Intelligent prediction method and system for wind and light power generation based on multi-source data fusion

Abstract

The application relates to the technical field of wind and light power generation, and particularly discloses a wind and light power generation capacity intelligent prediction method and system based on multi-source data fusion. The method realizes high-precision intelligent prediction of new energy power generation capacity through three core steps. Multi-source data of power generation performance, environment and equipment state are collected and fused in real time. Intelligent fusion adjustment is carried out on the basis of preprocessing, so that the accuracy and integrity of input data are ensured. Temporal and spatial correlation features are constructed by using the multi-source data. The equipment layout of a wind power plant and a photovoltaic power station is modeled as a graph structure, so that the spatial interaction between the equipment is effectively captured. After the features are input into a prediction model, the model can intelligently adjust parameters based on an error index, continuously optimize prediction accuracy, compare the prediction value with the actual power generation capacity, form a closed-loop feedback, trigger parameter adjustment of the model again, and ensure long-term reliability of the prediction result.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wind and light power generation, in particular to a multi-source data fusion intelligent prediction method and system for wind and light power generation. BACKGROUND

[0002] The multi-source data fusion intelligent prediction method and system for wind and light power generation collects data from multiple sources, including numerical weather prediction data, historical meteorological observation data, geographic information data, historical power output data of wind farms and photovoltaic power stations, and real-time operation state data. Through data cleaning, space-time alignment, feature engineering, and other technologies, the multi-source heterogeneous data is fused and processed to improve the quality. Based on the high-quality data set after fusion, a hybrid intelligent prediction model integrating physical mechanisms and data-driven is constructed and trained. The model can capture the complex nonlinear spatiotemporal variation of wind energy and solar energy. The trained model receives real-time fusion data input to predict future short-term or ultra-short-term wind power and photovoltaic power generation. After the prediction result is output, the system performs error analysis and result visualization, and can dynamically update and optimize the model parameters based on feedback information, forming a continuous improvement closed-loop prediction process.

[0003] For example, the wind and light complementary power generation system power generation prediction method disclosed in Chinese patent application CN119834203A includes collecting data to obtain historical data, preprocessing the historical data to obtain preprocessed data, discretizing the preprocessed data to obtain discrete data, constructing an HMMC model, training the HMMC model based on the historical data set, collecting real-time data, calculating a heat index based on the real-time data, and predicting power based on the heat index.

[0004] For example, the distributed photovoltaic power generation prediction method and system disclosed in Chinese patent application CN118213970A includes collecting historical power generation data of a target area and preprocessing the historical power generation data, the historical power generation data including historical photovoltaic power generation and corresponding power generation time; dividing each day into several time periods and respectively counting the power generation power of the historical power generation data in each time period to obtain a photovoltaic power generation power set; obtaining historical weather forecast information of the target area, the historical weather forecast information including sunshine intensity, temperature, rain, wind speed, and seasonal change; processing the historical weather forecast information, marking the historical weather forecast information according to the daily division time period of the photovoltaic power generation power set, and obtaining a power generation power correlation set.

[0005] The above technology has at least the following technical problems: in the data fusion process, raw data is first collected directly from multiple heterogeneous data sources; then preliminary data cleaning and conversion are performed; then a specific fusion algorithm is applied to combine the data; and finally the fusion result is output; this process has significant defects, the fusion algorithm itself often lacks adaptability and has high computational complexity, and the adjustment and optimization link in the process is weak or unsystematic, ultimately resulting in low efficiency of the entire data fusion process. SUMMARY

[0006] In view of the deficiencies of the prior art, the present application provides a multi-source data fusion wind and light power generation intelligent prediction method and system, which can effectively solve the problems involved in the above background art.

[0007] To achieve the above purpose, the present application is realized by the following technical scheme: the present application provides a multi-source data fusion wind and light power generation intelligent prediction method in the first aspect, comprising: step one, collecting real-time power generation performance parameters, environmental data and equipment operation state data monitored by the regional new energy control center, respectively preprocessing the power generation performance parameters, environmental data and equipment operation state data, fusing the preprocessed power generation performance parameters, environmental data and equipment operation state data, collecting and analyzing data fusion process parameters, and judging whether to adjust the data fusion process, so as to obtain accurate multi-source data set; step two, constructing the space-time correlation feature through the multi-source data set and inputting it into the power generation prediction model, the power generation prediction model outputs the power generation prediction value, collects and analyzes the model error index, judges whether to adjust the parameters of the prediction model, so as to intelligently optimize the precision of the power generation prediction model; the space-time correlation feature represents modeling the equipment layout of the wind farm and the photovoltaic power station as a graph structure, the node represents the wind turbine or the photovoltaic array, and the edge represents the spatial correlation between the equipment; step three, collecting the actual total power generation and comparing it with the power generation prediction value output by the prediction model, based on the comparison result, judging whether to readjust the prediction model parameters, so as to optimize the result of intelligently predicting the wind and light power generation in the prediction model.

[0008] The second aspect of the present application provides a multi-source data fusion wind and light power generation intelligent prediction system, comprising: a multi-source data fusion module, a prediction model module, an intelligent optimization module and a database.

[0009] Compared with the prior art, the embodiments of the present application have at least the following advantages or beneficial effects:

[0010] (1) The present application provides a multi-source data fusion wind and light power generation intelligent prediction method and system, which realizes high-precision intelligent prediction of new energy power generation through three core steps. First, real-time collection and fusion of power generation performance, environmental and equipment state multi-source data are carried out, intelligent fusion adjustment is carried out on the basis of preprocessing, and the accuracy and integrity of the input data are ensured. Secondly, the spatial and temporal correlation features are constructed by using multi-source data, the device layout of the wind farm and the photovoltaic power station is modeled as a graph structure, and the spatial interaction between the devices is effectively captured. After these features are input into the prediction model, the model can intelligently adjust the parameters based on the error index, continuously optimize the prediction accuracy. Finally, by comparing the predicted value with the actual power generation, a closed-loop feedback is formed, and the model parameter adjustment is triggered again, ensuring the long-term reliability of the wind and light power generation prediction results.

