New energy power prediction method, system, device and medium

Abstract

The invention relates to the technical field of new energy power prediction, and discloses a new energy power prediction method, system and device and a medium, and the method comprises the steps: carrying out the fine preprocessing and feature construction of historical operation data and comprehensive external data, fully mining the multi-dimensional incidence relation between data, constructing a standard and unified third feature set, and obtaining a new energy power prediction result; high-quality input is provided for a prediction model, and the accuracy and reliability of prediction are improved. According to unique power generation characteristics of different types of new energy power stations, accurate and personalized prediction can be carried out. In the face of complex and changeable external meteorological conditions and dynamic change of the operation state of power generation equipment, the actual change condition of the new energy power can be captured and reflected timely and accurately.

Description

Technical Field

[0001] This invention relates to the field of new energy power prediction technology, and in particular to a new energy power prediction method, system, device and medium. Background Technology

[0002] The inherent intermittency, volatility, and randomness of renewable energy generation pose unprecedented challenges to the safe and stable operation of the power grid. Accurate renewable energy power forecasting is crucial to effectively reduce the negative impacts of renewable energy integration on the grid system and significantly improve the grid's capacity to absorb renewable energy. However, many problems remain to be solved in the field of renewable energy power forecasting. On the one hand, data processing and feature construction are still relatively rudimentary, failing to fully explore and utilize the deep-seated potential correlations between data, directly resulting in the accuracy and reliability of forecasting models falling short of ideal levels. On the other hand, different types of renewable energy power plants, such as wind power plants and solar power plants, each possess unique power generation characteristics, and existing forecasting methods often struggle to provide accurate and personalized forecasts for these different types of power plants.

[0003] In practical applications, with the continuous and rapid growth of installed capacity of new energy sources, the power grid's requirements for the accuracy and real-time performance of new energy power forecasting are constantly increasing. Existing forecasting methods often struggle to capture and reflect the actual changes in new energy power in a timely and accurate manner when faced with complex and ever-changing external weather conditions and dynamic changes in the operating status of power generation equipment. This, to some extent, restricts the efficient utilization of new energy sources and the stable operation of the power grid. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a new energy power prediction method, system, device and medium that can solve the problems of crude data processing and feature construction, difficulty in accurate prediction for different types of power plants and difficulty in adapting to complex external conditions and equipment status changes in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for predicting the power output of new energy sources, comprising:

[0008] Acquire historical operating data of new energy power plants and comprehensive external data, and perform a first preprocessing on the historical operating data and comprehensive external data;

[0009] A second feature construction operation is performed on the preprocessed historical running data and the integrated external data to obtain the constructed third feature set;

[0010] The third feature set is subjected to model training operations, which are used to train a new energy power prediction model.

[0011] A fourth prediction calculation operation is performed based on the new energy power prediction model. The fourth prediction calculation operation is used to generate the predicted value of new energy active power for future periods.

[0012] The predicted active power of the new energy source is fused using a fifth result fusion operation, and the fused prediction result is used to predict the power of the new energy source.

[0013] The new energy power prediction is used to provide decision support for power grid dispatch and operation.

[0014] As a preferred embodiment of the new energy power prediction method described in this invention, the second feature construction operation includes:

[0015] Establish multi-dimensional correlations between historical operational data and comprehensive external data, and construct features based on these correlations;

[0016] The multi-dimensional relationships include at least one or more of the following relationships:

[0017] Power-meteorological correlation, spatial correlation, time-series autocorrelation, equipment operating status correlation, and historical power pattern matching.

[0018] As a preferred embodiment of the new energy power prediction method of the present invention, the second feature construction operation further includes:

[0019] The historical operational data is matched using field matching based on the power plant number and timestamp.

[0020] The integrated external data is initially aligned with the historical operational data;

[0021] Establish a candidate feature set based on the initial alignment;

[0022] Obtain a set of candidate features that satisfy multi-dimensional correlations.

[0023] As a preferred embodiment of the new energy power prediction method of the present invention, the second feature construction operation further includes:

[0024] Integrate the candidate feature set that satisfies multi-dimensional correlation;

[0025] Construct a feature dataset with a unified field structure as the third feature set;

[0026] The unified field structure includes feature name, data source, data type, and feature description.

