about the wind turbine, such as model, rated power, and installation time. It can also include real-time or historical data on generator speed, blade angle, bearing temperature, gearbox oil temperature, etc., which reflect the health and efficiency of the equipment. It can also include historical records of equipment failures and their repairs, which help analyze the causes of equipment performance degradation and predict possible future failures.

[0070] Power generation curve: The actual power generation of wind turbines at different wind speeds is recorded, used to establish a mapping relationship between power generation and wind speed. The power generation data within a statistical period (such as a day, month, or year) is used to evaluate the overall power generation capacity and efficiency of the power plant.

[0071] It can be understood that the above steps can achieve a comprehensive screening of data related to power generation to obtain all potential basic data that can have positive and negative effects on power prediction.

[0072] In step 2.2, in order to ensure the accuracy of model training, it is necessary to ensure the integrity of the data;

[0073] Therefore, it is necessary to screen the initial power-related data based on the parameter integrity rule to obtain multiple power data with data integrity. That is, the selected data must include all parameter items related to the power generation, and no item can be missing; on the premise of ensuring data integrity, a power database is constructed based on multiple power data.

[0074] Specifically, to ensure data integrity, a "parameter integrity rule" must be established. This means that when selecting power-related data, all key parameters that affect power generation must be included. Examples include time of day, temperature, humidity, wind speed, wind direction, equipment model, altitude, and air pressure.

[0075] It's important to note that, generally speaking, the more factors included, the more accurate the coverage of parameters affecting power output. For example, factors such as time of day, temperature, humidity, wind speed, wind direction, equipment type, altitude, and air pressure can all affect wind turbine power output. By incorporating more of these parameters, the model can more accurately fit the data, thereby improving forecast accuracy.

[0076] However, not all power plants are able to collect all of the parameters that affect power. Some power plants may lack data on certain parameters due to equipment limitations, cost issues, or other reasons.

[0077] In this situation, insisting on completeness for all parameters can lead to a significant reduction in sample data. Therefore, in this embodiment, a trade-off needs to be struck between pursuing parameter completeness and ensuring a sufficient number of samples during data collection and processing. Parameters can be flexibly set based on actual forecast requirements and the actual conditions of the wind farm to ensure sufficient sample data.

[0078] By flexibly setting parameters as described above, we can expand the range of sample data available, avoiding situations where overly stringent data screening criteria lead to insufficient available data. This helps ensure that the predictive model has sufficient data support, thereby finding the optimal balance between accuracy and data coverage.

[0079] It is understandable that after screening, multiple power data with data integrity belonging to different electric fields are obtained, and a complete power database can be further constructed based on these power data.

[0080] It should also be noted that after the above steps, the data is filtered according to the parameter completeness rules, and any data with missing parameters is marked. The data is then organized and packaged according to the electric field as a unit to form a data file to be supplemented. This data file is then sent to the electric field where the sub-control data platform is deployed. These data files are then sent to the sub-control data platform of the relevant electric field. The electric field can then complete the data file by supplementing the missing data.

[0081] Once the electric field has completed data supplementation, the supplemented data can be uploaded through the sub-control data platform. The uploaded complete power data will be transferred to the "complete power database", thereby improving the integrity and accuracy of the overall data set.

[0082] In Step 2.2 above, a rigorous screening and tagging mechanism is implemented to ensure the high integrity of the data included in the "Complete Power Database," helping to improve the accuracy of the prediction model. Furthermore, the distributed control data platform is not only used for local data processing and model training but also plays a vital role in data supplementation and upload, enabling data updates and improvements to be performed locally.

[0083] Step 3: Divide the data in the power database into multiple levels of power data sets based on different types of power impact parameters. The project classification of the low-level power data sets is more detailed than that of the high-level power data sets.

[0084] Specifically, through the dataset architecture module, a multi-level data set is set for all data in the complete power database. The first-level set is all the complete power data, and the second and third levels set more levels of sample sets based on the weights of power impact parameter items at different granularities.

[0085] In an exemplary embodiment, a multi-level architecture design example of a data set is as follows:

[0086] 1.1 Initial Definition of the Dataset

[0087] Level 1: Complete power data set

[0088] This level contains all raw power data, regardless of the specific influencing factors.

[0089] 1.2 Second level: classification based on main parameters affecting power

[0090] The parameters that mainly affect power (such as wind speed, wind direction, wind stability, etc.) are selected to further divide the data set.