[0011] (2) The present application preprocesses the power generation performance, environmental data and equipment state, effectively eliminates noise, outliers and missing values, and ensures the reliability and consistency of the data. The fusion of the three types of heterogeneous data breaks the data silos and constructs a more comprehensive multi-dimensional view reflecting the actual operation state of the new energy power generation system. Real-time monitoring and dynamic adjustment of the fusion process, continuous and intelligent quality optimization and adaptive adjustment of the key basic data input into the wind and light prediction model ensure that the data foundation relied on by the prediction model is always the most accurate, freshest and most consistent with the current actual operation state.

[0012] (3) The present application constructs the spatio-temporal correlation features using the high-quality fusion data obtained, which can more effectively capture the complex spatio-temporal dependence relationship affecting wind and light power generation, significantly enhancing the model's understanding ability of the power generation change pattern. After inputting the constructed features into the prediction model, the model error index is collected and analyzed in real time, and the model parameters are intelligently adjusted according to the analysis results, realizing the dynamic closed-loop optimization of the prediction model. This greatly improves the wind and light power generation prediction accuracy and the robustness of the model, making it self-adaptive to different weather conditions, seasonal changes and equipment state fluctuations.

[0013] (4) The present application realizes the continuous improvement of the model. The predicted value of the power generation output by the prediction model is compared with the actual measured total power generation in real time or periodically, and the performance of the model in the real world is directly evaluated. Based on the comparison results, it is judged whether the model parameters need to be adjusted again, forming a closed-loop mechanism of prediction, verification, adjustment and re-prediction. This mechanism ensures that the prediction model does not stagnate, but can continuously learn new operation data and pattern changes, and continuously narrow the gap between the predicted value of wind and light power generation and the actual wind and light power generation value. BRIEF DESCRIPTION OF DRAWINGS

[0014] The application is further described with the help of the accompanying drawings, but the embodiments in the drawings do not constitute any limitation to the application, and other embodiments can be obtained by those skilled in the art without creative labor on the basis of the following drawings.

[0015] Figure 1 The figure is a schematic diagram of the method steps of the application.

[0016] Figure 2 The figure is a schematic diagram of the system module connection of the application.

[0017] Figure 3 The figure is a schematic diagram of the adjustment process of the data fusion process of the application.

[0018] Figure 4 The figure is a schematic diagram of the secondary adjustment process of the data fusion process of the application.

[0019] Figure 5 The figure is a schematic diagram of the adjustment process of the prediction model of the application.

[0020] Figure 6 The figure is a schematic diagram of the secondary adjustment process of the prediction model of the application.

[0021] Figure 7 The figure is a schematic diagram of the relationship between the power generation difference and the actual power generation of the application. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the application will be described clearly and completely below with the help of the accompanying drawings in the embodiments of the application. Obviously, the described embodiments are only a part of the embodiments of the application, but not all the embodiments of the application. Based on the embodiments in the application, all the other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application.

[0023] REFERENCE Figure 1As shown, the first aspect of the application provides a multi-source data fusion wind and light power generation intelligent prediction method, comprising: step one, collecting the power generation performance parameters, environmental data and equipment operation state data monitored by the regional new energy centralized control center in real time, preprocessing the power generation performance parameters, environmental data and equipment operation state data respectively, fusing the preprocessed power generation performance parameters, environmental data and equipment operation state data, collecting and analyzing data fusion process parameters, judging whether to adjust the data fusion process, so as to obtain accurate multi-source data set; step two, constructing the space-time correlation characteristics through the multi-source data set, and inputting into the power generation prediction model, the prediction model outputs the power generation prediction value, collects and analyzes the model error index, judges whether to adjust the parameters of the prediction model, so as to intelligently optimize the precision of the power generation prediction model; the space-time correlation characteristics represent that the equipment layout of the wind farm and the photovoltaic power station is modeled as a graph structure, the node represents the fan or the photovoltaic array, and the edge represents the spatial correlation between the equipment; step three, collecting the actual total power generation, and comparing with the power generation prediction value output by the prediction model, based on the comparison result, judging whether to re-adjust the prediction model parameters, so as to optimize the result of intelligently predicting the wind and light power generation in the prediction model.

[0024] Figure 3 The adjustment flow chart of the data fusion process of the application is to judge whether the fusion result is reliable by calculating the data fusion process anomaly index. If the data fusion process anomaly index is not greater than the data fusion process anomaly threshold, the fusion result is stored and subsequent processing is performed; if the data fusion process anomaly index is greater than the data fusion process anomaly threshold, the fusion is cancelled, the original data is saved, and the fusion process is tried to be adjusted, and after adjustment, it is detected again, if successful, the normal process is followed, if still failed, it is decided whether to perform secondary adjustment; Figure 4 The secondary adjustment flow chart of the data fusion process of the application is to judge whether the data fusion process is adjusted twice, and then the core link is to check whether the data fusion process anomaly index after adjustment is not more than the data fusion process anomaly threshold: if the data fusion process anomaly index is not more than the data fusion process anomaly threshold, the multi-source data is stored in the normal area and the space-time correlation characteristics are constructed; if the data fusion process anomaly index exceeds the data fusion process anomaly threshold, the preprocessed data is first stored in the abnormal area, then the fusion process is adjusted and the new anomaly index is calculated to judge again; if the data fusion process anomaly index after re-adjustment is not more than the data fusion process anomaly threshold, normal storage and feature construction are also performed; if the data fusion process anomaly index after re-adjustment exceeds the data fusion process anomaly threshold, a warning is triggered.

[0025] Specifically, whether to adjust the data fusion process is judged, and the specific judgment process is: collecting and analyzing the data fusion process parameters to obtain the data fusion process anomaly index.

[0026] The power generation performance parameter reflects the real-time power generation capacity and efficiency of the power generation equipment.

[0027] The environmental data reflects external environmental factors that affect the power generation performance and operation of the power generation equipment.