[0027] This preferred solution makes feature data more standardized and organized, facilitating subsequent processing and analysis. By integrating a set of candidate features that satisfy multi-dimensional relationships, it avoids data fragmentation and chaos, improving data usability. A unified field structure clarifies the key information of each feature; feature names allow users to quickly identify the meaning of features; data source information helps trace the origin of data, ensuring its reliability and verifiability; and data type information provides an important basis for subsequent data analysis, as different data types may require different analysis methods. Feature descriptions further explain the specific meaning and purpose of the features, enabling data users to understand the data more deeply.

[0028] This constructed third feature set can provide higher-quality input data for new energy power prediction models. A standardized feature dataset enables the model to better learn and capture patterns and features in the data, thereby improving the accuracy and stability of predictions. Furthermore, a unified field structure facilitates data sharing and communication between different departments or personnel, promoting collaborative work in new energy power prediction. In practical applications, when it is necessary to optimize or adjust the prediction model, the standardized third feature set also makes data filtering and feature selection more convenient, improving work efficiency and effectiveness.

[0029] As a preferred embodiment of the new energy power prediction method described in this invention, the model training operation includes:

[0030] The input characteristics and output target variables of the pre-defined new energy power prediction model are defined.

[0031] The input features are features in the third feature set, and the output target variable is the active power of new energy sources in the future time period;

[0032] Model training is performed based on the input features and the output target variable.

[0033] As a preferred embodiment of the new energy power prediction method of the present invention, the fourth prediction calculation operation based on the new energy power prediction model includes:

[0034] Obtain comprehensive external data and the latest historical operational data for the period to be predicted;

[0035] Perform the same preprocessing and feature construction operations on the acquired data as in the training phase;

[0036] The constructed feature data is input into the trained new energy power prediction model to obtain the predicted value of new energy active power.

[0037] As a preferred embodiment of the new energy power prediction method described in this invention, the historical operating data and comprehensive external data include:

[0038] The historical operating data includes historical active power curves of new energy power plants, equipment operating status, and geographical location information synchronized with the OCS system.

[0039] The integrated external data includes real-time meteorological data, numerical weather forecast data, and GIS data.

[0040] Secondly, the present invention provides a new energy power prediction system, comprising:

[0041] The data acquisition and processing module is used to acquire historical operating data of new energy power plants and comprehensive external data, and to perform a first preprocessing on the historical operating data and comprehensive external data.

[0042] The feature set acquisition module is used to perform a second feature construction operation on the preprocessed historical running data and integrated external data to obtain the constructed third feature set.

[0043] The prediction model building module is used to perform model training operations on the third feature set, and the model training operations are used to train the new energy power prediction model.

[0044] The prediction module is used to perform a fourth prediction calculation operation based on the new energy power prediction model. The fourth prediction calculation operation is used to generate the predicted value of new energy active power for future periods.

[0045] The fusion module is used to perform a fifth result fusion operation on the predicted value of the active power of the new energy source, and use the fused prediction result to predict the power of the new energy source.

[0046] The new energy power prediction is used to provide decision support for power grid dispatch and operation.

[0047] Thirdly, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0048] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0049] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention proposes a new energy power prediction method. This method, through meticulous preprocessing and feature construction of historical operating data and comprehensive external data, fully explores the multi-dimensional correlations between data, constructs a standardized and unified third feature set, provides high-quality input for the prediction model, and improves the accuracy and reliability of prediction. It can perform accurate and personalized predictions based on the unique power generation characteristics of different types of new energy power plants. When facing complex and ever-changing external meteorological conditions and dynamic changes in the operating status of power generation equipment, it can timely and accurately capture and reflect the actual changes in new energy power. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart of a new energy power prediction method provided in one embodiment of the present invention.

[0052] Figure 2 This is an internal structure diagram of an electronic device for a new energy power prediction method provided in one embodiment of the present invention. Detailed Implementation

[0053] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0054] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for predicting new energy power, including:

[0055] Existing technologies suffer from several problems, such as insufficient precision in data processing and feature construction, making it difficult to uncover deep correlations between data points, resulting in poor accuracy and reliability of prediction models. Furthermore, existing prediction methods cannot accurately and individually predict the unique power generation characteristics of different types of renewable energy power plants. They also struggle to capture and reflect the actual changes in renewable energy power output in a timely and accurate manner when faced with complex and variable external weather conditions and dynamic changes in the operating status of power generation equipment.

[0056] This invention provides a method that can effectively solve the problems mentioned above. The following will describe in detail how to implement this new energy power prediction method with reference to several embodiments.