[0091] Example:

[0092] Set A1: Wind speed > a certain threshold

[0093] Set A2: Wind speed <= a certain threshold

[0094] Set B1: Wind direction is within a specific range (e.g. 0°-90°)

[0095] Set B2: Wind direction is within other ranges (e.g. 90°-180°)

[0096] 1.3 The third level: more detailed project classification

[0097] Based on the second level, a more detailed classification is carried out according to secondary power influencing parameters (such as wind stability, ambient temperature, etc.).

[0098] Example:

[0099] Set A1a: Wind speed > a certain threshold and high wind stability

[0100] Set A1b: Wind speed > a certain threshold and low wind stability

[0101] Set B1a: Wind direction is between 0° and 90°, and ambient temperature is greater than a certain value

[0102] Set B1b: Wind direction is between 0° and 90°, and ambient temperature is less than or equal to a certain value

[0103] Through the above steps, in the complete power database, the first-level data set contains all complete power data; this set is the most extensive and includes all data that meet the parameter completeness rules.

[0104] Starting at the second level, the data set is partitioned based on the weights of the parameters that influence power generation. "Weight" here refers to the degree to which each parameter affects power generation. During initial training, these weights can be set based on theoretical knowledge. For example, wind speed might be considered the most important parameter affecting power generation and therefore be used as a key parameter to partition the data set.

[0105] In this embodiment, the hierarchical division is based on the weights of different parameters. For example, key parameters such as wind speed, wind direction, and wind stability are used to further subdivide the data set. As the hierarchy decreases (such as the third level, the fourth level, etc.), the data set becomes increasingly specific. For example, in the lower level data sets, it is possible to classify according to more detailed wind speed ranges or wind direction angles.

[0106] It should be noted that each electric field may have different requirements and data characteristics, so the design of these hierarchical sets is flexible and can be adjusted and optimized according to actual forecasting needs. This ensures that the system can still work efficiently under different conditions and data types.

[0107] Step 4: Set classification labels for each level of power data set and send them to the sub-control platform. The classification labels include the various parameters used for classification and the corresponding parameter ranges.

[0108] Specifically, detailed labels are set for the sample sets at each level. These labels include various parameters used for division and the corresponding parameter ranges. For example, a set label may describe the data of the set within a certain wind speed range, as well as the corresponding ranges of parameters such as wind direction and air pressure.

[0109] As you can see, these labels not only help identify the characteristics of the dataset, but also provide a reference for subsequent data selection, model training, and prediction. Through labels, you can quickly locate and call the dataset that best suits your current prediction needs.

[0110] In one exemplary embodiment:

[0111] Parameter name: such as wind speed, wind direction, wind stability, ambient temperature, etc.

[0112] Parameter range: corresponding parameter range or conditions (for example, wind speed > a certain value, wind direction is within a certain range).

[0113] Collection description: The descriptive information of the corresponding collection, explaining the characteristics or categories of the collection.

[0114] Among them, data sets that reach the training quantity threshold can be labeled and sent to the sub-control data platform; considering that model training requires a sufficient amount of data, data sets that do not reach the set training quantity threshold have no training value, and therefore will not be sent to the sub-control data platform; at the same time, there are no more detailed data sets under the data sets that do not reach the set training data threshold.

[0115] In addition, in this embodiment, similarity calculation or graphical analysis is performed on all power data sets under the target level to determine whether there is abnormal data in the power data set that does not conform to the overall data rules. If so, the target power data set containing abnormal data and the upper-level power data set of the target power data set are deleted from the power database, and the data source of the abnormal data is obtained. According to the data source, the abnormal data is sent to the sub-control platform of the corresponding wind farm.

[0116] Through the above steps, the overall power data set is divided into multiple hierarchical data sets according to different levels of parameters. By building a multi-level data set, the system can better organize and utilize data. This structured hierarchical system allows different influencing factors to be considered separately, improving the efficiency of data management and facilitating the accuracy of subsequent model training. In addition, the flexible setting of the hierarchy based on the weights of different power-influencing parameters ensures that the system can adapt to different electric field conditions and actual needs, thereby optimizing the performance of the power prediction model.

[0117] S103, in response to an interactive instruction with any sub-control platform, returning a specific finished product power prediction model matching the wind farm according to the specific conditions of the wind farm contained in the interactive instruction;

[0118] Prior to step S103, the master control platform, through the training module, allocated data to the first-level set and the second-level set, and subsequently trained multiple blank power prediction models. Each data set (i.e., the first-level set and multiple second-level sets) generated a corresponding finished prediction model. Therefore, the master control training module trained at least multiple models and stored them in the system.