[0028] The equipment operation state data reflects the operation state of the power generation equipment.

[0029] The data fusion process anomaly index is compared with the data fusion process anomaly threshold value, and if the data fusion process anomaly index is less than or equal to the data fusion process anomaly threshold value, the multi-source data set is stored in the normal area of the database, and the multi-source data set is constructed to have spatial correlation characteristics.

[0030] The multi-source data set is a comprehensive data set formed by preprocessing, fusion optimization of the power generation performance parameter, environmental data and equipment operation state data monitored by the regional new energy centralized control center.

[0031] If the data fusion process anomaly index is greater than the data fusion process anomaly threshold value, the fusion is cancelled, and the preprocessed power generation performance parameter, environmental data and equipment operation state data are stored in the abnormal area of the database, and the observation noise covariance matrix in the data fusion algorithm is increased based on the data fusion process anomaly index, and the preprocessed power generation performance parameter, environmental data and equipment operation state data stored in the abnormal area of the database are fused, and the data fusion process parameters are collected and analyzed again to obtain the adjusted data fusion process anomaly index, and then it is judged whether to adjust the data fusion process twice.

[0032] The normal area is a standard data warehouse for system operation, and the abnormal area is an isolation buffer area and repair work area to protect the standard from being polluted. The two areas have clear division of labor and jointly build a more robust, reliable and maintainable data fusion and management system.

[0033] The data fusion process anomaly threshold value is used to represent the maximum upper limit of the data fusion process anomaly index, and is determined by analyzing the multi-batch normal data fusion process anomaly index generated by the regional new energy centralized control center during the historical normal operation, combining the experience of field experts and the requirements of the system for reliability, and stored in the database.

[0034] Based on the data fusion process anomaly index increasing the observation noise covariance matrix in the data fusion algorithm, the data fusion process anomaly index is divided into different level intervals in the database, and each level interval corresponds to a different increase value of the observation noise covariance matrix in the data fusion algorithm. The observation noise covariance matrix in the data fusion algorithm is added to the increase value of the observation noise covariance matrix in the data fusion algorithm to obtain the observation noise covariance matrix in the next step of the data fusion algorithm.

[0035] The adjustment of the observation noise covariance matrix in the data fusion algorithm based on the increasing data fusion process anomaly index can effectively reduce the interference of abnormal data on the fusion result and improve the robustness of the algorithm. In the intelligent prediction of wind and solar power generation, multi-source observation data from meteorological stations, historical power, numerical weather forecasts, and even adjacent station data may introduce abnormal values due to sensor failure, communication interference, or local extreme weather. By adjusting the observation noise covariance matrix, the fusion algorithm is more inclined to rely on the predicted value of the system model rather than the possibly abnormal observation data, thereby suppressing the negative impact of abnormal values on the fusion result. By dynamically adjusting the noise covariance matrix, the data quality fluctuations caused by the inherent intermittency and volatility of wind and solar data sources can be adaptively addressed, improving the stability of the fusion result. This adjustment can indirectly reduce the data fusion process anomaly index, as the contribution of abnormal data in the data fusion process is weakened, reducing inconsistencies in the fusion process, which may cause the data fusion process anomaly index calculated twice to fall below the data fusion process anomaly threshold, avoiding unnecessary repeated adjustments. For wind and solar prediction systems, this means that even in the case of temporary unreliability of some data sources, the system can maintain relatively stable and accurate prediction output, reducing prediction jumps or deviations caused by data anomalies.

[0036] The power generation performance parameters include active power, reactive power, voltage, current, frequency, conversion efficiency, capacity factor, and performance ratio, etc.

[0037] The environmental data include light intensity, wind speed, wind direction, ambient temperature, humidity, air pressure, and local wind speed turbulence, etc.

[0038] The equipment operating state data include vibration spectrum, bearing temperature, gear box oil quality, blade pitch angle, insulation resistance, winding temperature, circuit breaker state, and grounding resistance, etc.

[0039] The data preprocessing includes cleaning the real-time collected wind power and photovoltaic operating parameters and equipment state data, removing abnormal values and filling missing data, and then performing standardization processing to unify the dimensions and sampling frequencies of multi-source data, and analyzing the equipment performance trend through time series decomposition.

[0040] The process of unifying the dimensions and sampling frequencies of multi-source data is to standardize all these different units and ranges of data and align all time series data to the same timestamp grid with the same sampling interval.

[0041] By collecting and analyzing the power generation performance parameters, environmental data and equipment operation state data in real time, a multi-source data set is constructed and a data fusion process anomaly index is calculated, which can comprehensively and objectively evaluate the quality and reliability of data fusion. When the data fusion process anomaly index is greater than the data fusion process anomaly threshold, the system automatically cancels the fusion and stores the original data to the abnormal area, and at the same time optimizes the fusion algorithm by increasing the observation noise covariance matrix, effectively reducing the interference of abnormal data on the fusion result and improving the accuracy and robustness of data fusion. Through secondary adjustment and re-fusion, the effective use of data is further ensured, and data waste is avoided. The whole process not only realizes the accurate monitoring of the operation state of new energy power generation equipment, but also provides a high-quality data basis for subsequent fault diagnosis and performance optimization, thereby improving the operation efficiency and reliability of the new energy control center.

[0042] Specifically, whether to perform secondary adjustment on the data fusion process is determined, and the specific determination process is as follows: if the adjusted data fusion process anomaly index is less than or equal to the data fusion process anomaly threshold, the multi-source data set is stored to the normal area of the database, and the spatio-temporal correlation features are constructed based on the multi-source data set.

[0043] If the adjusted data fusion process anomaly index is greater than the data fusion process anomaly threshold, the preprocessed power generation performance parameters, environmental data and equipment operation state data are stored to the abnormal area of the database again, the observation noise covariance matrix in the data fusion algorithm is increased based on the adjusted data fusion process anomaly index, and the buffer size in the data fusion algorithm is increased, and the power generation performance parameters, environmental data and equipment operation state data in the abnormal area of the database are re-fused to obtain a re-adjusted data fusion process anomaly index.