[0057] Figure 1 A flowchart of a new energy power prediction method is shown, including:

[0058] S101, acquire historical operating data and comprehensive external data of the new energy power plant, and perform the first preprocessing on the historical operating data and comprehensive external data, including:

[0059] It should be noted that in order to predict the power output of new energy sources, it is necessary to obtain relevant data from new energy power plants so that these data can be analyzed and processed in depth.

[0060] In some specific implementations, the relevant data for new energy power plants may include equipment parameters such as the rated power and efficiency of the generators. These parameters reflect the basic performance of the power generation equipment and are of significant reference value for power prediction. Maintenance records may also be included, as the maintenance status of the equipment affects its operating efficiency and stability, thus impacting power generation. For example, regularly maintained and well-maintained equipment may generate more stable power with smaller power fluctuations.

[0061] In some specific implementations, market demand data can also be used as relevant data, and the output power of new energy power generation needs to match market demand. Understanding the market demand trend for new energy power can better predict the power output that power plants should provide. Historical meteorological data is also crucial. In addition to current meteorological data and numerical weather forecast data, meteorological data over a period of time can help analyze the patterns and trends of meteorological changes, providing more comprehensive meteorological information support for power forecasting.

[0062] In some specific implementations, relevant data from renewable energy power plants can be collected through various means. Firstly, it can be obtained from the power plant's own monitoring system, which records real-time operating parameters of the power generation equipment, such as power output, equipment temperature, voltage, and current. This data directly reflects the actual operating status of the power plant. Secondly, meteorological data is crucial for forecasting renewable energy power output. Comprehensive external meteorological data, such as sunlight intensity, wind speed, wind direction, temperature, and humidity, can be obtained through professional meteorological monitoring stations, meteorological satellites, or meteorological service platforms. These meteorological factors are closely related to renewable energy power generation, especially solar and wind power generation.

[0063] In some specific implementations, data related to the power plant's connection to the grid, such as grid load and power quality, can also be obtained from the power grid company. This helps in understanding the power plant's operational status within the overall grid. Simultaneously, equipment technical parameters and maintenance records provided by equipment manufacturers are also important data sources, helping to predict equipment performance changes and potential failures. By integrating and analyzing this multi-source data, more comprehensive and accurate foundational data can be provided for the new energy power prediction system, thereby improving the accuracy and reliability of predictions.

[0064] It should be noted that after obtaining historical operational data and comprehensive external data, considering that these data may contain noise, missing values, and inconsistent formats, which may adversely affect subsequent data processing and model training, it is necessary to perform the first preprocessing.

[0065] In some specific implementations, the first preprocessing can employ various methods. For noise in the data, filtering algorithms such as median filtering and mean filtering can be used to remove abnormal fluctuations, making the data smoother and reducing noise interference with power prediction. For missing values, appropriate processing methods are selected based on the characteristics and extent of the missing values. If there are few missing values, interpolation methods such as linear interpolation and spline interpolation can be used to estimate the missing values ​​based on the values ​​of adjacent data points. If there are many missing values, deleting records containing missing values ​​can be considered, but care must be taken not to delete excessively, which could reduce the data volume and affect model training. Regarding format standardization, data from different sources and in different formats are converted into a unified data format, such as a unified timestamp format and data type, to facilitate subsequent data processing and analysis.

[0066] It should be noted that after the first preprocessing, the historical operating data and the comprehensive external data will be cleaner, more accurate, and more standardized, laying a solid foundation for the subsequent second feature construction operation. The second feature construction operation aims to further mine the potential information in the data. By establishing multi-dimensional correlations between historical operating data and comprehensive external data, a more representative and discriminative feature set can be constructed, thereby improving the performance and prediction accuracy of the new energy power prediction model.

[0067] In this embodiment of the invention, historical operating data of the new energy power plant and comprehensive external data are obtained. The historical operating data and comprehensive external data include:

[0068] Historical operating data includes historical active power curves of new energy power plants, equipment operating status, and geographical location information synchronized with the OCS system;

[0069] The integrated external data includes real-time meteorological data, numerical weather forecast data, and GIS data.

[0070] Specifically, this invention simultaneously acquires two types of core data: historical operating data (denoted as H) originates from the power grid dispatching system (OCS), which includes the historical active power P of each new energy power station. t Equipment operating status S t The power station's unique ID, precise timestamp t, and geographical location information (longitude Lon, latitude Lat).

[0071] Comprehensive external data (denoted as E) includes real-time meteorological observation data (such as real-time wind speed V). t Temperature T t Light intensity I t Numerical weather forecasts (NWP) provide weather predictions for the next k hours (such as...). ), and the topographic elevation H provided by the Geographic Information System (GIS) elev and surface roughness R surf .