[0119] In specific cases, if there are more layers of data sets and the master control module performs model training for each layer, the number of trained models may be even greater. In general, the number of trained models depends on how many different layers of data sets are used for training, that is, the master control platform aggregates the data models of the entire candidate wind farm.

[0120] S104, the sub-control platform deploys a specific finished product power model in combination with local business logic, and performs power prediction based on local power operation and maintenance data through the specific finished product power prediction model.

[0121] It can be understood that in the above step S103, after data acquisition and training, the data of all wind farms in the entire system are obtained, and further based on the layered data set, a variety of different types of models are trained; these models are uniformly stored in the master control platform.

[0122] In this step, each sub-control platform obtains a suitable model from the master control platform based on its own characteristics and deploys the model for local actual power prediction;

[0123] Specifically, the sub-control platform issues an interactive command containing the specific conditions of the wind farm. These conditions may include information such as the wind farm's geographic location, equipment type, and meteorological conditions. Furthermore, in response to this command, the master control platform selects the most suitable off-the-shelf power forecast model for the wind farm based on the specific conditions. This process may involve screening multiple off-the-shelf models to find the one that best matches the actual conditions of the wind farm. The selection criteria may include parameter ranges, similarity of historical data, and compatibility of equipment types. The fundamental goal is to ensure that the selected model can be effectively applied to power forecasting for that specific wind farm.

[0124] Furthermore, the sub-control platform receives the matching finished power prediction model and deploys it in conjunction with localized business logic. Localized business logic refers to the operational procedures and data processing methods specific to the specific conditions of the wind farm. On the sub-control platform, the model is integrated into existing systems to meet the unique needs of the wind farm.

[0125] As you can see, once the model is deployed, the sub-control platform will use this specific finished product power prediction model to generate power forecasts based on local, real-time power operation and maintenance data. As the model continues to run, its internal parameters will adjust to the characteristics of local data, gradually improving the accuracy and reliability of the forecasts.

[0126] In addition, the latest actual power data predicted by the sub-control platform will also be uploaded to the master control platform. The master control platform will re-execute the above step S101 based on the latest actual power data, and then continue to optimize and train the centrally stored model.

[0127] At the same time, this application is also equipped with a data update function. With the development of power generation technology and the update of electric field power generation equipment, some equipment is eliminated. At this time, the corresponding power generation data will not be used for subsequent actual data reference. Therefore, the master control data platform will update the data set of this part through the data update module.

[0128] For some devices, their power performance varies with different years of use. Therefore, even if the hardware parameters of each device remain unchanged, the data power data will be different. Therefore, in the corresponding data set, if there is sufficient data in the latest stage, the data with an older time will be deleted and the data set will be updated.

[0129] After the master control platform's data is updated, the data sets of the sub-control platforms are also updated synchronously. During the idle period of the sub-control platform's deployed equipment, the blank prediction model is trained. After the training is completed, the previous prediction model is replaced. This ensures that the running prediction model is trained with the latest data and its prediction results are the most accurate.

[0130] Through the above steps S101 to S104, compared with the existing technology, the problem of insufficient historical data on the electric field side making model training impossible is solved. The present application solution uses the master control platform to perform layered training based on the data of all candidate electric fields to obtain multiple labeled data sets, and then trains each data set separately to obtain multiple models. Therefore, the master control platform can provide each wind farm with models of other electric fields with similar conditions for power forecasting; and continue to optimize based on the actual data of the electric field during the operation and development of the electric field, establishing a sustainable optimization model unique to each electric field. This greatly improves the versatility and resource cost of centralized power forecasting.

[0131] In another embodiment, the site-side model is a model deployed locally at each wind farm to cope with network instability or disconnection. If the site's network connection with the group is interrupted, the site-side model can run independently. If the site has sufficient computing resources, it can also obtain data similar to the local environment from the master control platform through the sub-control platform and train a model suitable for local deployment. The specific steps include the following:

[0132] Step 1: The sub-control platform obtains a specific power data set from the master control platform based on its own equipment parameters and local climate parameters;

[0133] Among them, the sub-control data platform will select the appropriate data set for model training through data set labels based on the actual situation of the wind farm, such as equipment type, climatic conditions, etc.; among them, the data set label is a description of the data set, indicating the main characteristics of the data in the set, such as wind speed, wind direction, temperature, humidity and other parameter ranges.