[0044] If the re-adjusted data fusion process anomaly index is less than or equal to the data fusion process anomaly threshold, the multi-source data set is stored to the normal area of the database, and the spatio-temporal correlation features are constructed based on the multi-source data set.

[0045] The multi-source data set constructs spatio-temporal correlation features, which means that by integrating heterogeneous data from different sources, these data are aligned and fused in a unified geographic spatial framework and time reference, and specific algorithms or models are used to mine and extract key indicators or patterns that can represent the state and dynamic change rule of target objects or phenomena at a specific geographic location and a specific time point. The core is to reveal the implicit spatio-temporal dependence relationship, interaction and evolution trend between data, so as to form a comprehensive feature expression containing spatio-temporal context information beyond the perspective of a single data source, and provide a deeper and more comprehensive insight basis for subsequent analysis, prediction or decision-making.

[0046] If the adjusted data fusion process anomaly index is greater than the data fusion process anomaly threshold, a pre-warning is triggered.

[0047] The pre-warning is triggered, and a sound and light pre-warning is performed to remind the staff.

[0048] In the adjustment process of data fusion, increasing the observation noise covariance matrix has a more significant effect than increasing the buffer size. The noise covariance matrix directly controls the trust degree of the fusion algorithm for observation data, and its increase reduces the sensitivity of abnormal data, thereby more directly suppressing the data fusion process anomaly index; while the increase of the buffer size mainly indirectly affects the fusion result by smoothing historical data, and the effect is relatively weak. Therefore, if the data fusion process anomaly index needs to be adjusted quickly, the noise covariance matrix in the data fusion algorithm is adjusted first; if the anomaly has time correlation, the buffer size is optimized.

[0049] Based on the adjusted data fusion process anomaly index, the observation noise covariance matrix in the data fusion algorithm and the buffer size in the data fusion algorithm are increased, the data fusion process anomaly index is divided into different level intervals in the database, each level interval corresponds to a different increase value of the observation noise covariance matrix in the data fusion algorithm, the observation noise covariance matrix in the current data fusion algorithm plus the increase value of the observation noise covariance matrix in the data fusion algorithm, to obtain the observation noise covariance matrix in the next step data fusion algorithm; the data fusion process anomaly index is divided into different level intervals in the database, each level interval corresponds to a different increase value of the buffer size in the data fusion algorithm, the buffer size in the current data fusion algorithm plus the increase value of the buffer size in the data fusion algorithm, to obtain the buffer size in the next step data fusion algorithm.

[0050] Based on the adjusted data fusion process anomaly index, the observation noise covariance matrix and the buffer size adjustment strategy can effectively improve the fault tolerance and stability of the data fusion algorithm to abnormal data. Increasing the observation noise covariance matrix can reduce the sensitivity of the algorithm to the current adjusted data fusion process anomaly index, avoiding the deviation of the fusion result from the true state due to individual abnormal values; while increasing the buffer size can smooth abnormal fluctuations by introducing more historical data or adjacent data, improving the robustness of the fusion result. The combined effect of this double adjustment can significantly suppress the further increase of the data fusion process anomaly index, and even reduce it to below the data fusion process anomaly threshold through re-fusion, thereby avoiding unnecessary pre-warning and ensuring the reliability of the fusion result. If the adjustment is appropriate, the adjusted data fusion process anomaly index may quickly converge due to the enhancement of the algorithm adaptability, and eventually meet the normal fusion conditions.

[0051] By dynamically adjusting the observation noise covariance matrix of the data fusion algorithm and the buffer size, the system can adaptively optimize the fusion process, effectively reduce the interference of abnormal data, and improve the accuracy and stability of the fusion result. When the adjusted data fusion process anomaly index is greater than the data fusion process anomaly threshold, the system adopts a hierarchical processing strategy, first performs secondary fusion optimization, and if it still does not meet the standard, triggers the sound control warning to ensure that abnormal data is processed in time and avoid false data entering the normal database. This multi-level adjustment mechanism not only enhances the robustness of data fusion, but also improves the data reliability of the new energy centralized control center, providing more accurate data support for the state monitoring, fault diagnosis and performance optimization of power generation equipment. At the same time, the triggering of the warning mechanism can remind the operation and maintenance personnel to intervene and process, further ensuring the safe and stable operation of the system.

[0052] Further, the data fusion process parameters are collected and analyzed, and the specific analysis process is as follows: the data fusion process parameters include the noise covariance of the data fusion process, the coverage of the data fusion process, and the error covariance of the data fusion process.

[0053] The noise covariance of the data fusion process can be calculated by statistically analyzing the output data of the sensors (temperature sensors, radars, cameras, IMUs, etc.) in a stable state, that is, by collecting multiple sets of data when there is no target change, calculating the variance matrix to represent the noise characteristics; the coverage of the data fusion process needs to be evaluated in terms of space and information dimension, and quantified by the overlap rate of the sensor detection range; the error covariance of the data fusion process is usually calculated based on the residual error between the fusion result and the true value, and the covariance matrix of the residual error is calculated through multiple experiments.

[0054] The noise covariance matrix is used to quantify the statistical characteristics of sensor measurement noise, and its calculation method is as follows: in a stable state, multiple sets of data output by the sensor are collected, the mean value is calculated, and then the noise covariance matrix is obtained by using the covariance formula. The diagonal elements of this matrix represent the noise variance of each sensor, and the non-diagonal elements reflect the correlation between the noise of different sensors.

[0055] The coverage measures the spatial or information redundancy of data fusion, which can be quantified by the overlap rate of the sensor detection range. In terms of information dimension, the feature complementarity of different sensors observing the same target is evaluated, such as the proportion of the common observed features to the total features.

[0056] The error covariance matrix reflects the accuracy of the fusion result, and its calculation is based on the residual error between the fusion estimate and the true value. The specific steps are as follows: perform multiple independent experiments, record the error between each fusion estimate and the true value, and then calculate the error covariance matrix using the formula. If the true value cannot be directly obtained, the cross-validation or confidence weighting method can be used to indirectly estimate the error to ensure the optimization basis of the fusion algorithm.