[0072] Furthermore, the first preprocessing often results in issues such as missing data, noise, and time misalignment. Therefore, uniform preprocessing must be performed to ensure the consistency of data quality.

[0073] The linear imputation of missing values ​​is performed using the following formula:

[0074]

[0075] Where, x t Let x represent any observed variable at time t, such as wind speed, power, temperature, etc. t ′ represents the value after padding;

[0076] It should be noted that when data is lost at a certain moment, the average of the two adjacent moments is used for interpolation, which is suitable for short-term breakpoints. This operation ensures the continuity of the time series and avoids feature construction failure due to missing individual data points.

[0077] Furthermore, outlier identification and correction are performed using the following formula:

[0078] Keep x t If and only if |x t -μ|≤3σ

[0079] Where μ is the mean of the variable within a sliding window (e.g., the past 24 hours); σ is the standard deviation within the same window;

[0080] It should be noted that if a point deviates from the mean by more than three standard deviations, it is considered an outlier and replaced with an interpolated result. This strategy effectively filters out "outliers" caused by sensor malfunctions or communication errors, improving data reliability.

[0081] Furthermore, cross-source data time alignment can be performed using the following formula:

[0082] D clean =align(H,E,key:ID,t)

[0083] Among them, D clean It is a cleaned and aligned unified dataset; align(·) indicates an operation to perform exact matching based on the power plant ID and timestamp t;

[0084] It should be noted that data from different sources are fully aligned in both "space" (which power plant) and "time" (which moment). At this point, the invention completes the transformation from "messy raw data" to a "structured, clean, and spatiotemporally consistent" initial dataset, preparing it for the next step of feature engineering.

[0085] S102, perform a second feature construction operation on the preprocessed historical running data and integrated external data to obtain the constructed third feature set, where:

[0086] It should be noted that after obtaining the historical operating data and comprehensive external data that have undergone the first preprocessing, a second feature construction operation is needed to further explore the potential information in the data, thereby obtaining a more representative and discriminative third feature set. This process will fully consider the multi-dimensional correlation between historical operating data and comprehensive external data to comprehensively reflect the characteristics and influencing factors of new energy power generation.

[0087] In some specific implementations, the second feature construction operation can be carried out from multiple aspects. For the historical active power curve of new energy power plants in historical operating data, the periodic characteristics of its power change can be analyzed, such as the daily, weekly or monthly power fluctuation patterns, and the power mean, variance, peak value and other statistics can be calculated. These statistics can reflect the stability and variation range of the power plant's power.

[0088] In some specific implementations, features such as normal operating time and failure frequency can be constructed based on the equipment's operating status to understand the impact of equipment reliability on power generation. Regarding geographic location information, differences in meteorological resources such as sunlight intensity and wind speed at different geographical locations can be considered and correlated with historical active power to construct regionally specific features.

[0089] In some specific implementations, meteorological characteristics can be combined and derived from real-time meteorological data and numerical weather forecast data in integrated external data. For example, wind speed and solar radiation intensity can be weighted and combined to construct a comprehensive meteorological energy index, which can more comprehensively reflect the impact of meteorological conditions on new energy power generation.

[0090] In some specific implementations, the rate of change of meteorological parameters, such as the rate of change of wind speed and the rate of change of temperature, can also be calculated based on numerical weather forecast data. These rates of change can reflect the dynamic changes of meteorological conditions in advance and are of great significance for predicting future power generation.

[0091] In an embodiment of the present invention, the second feature construction operation includes:

[0092] Establish multi-dimensional correlations between historical operational data and comprehensive external data, and construct features based on these correlations;

[0093] Multidimensional relationships include at least one or more of the following:

[0094] Power-meteorological correlation, spatial correlation, time-series autocorrelation, equipment operating status correlation, and historical power pattern matching.

[0095] In this embodiment of the invention, the second feature construction operation further includes:

[0096] Historical operational data is matched using fields such as power plant number and timestamp.

[0097] Initial alignment of external data with historical operational data;

[0098] Establish a candidate feature set based on the initial alignment;

[0099] Obtain a set of candidate features that satisfy multi-dimensional correlations.

[0100] In this embodiment of the invention, the second feature construction operation further includes:

[0101] Integrate the candidate feature set that satisfies multi-dimensional correlation;

[0102] Construct a feature dataset with a unified field structure as the third feature set;

[0103] The unified field structure includes feature name, data source, data type, and feature description.