[0134] In addition, the distributed control data platform can automatically select the most appropriate dataset by calculating parameter fit. Parameter fit refers to the degree of match between the actual conditions of the wind farm and different datasets. An algorithm automatically evaluates the match between the current environment and the labels of each dataset, recommending or automatically selecting the dataset that best suits the current situation.

[0135] Different data sets may belong to different hierarchies. For example, one set may be a subset or lower-level set of a larger set. Lower-level sets contain more specific and detailed categorized data, while higher-level sets may contain more general data. When selecting a data set, you can choose a single-level set or combine data sets from multiple levels for comprehensive training to improve the accuracy of the predictive model.

[0136] Step 2: Based on the specific power data set and local power-related data, a power prediction blank model is trained to obtain a local power prediction model, and the local power prediction model is deployed in the actual power prediction task of the wind farm.

[0137] The sub-control data platform first performs model training based on the multiple specific power data sets selected in Step 1. These data sets may contain historical data at different levels or under specific conditions, such as a specific wind speed range, climate conditions, equipment status, etc.

[0138] Furthermore, the selected data sets are divided into training and test sets. Typically, the data sets are divided into two parts: one for training the model (training set) and the other for validating the model's performance (test set). The training set is used to build the model, allowing it to learn the relationship between different parameters and power generation; the test set is used to evaluate the model's accuracy and robustness, ensuring that the model can make accurate predictions even on unseen data.

[0139] During the training process, the training set data is used to train one or more blank prediction models. These blank prediction models (blank models) are initially untrained models. During the training process, these models gradually improve their predictive capabilities by learning from historical data. Each data set (or combination thereof) produces a separate blank prediction model. The training process involves multiple iterations, in which model parameters are adjusted to minimize prediction error, thereby improving model accuracy.

[0140] Finally, all the finished prediction models on the sub-control data platform are deployed to the centralized power prediction of the actual electric field. The current forecast meteorological data and known data are input. Unknown data is supplemented by the average data of the same period. The finished prediction model that meets the data set label is selected for prediction calculation. Generally, multiple prediction models are run simultaneously here to obtain multiple prediction data to provide multi-angle data reference. During this process, the prediction data is recorded simultaneously with the subsequent actual power data.

[0141] It should be noted that the steps shown in the above process or the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0142] This application also provides a gas turbine state recognition system based on a multi-scale atrous variational graph convolutional network. Figure 2 This is a structural block diagram of a wind farm centralized power prediction system based on big data according to an embodiment of the present application, such as Figure 2 As shown, the system includes: a master control platform 20 and a sub-control platform 21, wherein:

[0143] The master control platform 20 is used to obtain power operation and maintenance data of candidate wind farms, and use big data technology to filter and pre-process the power operation and maintenance data to obtain power data sets at multiple levels;

[0144] Furthermore, based on data sets at multiple levels, blank models are trained separately to obtain multiple finished power prediction models corresponding to different levels and types. In response to interactive instructions with any sub-control platform, a specific finished power prediction model matching the wind farm is returned based on the specific conditions of the wind farm contained in the interactive instructions.

[0145] The sub-control platform 21 is used to deploy a specific finished product power prediction model in combination with local business logic, and to perform power prediction based on local power operation and maintenance data through the specific finished product power prediction model.

[0146] The above system solves the problem of insufficient historical data on the electric field side, making it impossible to implement model training. This application solution uses the master control platform to perform layered training based on the data of all candidate electric fields to obtain multiple labeled data sets, and then trains each data set separately to obtain multiple models. Therefore, the master control platform can provide each wind farm with models of other electric fields with similar conditions for power forecasting; and continue to optimize according to the actual data of the electric field during the operation and development of the electric field, establishing a sustainable optimization model unique to each electric field. This greatly improves the versatility and resource cost of centralized power forecasting.

[0147] In one embodiment, Figure 3 is a schematic diagram of the internal structure of an electronic device according to an embodiment of the present application, such as Figure 3 As shown, an electronic device is provided, which may be a server, and its internal structure diagram may be as shown in FIG. Figure 3 The electronic device includes a processor, memory, a network interface, and a database connected via a system bus. The processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the electronic device is used to store data. The network interface of the electronic device is used to communicate with an external terminal via a network connection. The computer program is executed by the processor to implement a wind farm centralized power prediction method based on big data.