[0057] Multiple independent experiments are performed, which means that the fusion algorithm is repeatedly executed multiple times, and each execution is based on newly generated, statistically independent input data. Each execution produces a fusion result and a corresponding error vector. After collecting a sufficient number of these independent error vectors, statistical methods can be used to estimate the error covariance matrix of the fusion result, which quantifies the uncertainty distribution of the fusion estimate around the true value and the correlation between the errors of each state component. It is the core basis for evaluating and optimizing the accuracy of the fusion algorithm.

[0058] The calculation of the data fusion process anomaly index is based on the proportional relationship between the noise covariance of the data fusion process and the defined noise covariance, the proportional relationship between the coverage of the data fusion process and the defined coverage, and the proportional relationship between the error covariance of the data fusion process and the defined error covariance. The specific process is as follows: the defined noise covariance, the defined coverage, and the defined error covariance contrast are taken as the benchmark, the proportional relationship between the noise covariance of the data fusion process, the coverage of the data fusion process, and the error covariance of the data fusion process and the corresponding defined values is calculated respectively, different measurement proportions are given according to the influence degree of each proportional relationship on the data fusion process anomaly index, and the data fusion process anomaly index is finally obtained by summarizing.

[0059] The data fusion process anomaly index is used to quantify the comprehensive deviation degree of the data fusion result from the expected normal state.

[0060] ;

[0061] SJH is the data fusion process anomaly index, SRZ is the noise covariance of the data fusion process, SRF is the coverage of the data fusion process, SRW is the error covariance of the data fusion process, SRZ_L is the preset defined noise covariance in the database, SRF_L is the preset defined coverage in the database, SRW_L is the preset defined error covariance in the database, A1 is the measurement proportion corresponding to the noise covariance in the database, A2 is the measurement proportion corresponding to the coverage in the database, and A3 is the measurement proportion corresponding to the error covariance in the database.

[0062] The defined noise covariance is used to represent the upper limit value of the noise covariance of the data fusion process; the defined coverage is used to represent the lower limit value of the coverage of the data fusion process; and the defined error covariance is used to represent the upper limit value of the error covariance of the data fusion process.

[0063] The quality and reliability of the data fusion process are determined by three parameters: the noise covariance of the data fusion process, the coverage of the data fusion process, and the error covariance of the data fusion process. The noise covariance of the data fusion process measures the degree of noise interference in the fusion process. The coverage of the data fusion process represents the completeness of the coverage of the target area by the data source. The error covariance of the data fusion process directly reflects the accuracy deviation of the fusion result. These three parameters interact with each other: insufficient coverage of the data fusion process will reduce the ability of the system to distinguish between signal and noise, indirectly amplify the influence of noise and directly lead to an increase in the error covariance of the data fusion process; high noise covariance will directly pollute the fusion result and significantly increase the error covariance of the data fusion process; finally, the error covariance of the data fusion process, as the core indicator of fusion accuracy, its rise is the direct result of the combined action of the lack of coverage of the data fusion process and noise interference, and the three form a mutually dependent triangular relationship. The data fusion process anomaly index is a comprehensive evaluation of the process anomaly by calculating the deviation of the three parameters from their respective preset baseline values: the noise covariance of the data fusion process is out of range, the coverage of the data fusion process is lower than the standard, or the error covariance of the data fusion process is out of range, all of which contribute to an increase in the index. The higher the index, the more serious the problems of noise interference, information loss, and result accuracy deviation, which are interlocked, and the lower the reliability and effectiveness of the data fusion process.

[0064] The measurement ratio corresponding to the noise covariance of the data fusion process represents the relative ratio between the noise covariance of the data fusion process and the preset defined noise covariance, and the degree of influence on the data fusion process anomaly index. This ratio value quantifies the contribution weight of the noise covariance in the overall anomaly evaluation, reflecting the key role of data noise fluctuation in the detection of fusion result deviation. The measurement ratio corresponding to the coverage of the data fusion process represents the relative ratio between the coverage of the data fusion process and the preset defined coverage, and the degree of influence on the anomaly index. This value determines the importance of coverage in the comprehensive evaluation and is used to measure the sensitivity of the data coverage range to the consistency of the fusion result. The measurement ratio corresponding to the error covariance of the data fusion process represents the relative ratio between the error covariance of the data fusion process and the preset defined error covariance, and the degree of influence on the anomaly index, which is used to evaluate the diagnostic value of error fluctuation in the fusion anomaly detection, highlighting the significant influence of error stability on the overall quantitative result.

[0065] The database stores preset data fusion process evaluation benchmark parameters, including defined noise covariance, defined coverage, and defined error covariance; these parameters are dynamically bound with key indicators of real-time data fusion processes and preset anomaly evaluation benchmarks through a structured parameter mapping table, and together constitute a complete data fusion process anomaly evaluation system. When it is necessary to calculate the anomaly index of a specific data fusion process, the system extracts data fusion process parameters including data fusion process noise covariance, data fusion process coverage, and data fusion process error covariance; then, based on a preset rule base, historical data similarity matching, or a statistical analysis model, these indicators are compared or calculated with the parameter mapping table in the database; finally, the defined noise covariance, the defined coverage, and the defined error covariance suitable for the fusion process are dynamically output, as well as the corresponding weight coefficients. Among them, A1, A2, and A3 measure the proportional value as the weight coefficient, and the value range is between 0 and 1.

[0066] Specifically, the model error indicators are collected and analyzed. The specific analysis process is: collect and analyze model error indicators, including power generation prediction model mean absolute error, power generation prediction model root mean square error, and power generation prediction model mean square error.