[0104] Specifically, the present invention will use the D obtained in the previous stage clean Field expansion is performed using ID and t as primary keys to generate a preliminary feature candidate pool F. raw For example, the power, wind speed, temperature, altitude, etc. of a wind farm at a certain moment can be combined into a single record.

[0105] Furthermore, this invention constructs explanatory high-level features around five key relationships.

[0106] Furthermore, regarding the power-meteorological correlation characteristics, the output of new energy sources is directly driven by weather conditions. This invention constructs features reflecting this physical relationship. The power-meteorological correlation characteristics are represented by the wind turbine efficiency ratio:

[0107]

[0108] Among them, f pm P represents the actual power generation efficiency per unit of wind energy. t V represents the actual active power at the current moment. t Indicates the wind speed at the current moment;

[0109] It should be noted that wind energy is proportional to the cube of wind speed, so this ratio can reflect whether the wind turbine is operating efficiently.

[0110] When f pm A significant decrease may indicate blade icing or equipment aging, which serves as an early warning for predictive adjustments.

[0111] Furthermore, considering that adjacent power plants are often affected by similar weather systems and their power output is synergistic, spatially weighted proximity power is used to represent spatial correlation characteristics:

[0112]

[0113] Among them, f sp This represents the weighted average power of N neighboring power plants surrounding the target power plant; d represents the power of the i-th neighboring power station at the current moment; i Indicates the geographical distance between the target station and the i-th neighboring station; w i σ represents the weighting coefficient, which decreases exponentially with increasing distance; σ represents the parameter that controls the rate of decay, and is usually set to around 50km.

[0114] It should be noted that this feature can capture weather movement trends at a regional scale, providing additional information, especially during sudden weather changes.

[0115] Furthermore, considering the significant diurnal periodicity and seasonality of new energy output, the autocorrelation strength is used to represent the time series autocorrelation characteristics:

[0116]

[0117] Among them, f ac This indicates the degree of linear correlation between the current power and the power k hours ago; P t P represents the power at the current moment. t-k This represents the power at k hours ago (e.g., k = 24 means the same time yesterday); This represents the average power over a period of time. It represents the variance of the power sequence; the closer the autocorrelation strength is to 1, the more stable the power output mode.

[0118] It should be noted that this feature helps the model identify whether "today is like yesterday," thus enhancing its ability to model cyclical behavior.

[0119] Furthermore, considering that the equipment is not always operating at full capacity, its status directly affects the maximum output. An equipment efficiency factor is used to represent the correlation characteristics of the equipment's operating status:

[0120]

[0121] Among them, f st This indicates that the textual status code is encoded as a numerical efficiency factor; S t This represents the device's state label at time t; this factor can be directly multiplied into the prediction result to achieve "state-aware" prediction.

[0122] For example, if a dispatch order requires power rationing, the model will automatically lower the forecast even if the weather is good.

[0123] Furthermore, considering that certain weather patterns may repeat, leading to highly similar power output curves, historical curve similarity is used to represent historical power pattern matching characteristics:

[0124]

[0125] Among them, f pt This represents the current 24-hour power vector. Indicates a date in history Cosine similarity; This indicates that the most similar day is retrieved from the historical database; the higher the similarity, the more likely the current weather is to be similar to a historical day, which can be used for extrapolation and prediction.

[0126] Furthermore, all candidate features satisfying multi-dimensional correlations are integrated, redundancy is removed, and a unified naming convention is adopted to ultimately form a clearly structured and semantically explicit third feature set F3. All features are labeled with name, source, type, and meaning; a unified dimension is used; and a feature vector is constructed: F3 = [f pm ,f sp ,f ac ,f st ,f pt ,...];

[0127] It should be noted that this invention successfully achieves a significant leap from "raw data" to "intelligent features." These transformed intelligent features are no longer merely isolated numerical representations, but rather "knowledge units" that deeply integrate multi-dimensional information such as physical laws, spatiotemporal structures, and operational logic. These knowledge units not only more comprehensively reflect the intrinsic nature of the data but also provide richer and more accurate support for subsequent data analysis and applications, thereby greatly enhancing the value and application potential of the data.

[0128] S103, perform model training operations on the third feature set. These operations are used to train the new energy power prediction model, where:

[0129] It should be noted that once the completed third feature set is obtained, the model training operation can begin to train the new energy power prediction model.