[0148] Those skilled in the art will understand that Figure 3The structure shown in the figure is merely a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the electronic device to which the solution of the present application is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0149] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, which can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0150] Those skilled in the art should understand that the various technical features of the above-described embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the various technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0151] In addition, in conjunction with the big data-based centralized wind farm power prediction method in the above-mentioned embodiments, embodiments of the present application may provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when executed by a processor, the computer program instructions implement any of the big data-based centralized wind farm power prediction methods in the above-mentioned embodiments.

[0152] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0153] The above embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for predicting concentrated power of wind farms based on big data, characterized in that: Centralized power forecasting is achieved through a master control platform and sub-control platforms located in each wind farm. The method includes: The master control platform obtains power operation and maintenance data of candidate wind farms and uses big data technology to filter and preprocess the power operation and maintenance data to obtain power data sets at multiple levels; Based on the data sets of the multiple levels, blank models are trained respectively to obtain multiple finished product power prediction models corresponding to different levels and different types; In response to an interactive instruction with any sub-control platform, based on the specific conditions of the wind farm contained in the interactive instruction, a specific finished product power prediction model matching the wind farm is returned; The sub-control platform deploys the specific finished product power prediction model in combination with local business logic, and performs power prediction based on local power operation and maintenance data through the specific finished product power prediction model.

2. The method according to claim 1, characterized in that By using big data technology to filter and pre-process the power operation and maintenance data, we can obtain power data sets at multiple levels, including: Using big data technology to filter the power operation and maintenance data to obtain data related to power generation, and building a power database based on the data related to power generation; According to different types of power impact parameters, the data in the power database is divided into multiple levels of power data sets, wherein the item classification of the low-level power data sets is more detailed than that of the high-level power data sets; A classification label is set for each level of power data set and sent to the sub-control platform. The classification label includes various parameters used for classification and corresponding parameter ranges.

3. The method according to claim 2, characterized in that By using big data technology to filter the power operation and maintenance data, the data related to power generation are as follows: Acquiring data related to generated power from the power operation and maintenance data to obtain initial power-related data, wherein the initial power-related data includes: wind farm geographic data, climate data, power generation-related equipment data, and historical generated power; Based on parameter integrity rules, the initial power-related data are screened to obtain a plurality of power data with data integrity, and the power database is constructed based on the plurality of power data.

4. The method according to claim 3, characterized in that After screening the initial power-related data based on parameter integrity rules, the method further includes: Obtain data with missing parameters, and package the missing data corresponding to each wind farm to generate data to be supplemented; The data to be supplemented is sent to the corresponding wind farm sub-control platform to instruct the sub-control platform to fill in the missing data to obtain complete supplemented power data, and return the supplemented power data.

5. The method according to claim 2, characterized in that The method further comprises: When the amount of historical power operation and maintenance data of any target wind farm is greater than a preset threshold, a dedicated database is set up for the target wind farm, and the dedicated database is divided into multiple levels of power data sets according to the operation and maintenance information of the target wind farm.

6. The method according to claim 2, characterized in that The method further comprises: By performing similarity calculation or graphical analysis on all power data sets under the target level, it is determined whether there are abnormal data in the power data sets that do not conform to the overall data rules. If yes, in the power database, the target power data set with abnormal data and the upper power data set of the target power data set are deleted. Furthermore, a data source of the abnormal data is obtained, and the abnormal data is sent to a sub-control platform of a corresponding wind farm according to the data source.

7. The method according to claim 1, characterized in that After using big data technology to filter and preprocess the power operation and maintenance data to obtain multiple levels of power data sets, the method further includes: The sub-control platform obtains a specific power data set from the master control platform based on its own equipment parameters and local climate parameters; Based on the specific power data set and local power related data, the blank model is trained to obtain a local end power prediction model, and the local end power prediction model is deployed in the actual power prediction task of the wind farm.

8. A wind farm centralized power prediction system based on big data, characterized in that: Centralized power forecasting is achieved through a master control platform and sub-control platforms located in each wind farm. The system includes the master control platform and the sub-control platforms, wherein: The master control platform is used to obtain power operation and maintenance data of candidate wind farms, and use big data technology to filter and preprocess the power operation and maintenance data to obtain power data sets at multiple levels; Furthermore, based on the data sets of the multiple levels, blank models are trained respectively to obtain multiple finished product power prediction models corresponding to different levels and types, and in response to an interactive instruction with any sub-control platform, a specific finished product power prediction model matching the wind farm is returned according to the specific conditions of the wind farm contained in the interactive instruction; The sub-control platform is used to deploy the specific finished product power prediction model in combination with local business logic, and perform power prediction based on local power operation and maintenance data through the specific finished product power prediction model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images