[0067] To measure the error indicators of the power generation prediction model, the following steps are needed: First, collect historical model prediction values and corresponding historical actual observation values, and ensure that the data time range is consistent and has no missing values. Calculate the historical model mean absolute error, the absolute value of the difference between each time point's historical model prediction value and the corresponding historical actual value, sum and take the average, reflecting the overall amplitude of the prediction bias. Calculate the power generation prediction model mean square error, square each error and take the average, which amplifies the influence of larger errors and is commonly used for gradient optimization. Calculate the power generation prediction model root mean square error, which takes the square root of the power generation prediction model mean absolute error, and its dimension is consistent with the actual value, which is used to evaluate the stability of the prediction.

[0068] Obtain the data fusion process anomaly terminal value, and match the data fusion process anomaly increment from the database.

[0069] The data fusion process anomaly terminal value is dynamically generated by analyzing the latest data generated in real time in the data fusion process using a specific calculation method. The data fusion process anomaly index intuitively and quantitatively represents the current health status and the degree of abnormality of the data fusion process. The data fusion process anomaly index will be updated continuously over time and with the arrival of new data.

[0070] The calculation of the prediction model evaluation index abnormal value is based on the proportional relationship between the average absolute error of the power generation prediction model and the defined average absolute error, the proportional relationship between the root mean square error of the power generation prediction model and the defined root mean square error, and the proportional relationship between the mean square error of the power generation prediction model and the defined mean square error. The specific process is as follows: taking the defined average absolute error, the defined root mean square error, and the defined mean square error as the benchmark, the proportional relationship between the average absolute error of the power generation prediction model, the root mean square error of the power generation prediction model, and the mean square error of the power generation prediction model and the corresponding defined values is calculated respectively. According to the influence degree of each proportional relationship on the prediction model evaluation index abnormal value, different measurement proportions are given, and then the data fusion process abnormal increment is added to obtain the prediction model evaluation index abnormal value.

[0071] The prediction model evaluation index abnormal value is used to quantify the abnormal degree of the overall performance of the prediction model.

[0072] ;

[0073] PGZ is the prediction model evaluation index abnormal value, MPW is the average absolute error of the power generation prediction model, MJC is the root mean square error of the power generation prediction model, MFC is the mean square error of the power generation prediction model, H is the data fusion process abnormal increment, MPW_L is the defined average absolute error in the database, MJC_L is the defined root mean square error in the database, MFC_L is the defined mean square error in the database, C1 is the measurement proportion corresponding to the average absolute error in the database, C2 is the measurement proportion corresponding to the root mean square error in the database, and C3 is the measurement proportion corresponding to the mean square error in the database.

[0074] The defined average absolute error is used to represent the upper limit value of the average absolute error of the power generation prediction model; the defined root mean square error is used to represent the upper limit value of the root mean square error of the power generation prediction model; and the defined mean square error is used to represent the upper limit value of the mean square error of the power generation prediction model.

[0075] The mean absolute error of the power generation prediction model, the root mean square error of the power generation prediction model, and the mean square error of the power generation prediction model reflect the prediction accuracy and stability of the model itself, while the data fusion process anomaly increment captures the additional uncertainty in the data preprocessing or integration link, both of which together determine the level of outliers: if the model error proportion or the fusion increment significantly increases, the outliers will also increase, indicating potential problems in model performance or data quality. The data fusion process anomaly increment is used to quantify the additional error introduced by information integration bias or noise in the data fusion process. This increment can improve the robustness of outlier evaluation, ensuring that the final result is more in line with the real fluctuations of model performance in complex scenarios, especially when data sources are heterogeneous or the fusion algorithm has limitations. When the mean absolute error of the power generation prediction model increases, it usually means that the overall prediction bias level rises, which will directly lead to a significant increase in the mean square error of the power generation prediction model, and further push up the root mean square error of the power generation prediction model. When the root mean square error of the power generation prediction model increases, it leads to a significant increase in the mean square error of the power generation prediction model and the mean absolute error, which is squared; when the mean square error of the power generation prediction model increases, since the root mean square error of the power generation prediction model is the square root of the mean square error of the power generation prediction model, the root mean square error of the power generation prediction model will inevitably increase, and the increase reflects the degree of increase in the mean square error of the power generation prediction model. At the same time, the mean absolute error of the power generation prediction model will also increase with the increase of the mean square error of the power generation prediction model, because they both reflect the overall level of error.

[0076] The metric proportion corresponding to the mean absolute error of the power generation prediction model represents the relative proportional relationship between the mean absolute error of the power generation prediction model and the preset defined mean absolute error, and the influence degree on the abnormal value of the evaluation index of the prediction model. This proportion value quantifies the contribution weight of the mean absolute error in the overall evaluation, and reflects the relative importance of prediction accuracy bias in anomaly detection. The metric proportion corresponding to the root mean square error of the power generation prediction model represents the relative proportional relationship between the root mean square error of the power generation prediction model and the preset defined root mean square error, and the influence degree on the abnormal value. This value determines the importance of the root mean square error in the comprehensive evaluation, and is used to evaluate the key role of prediction stability in anomaly identification. The metric proportion corresponding to the mean square error of the power generation prediction model represents the relative proportional relationship between the mean square error of the power generation prediction model and the preset defined mean square error, and the influence degree on the abnormal value, which is used to measure the sensitivity of the mean square error in the prediction model anomaly detection, and highlights the diagnostic value of the prediction error dispersion degree in the overall quantification.

[0077] The preset power generation prediction model evaluation benchmark parameters stored in the database include the defined average absolute error, the defined root mean square error and the defined mean square error; these parameters are dynamically bound with the error indicators output by the power generation prediction model and the preset abnormal value evaluation benchmark through a structured parameter mapping table, and together constitute a complete prediction model evaluation index abnormal value calculation system. When it is necessary to calculate the abnormal value of a specific prediction model, the system extracts the error indicators of the current model and the data fusion process abnormal increment; then, based on the preset rule base or historical data matching, the error indicators are compared with the parameter mapping table in the database; finally, the defined average absolute error, the defined root mean square error and the defined mean square error suitable for the model are dynamically output, as well as the corresponding weight coefficients. Among them, the C1, C2 and C3 measurement proportion values are used as weight coefficients, and the value range of each is between 0 and 1.