[0130] In some specific implementations, various machine learning or deep learning models can be used for model selection. For example, a random forest model can be chosen, which is an ensemble learning method consisting of multiple decision trees. During training, the random forest samples the samples in the training set with replacement to construct different decision trees, and then combines the results of these decision trees to make predictions. This model has high accuracy and stability and can handle high-dimensional data and complex nonlinear relationships.

[0131] In some specific implementations, a Long Short-Term Memory (LSTM) network model can also be selected. LSTM is a special type of recurrent neural network capable of effectively processing time-series data. By using a gating mechanism to control the flow of information, LSTM can capture long-term dependencies in the data, making it well-suited for tasks such as renewable energy power prediction that require consideration of historical data and time factors.

[0132] In this embodiment of the invention, the model training operation includes:

[0133] The input characteristics and output target variables of the pre-defined new energy power prediction model are defined.

[0134] The input features are the features in the third feature set, and the output target variable is the active power of new energy in the future period;

[0135] Model training is performed based on input features and output target variables.

[0136] Specifically, with high-quality features, a powerful predictive model can be trained. The goal of this stage is to teach the model to map from F3 to future power P. t+k This invention formalizes the problem into a supervised learning task, and sets the model learning objective as follows:

[0137]

[0138] in, F3 represents the power prediction value after the k-th hour of the model output; M(·) represents the third feature set of the input; θ represents the prediction model (such as XGBoost, LSTM, Transformer); and θ represents the parameters to be learned by the model. The model's task is to find the optimal θ such that... As close as possible to the true value P t+k .

[0139] Furthermore, we define the loss function to be minimized:

[0140]

[0141] Among them, P t+k The model represents the actual historical power; M(F3; θ) represents the model's predicted value; the summation range T is the entire training period; minimizing the mean squared error (MSE) improves the overall prediction accuracy. During training, a validation set is used to monitor overfitting, and training is stopped in a timely manner (early stopping method). In addition, dedicated models are trained for different types of power plants (wind power or photovoltaic) to achieve personalized modeling and avoid errors caused by a "one-size-fits-all" approach.

[0142] S104, the fourth prediction calculation operation is performed based on the new energy power prediction model. This fourth prediction calculation operation is used to generate the predicted value of new energy active power for future periods, wherein:

[0143] It should be noted that after the new energy power prediction model is trained, it can be used to perform the fourth prediction calculation to generate the predicted value of new energy active power for future periods. When making predictions, the latest historical operating data and comprehensive external data must first be collected. Following the previous preprocessing steps and the second feature construction operation, a corresponding third feature set is generated as the input to the model.

[0144] In some specific implementations, different types of new energy power plants, such as wind power and photovoltaic power plants, use their respective pre-trained dedicated models for prediction. Taking wind power as an example, when the third feature set at the current moment is input, the model will output the predicted power value for the k-th hour in the future based on the learned parameters and mapping relationships. In practical applications, the prediction results may be affected by various factors, such as sudden weather changes and equipment failures. Therefore, it is necessary to monitor and evaluate the prediction results in real time.

[0145] In some specific implementations, a certain threshold can be set, and an early warning mechanism can be activated when the deviation between the predicted and actual values ​​exceeds the threshold. For example, by comparing the power value predicted by the model with the real-time collected power value, if the deviation reaches 10%, it is necessary to further check the accuracy of the model or investigate whether there is any equipment malfunction. At the same time, in order to improve the accuracy and reliability of the prediction, the prediction model can be dynamically adjusted by combining real-time meteorological update data.

[0146] In this embodiment of the invention, the fourth prediction calculation operation based on the new energy power prediction model includes:

[0147] Obtain comprehensive external data and the latest historical operational data for the period to be predicted;

[0148] Perform the same preprocessing and feature construction operations on the acquired data as in the training phase;

[0149] The constructed feature data is input into the trained new energy power prediction model to obtain the predicted value of new energy active power.

[0150] Specifically, after the model training is completed, the actual prediction phase begins. This phase requires strictly replicating the data processing flow from training to ensure input consistency.

[0151] Furthermore, obtain the latest NWP forecast data. pred (Next 168 hours); Obtain the latest historical operation data H latest (Past 72 hours);

[0152] Furthermore, every step of the prediction phase must be completely consistent with that of the training phase. For E pred and H latest Perform the same cleaning, alignment, and missing fill procedures; construct features using the same formula; and obtain a prediction-specific feature set.

[0153] Furthermore, the model inference generates preliminary predictions, and the single-model prediction output is represented as:

[0154]

[0155] Where K represents the prediction duration (e.g., 168); the output is the hourly power prediction sequence for the next K hours; this result is the "preliminary prediction" of a single model. At this point, the invention has obtained the first version of the prediction results, but has not yet reached optimal accuracy.

[0156] S105, performs a fifth result fusion operation on the predicted active power of new energy sources, and uses the fused prediction results to predict the power of new energy sources, wherein:

[0157] It should be noted that after obtaining the preliminary predicted value of active power from new energy sources, the prediction results of a single model may have certain limitations and errors. To further improve the accuracy and reliability of the prediction, a fifth result fusion operation is needed. This invention employs a multi-model fusion method to combine the results of multiple prediction models of different types.

[0158] In some specific implementations, different models such as Random Forest, LSTM, and XGBoost can be selected for training to obtain prediction results from multiple models. These results are then fused using a weighted average, with the weights dynamically adjusted based on the performance of each model on the validation set. For example, models that perform better on the validation set can be assigned higher weights.

[0159] In some specific implementations, a stacked ensemble approach can also be used to fuse results. First, a base model (such as random forest or LSTM) is used for initial predictions. Then, the predictions from these base models are used as input to a meta-model (such as logistic regression) for secondary learning and prediction. This method fully leverages the advantages of different models, explores the complementarity between them, and further improves prediction accuracy.

[0160] In this embodiment of the invention, new energy power prediction is used to provide decision support for power grid dispatching and operation.

[0161] In summary, this invention proposes a new energy power prediction method. This method, through meticulous preprocessing and feature construction of historical operating data and comprehensive external data, fully explores the multi-dimensional correlations between data, constructs a standardized third feature set, provides high-quality input for the prediction model, and improves the accuracy and reliability of predictions. It can perform accurate and personalized predictions tailored to the unique power generation characteristics of different types of new energy power plants. When facing complex and ever-changing external meteorological conditions and dynamic changes in the operating status of power generation equipment, it can timely and accurately capture and reflect the actual changes in new energy power.

[0162] Example 2, in a preferred embodiment, involves performing a fifth result fusion operation on the predicted active power values ​​of new energy sources. The specific operation for predicting new energy power using the fused prediction results can be as follows:

[0163] Specifically, a single model is inevitably prone to bias. This stage further improves the stability and accuracy of predictions by fusing the results of multiple models or methods.

[0164] Furthermore, this invention can run multiple models (such as XGBoost, LSTM, and SVM) simultaneously, each outputting a predicted sequence. Weighted fusion prediction can be used for computation, as shown below:

[0165]

[0166] Among them, P final,t+k This represents the predicted value for the k-th hour after final fusion; w represents the predicted output of the i-th model; i The weights of the model satisfy ∑w i =1 and w i ≥0; weights can be dynamically allocated based on the performance of each model on the validation set (e.g., the inverse of MAE).

[0167] For example, if XGBoost performs well on sunny days and LSTM is more stable on cloudy or rainy days, an adaptive weight mechanism can be designed.

[0168] Furthermore, the final output is a fused power prediction curve:

[0169] P final =[P final,t+1 ,P final,t+2 ,...,P final,t+K ]

[0170] The results demonstrate stronger robustness and generalization ability, and can effectively cope with complex weather changes and equipment fluctuations.

[0171] Example 3, referring to Figure 2 This embodiment also provides a new energy power prediction system, including:

[0172] The data acquisition and processing module is used to acquire historical operating data of new energy power plants and comprehensive external data, and to perform the first preprocessing on the historical operating data and comprehensive external data.

[0173] The feature set acquisition module is used to perform a second feature construction operation on the preprocessed historical running data and integrated external data to obtain the constructed third feature set.

[0174] The prediction model building module is used to perform model training operations on the third feature set. The model training operations are used to train the new energy power prediction model.

[0175] The prediction module is used to perform the fourth prediction calculation operation based on the new energy power prediction model. The fourth prediction calculation operation is used to generate the predicted value of new energy active power for future periods.

[0176] The fusion module is used to perform a fifth result fusion operation on the predicted active power of new energy sources, and to use the fused prediction results to predict the power of new energy sources.

[0177] New energy power forecasting is used to provide decision support for power grid dispatching and operation.

[0178] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0179] This embodiment also provides an electronic device, which can be a terminal, and its internal structure diagram can be as follows: Figure 2 As shown, the electronic device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a new energy power prediction method. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the device's casing, or an external keyboard, touchpad, or mouse.