[0078] Figure 5 For the adjustment flowchart of the prediction model of the application, the flowchart starts from constructing the space-time correlation characteristics, and based on this, the model error indicators are collected and analyzed to obtain the prediction model evaluation index abnormal value; then, it is judged: if the prediction model evaluation index abnormal value is less than or equal to the set evaluation index abnormal threshold value, the current prediction model parameters are retained and the power generation prediction value is directly output; if the prediction model evaluation index abnormal value is greater than the set evaluation index abnormal threshold value, the related multi-source data space-time correlation characteristics are stored in the abnormal area of the database, then the prediction model parameters are adjusted and the adjusted prediction model evaluation index abnormal value is recalculated, and whether further secondary adjustment is needed is judged based on the adjusted prediction model evaluation index abnormal value; Figure 6 For the secondary adjustment flowchart of the prediction model of the application, it is judged whether secondary adjustment is needed, if the adjusted evaluation index abnormal value does not exceed the evaluation index abnormal threshold value, the prediction value is directly output using the model; if the adjusted evaluation index abnormal value exceeds the evaluation index abnormal threshold value, the related data is stored in the abnormal area of the database and secondary adjustment is performed, the prediction model evaluation index abnormal value after re-adjustment is obtained after the secondary adjustment is evaluated again, if the prediction model evaluation index abnormal value after re-adjustment does not exceed the evaluation index abnormal threshold value, the prediction result is output, if the prediction model evaluation index abnormal value after re-adjustment exceeds the evaluation index abnormal threshold value, a warning signal is generated, ensuring that the final prediction value is output only when the model reliability meets the requirements.

[0079] Specifically, it is judged whether to adjust the parameters of the prediction model, and the specific judgment process is: collecting and analyzing the model error indicators to obtain the prediction model evaluation index abnormal value.

[0080] The prediction model evaluation index abnormal value is compared with the evaluation index abnormal threshold value, and if the prediction model evaluation index abnormal value is less than or equal to the evaluation index abnormal threshold value, the parameters of the prediction model are retained, and a power generation prediction value output by the prediction model is obtained.

[0081] If the prediction model evaluation index abnormal value is greater than the evaluation index abnormal threshold value, the spatiotemporal correlation features of the multi-source data set are stored in the abnormal area of the database, the learning rate in the prediction model is reduced based on the prediction model evaluation index abnormal value, the prediction model is retrained and verified by constructing spatiotemporal correlation features of the multi-source data set, and an adjusted prediction model evaluation index abnormal value is obtained. It is determined whether to perform secondary adjustment on the parameters of the prediction model.

[0082] The evaluation index abnormal threshold value is used to represent the maximum upper limit of the prediction model evaluation index abnormal value, is determined by comprehensive analysis of model historical performance and business requirements, is usually calculated based on historical evaluation index data of the model during stable operation, and is set in combination with a maximum acceptable error range of the business and stored in the database.

[0083] The learning rate in the prediction model is reduced based on the prediction model evaluation index abnormal value, the data fusion process abnormal index is divided into different grade intervals in the database, each grade interval corresponds to a different reduction value of the learning rate in the prediction model, the learning rate in the prediction model is reduced by the reduction value of the learning rate in the prediction model to obtain the learning rate in the prediction model for the next step.

[0084] The adjustment strategy of reducing the learning rate based on the abnormal value of the evaluation index of the prediction model has the core advantage of enhancing the convergence stability of the model in the abnormal data area by dynamically reducing the parameter update step. When the abnormal value of the evaluation index of the prediction model is high, moderately reducing the learning rate can prevent the parameters from falling into local optimization or divergence due to the drastic fluctuation of the gradient, and promote the model to learn the effective patterns in the spatio-temporal correlation characteristics more finely. This adjustment directly suppresses the overfitting or underfitting risk corresponding to the abnormal value, and through iterative retraining, the abnormal value of the evaluation index gradually converges to the evaluation index abnormal threshold, avoiding training shock caused by aggressive adjustment, and improving the model's ability to capture complex relationships in multi-source data through progressive optimization. Ultimately, the abnormal value tends to decrease and stabilize within a reasonable range. By adaptively reducing the parameter update step, the model's robustness to noise, errors, and inherent drastic fluctuations in wind and light is significantly enhanced. When the prediction model detects an increase in the evaluation index abnormal value, the risk of local optimization or divergence caused by drastic gradient fluctuations is suppressed, and the model is prompted to capture effective patterns in complex spatio-temporal correlation characteristics more finely, avoiding overfitting problems caused by abnormal values. By balancing training speed and accuracy through progressive optimization, this mechanism ensures the stability of the model's convergence in a multi-source heterogeneous data environment, systematically driving the prediction error to converge steadily within the threshold range, thereby improving the generalization ability and reliability of power generation prediction, providing accurate support for power grid dispatching and energy management.

[0085] By constructing a systematic error monitoring and dynamic parameter adjustment mechanism, the adaptability and accuracy of the prediction model are significantly improved, providing reliable support for wind and light power generation prediction. Its core benefits are highlighted in three aspects: First, the automatic judgment process based on abnormal threshold comparison can quickly identify model performance degradation, effectively avoiding the lag of manual intervention, and ensuring timely response to complex variables of wind and light power generation; Second, when an anomaly is detected, by reducing the learning rate and retraining, it not only prevents the risk of overfitting caused by drastic parameter fluctuations, but also progressively optimizes the model's performance in multi-source data fusion scenarios, improving the stability of power generation prediction; Finally, the design of automatically archiving abnormal data to the database exception area not only provides rich historical reference samples for subsequent model iteration, but also builds a closed-loop optimization system that continuously accumulates abnormal cases and spatio-temporal features related to wind and light power generation, constantly enhancing the model's generalization ability for complex working conditions. This combination of real-time evaluation, parameter adjustment, and data sedimentation not only ensures the reliability of wind and light power generation prediction results, but also enables the model to self-.