[0180] This embodiment also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it performs the following steps:

[0181] Acquire historical operating data of new energy power plants and comprehensive external data, and perform the first preprocessing on the historical operating data and comprehensive external data;

[0182] A second feature construction operation is performed on the preprocessed historical running data and the integrated external data to obtain the constructed third feature set;

[0183] The model training operation is performed on the third feature set, and the model training operation is used to train the new energy power prediction model.

[0184] The fourth prediction calculation operation is performed based on the new energy power prediction model. The fourth prediction calculation operation is used to generate the predicted value of new energy active power for future periods.

[0185] The predicted active power of new energy sources is fused using a fifth result fusion operation, and the fused prediction results are used to predict the power of new energy sources.

[0186] New energy power forecasting is used to provide decision support for power grid dispatching and operation.

[0187] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0188] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0189] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for predicting the power output of new energy sources, characterized in that, include: Acquire historical operating data of new energy power plants and comprehensive external data, and perform a first preprocessing on the historical operating data and comprehensive external data; A second feature construction operation is performed on the preprocessed historical running data and the integrated external data to obtain the constructed third feature set; The third feature set is subjected to model training operations, which are used to train a new energy power prediction model. A fourth prediction calculation operation is performed based on the new energy power prediction model. The fourth prediction calculation operation is used to generate the predicted value of new energy active power for future periods. The predicted active power of the new energy source is fused using a fifth result fusion operation, and the fused prediction result is used to predict the power of the new energy source. The new energy power prediction is used to provide decision support for power grid dispatch and operation.

2. The new energy power prediction method as described in claim 1, characterized in that, The second feature construction operation includes: Establish multi-dimensional correlations between historical operational data and comprehensive external data, and construct features based on these correlations; The multi-dimensional relationships include at least one or more of the following relationships: Power-meteorological correlation, spatial correlation, time-series autocorrelation, equipment operating status correlation, and historical power pattern matching.

3. The new energy power prediction method as described in claim 2, characterized in that, The second feature construction operation also includes: The historical operational data is matched using field matching based on the power plant number and timestamp. The integrated external data is initially aligned with the historical operational data; Establish a candidate feature set based on the initial alignment; Obtain a set of candidate features that satisfy multi-dimensional correlations.

4. The new energy power prediction method as described in claim 3, characterized in that, The second feature construction operation also includes: Integrate the candidate feature set that satisfies multi-dimensional correlation; Construct a feature dataset with a unified field structure as the third feature set; The unified field structure includes feature name, data source, data type, and feature description.

5. The new energy power prediction method as described in claim 4, characterized in that, The model training operation includes: The input characteristics and output target variables of the pre-defined new energy power prediction model are defined. The input features are features in the third feature set, and the output target variable is the active power of new energy sources in the future time period; Model training is performed based on the input features and the output target variable.

6. The new energy power prediction method as described in claim 5, characterized in that, The fourth prediction calculation operation based on the new energy power prediction model includes: Obtain comprehensive external data and the latest historical operational data for the period to be predicted; Perform the same preprocessing and feature construction operations on the acquired data as in the training phase; The constructed feature data is input into the trained new energy power prediction model to obtain the predicted value of new energy active power.

7. The new energy power prediction method as described in claim 6, characterized in that, The historical operational data and comprehensive external data include: The historical operating data includes historical active power curves of new energy power plants, equipment operating status, and geographical location information synchronized with the OCS system. The integrated external data includes real-time meteorological data, numerical weather forecast data, and GIS data.

8. A new energy power prediction system, using the method described in any one of claims 1 to 7, characterized in that, include: The data acquisition and processing module is used to acquire historical operating data of new energy power plants and comprehensive external data, and to perform a first preprocessing on the historical operating data and comprehensive external data. The feature set acquisition module is used to perform a second feature construction operation on the preprocessed historical running data and integrated external data to obtain the constructed third feature set. The prediction model building module is used to perform model training operations on the third feature set, and the model training operations are used to train the new energy power prediction model. The prediction module is used to perform a fourth prediction calculation operation based on the new energy power prediction model. The fourth prediction calculation operation is used to generate the predicted value of new energy active power for future periods. The fusion module is used to perform a fifth result fusion operation on the predicted value of the active power of the new energy source, and use the fused prediction result to predict the power of the new energy source. The new energy power prediction is used to provide decision support for power grid dispatch and operation.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the new energy power prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the new energy power prediction method according to any one of claims 1 to 7.

Images