[0086] Further, if the abnormal value of the evaluation index of the adjusted prediction model is less than or equal to the evaluation index abnormal threshold, the parameters of the prediction model are retained, and the power generation prediction value output by the prediction model is obtained.

[0087] If the adjusted prediction model evaluation index abnormal value is greater than the evaluation index abnormal threshold, the spatiotemporal correlation features of the multi-source data set are stored in the database again in the abnormal area, and then the learning rate and the Dropout rate in the prediction model are reduced based on the adjusted prediction model evaluation index abnormal value, so as to obtain a re-adjusted prediction model evaluation index abnormal value.

[0088] If the re-adjusted prediction model evaluation index abnormal value is less than or equal to the evaluation index abnormal threshold, the parameters of the prediction model are retained, and the prediction model outputs the power generation prediction value.

[0089] If the re-adjusted prediction model evaluation index abnormal value is greater than the evaluation index abnormal threshold, a warning signal is generated.

[0090] The warning signal is generated to remind the staff through light and sound.

[0091] Based on the adjusted prediction model evaluation index abnormal value, the learning rate and the Dropout rate in the prediction model are reduced, the data fusion process abnormal index is divided into different level intervals in the database, each level interval corresponds to a different reduction value of the learning rate in the prediction model, the learning rate in the current prediction model is reduced by the reduction value of the learning rate in the prediction model to obtain the learning rate in the next prediction model, and the data fusion process abnormal index is divided into different level intervals in the database, each level interval corresponds to a different reduction value of the Dropout rate in the prediction model, the Dropout rate in the current prediction model is reduced by the reduction value of the Dropout rate in the prediction model to obtain the Dropout rate in the next prediction model. This method based on the adjusted prediction model evaluation index abnormal value to reduce the learning rate and the Dropout rate in the prediction model effectively suppresses the interference of data fluctuations on the prediction model and improves the stability and precision of wind power generation prediction in the multi-source heterogeneous data fusion scenario.

[0092] When adjusting the learning rate and the Dropout rate of the prediction model to reduce the evaluation index abnormal value, the adjustment of the learning rate usually has a more significant impact than the adjustment of the Dropout rate, because the learning rate directly controls the step size of parameter updates, and an excessively high learning rate can cause the optimization process to be unstable, making the model fluctuate violently or even diverge during training, thereby significantly affecting the evaluation index. The Dropout rate mainly improves the model generalization ability by preventing overfitting, and its adjustment effect is relatively mild. Therefore, if the prediction model evaluation index abnormal value is mainly caused by training instability, reducing the learning rate can more directly stabilize the model and reduce the prediction model evaluation index abnormal value; if the prediction model evaluation index abnormal value is caused by data noise or overfitting, adjusting the Dropout rate can be more effective, but overall, the adjustment of the learning rate has a more obvious and direct short-term impact on the prediction model evaluation index abnormal value.

[0093] Based on the adjusted prediction model evaluation index abnormal value reduction prediction model in learning rate and Dropout rate, the stability and generalization ability of the model can be improved by more conservative parameter adjustment strategy. Reducing the learning rate can slow down the step of parameter update, avoid the abnormal fluctuation of evaluation index caused by gradient shock, and more stably approach the optimal solution. Reducing the Dropout rate can retain more network nodes for training, enhance the learning ability of the model to the characteristics, and is especially suitable for the case that the abnormal value may be caused by overfitting or feature loss. This double adjustment can effectively suppress the amplitude of the evaluation index abnormal value, so that it is more likely to converge within the threshold range, while reducing the sensitivity of the model to noise or abnormal data. If the evaluation index abnormal value of the adjusted prediction model is still beyond the evaluation index abnormal threshold, it means that the problem may be caused by data quality or model structure itself. At this time, triggering the early warning signal can guide further investigation of the root cause.

[0094] By constructing an adaptive and multi-stage dynamic parameter optimization system, the performance of the model is precisely controlled through a hierarchical adjustment strategy. The core value lies in: adopting progressive optimization logic, firstly adjusting the learning rate for preliminary correction, and if it still does not meet the standard, further optimizing the learning rate and Dropout rate, which ensures the stability of training and enhances the generalization ability; At the same time, an abnormal data closed-loop feedback mechanism is established to continuously accumulate spatio-temporal feature data to enhance the adaptability of the model; Finally, an intelligent early warning threshold is set to terminate redundant calculation and prompt intervention in time when two optimizations are ineffective, effectively balancing the automation efficiency and result reliability. This structured adjustment process not only significantly improves the accuracy of power generation prediction, but also forms a complete intelligent operation and maintenance closed loop from parameter optimization to data accumulation to abnormal warning.

[0095] Figure 7 The flow chart of the relationship between the power generation difference and the actual power generation of the present application, which describes the closed-loop feedback adjustment mechanism of the power generation prediction system: the system first outputs the power generation prediction value, then collects the actual power generation and calculates the difference between the two; If the difference is within the power generation difference interval, the current model parameters remain unchanged; If the difference exceeds the power generation difference interval, the prediction value is stored in the difference area of the database, and the prediction model is adjusted accordingly to update its parameters to optimize the future prediction accuracy, thus forming a continuous closed-loop optimization mechanism.

[0096] Specifically, compared with the power generation prediction value output by the prediction model, the specific comparison process is: collecting the actual total power generation from the regional new energy control center.

[0097] Differential analysis is performed on the actual total power generation and the power generation prediction value output by the prediction model to obtain the power generation difference.

[0098] The absolute difference between the actual total power generation and the power generation prediction value output by the prediction model is calculated.

[0099] The power generation difference value is compared with the power generation difference value interval to determine whether to readjust the prediction model parameters.

[0100] By collecting the actual power generation data of the regional new energy centralized control center in real time and performing accurate difference analysis with the output value of the prediction model, the prediction deviation can be dynamically quantified to form a closed-loop feedback mechanism. Intelligent comparison of the power generation difference value with the power generation difference value interval can not only avoid frequent parameter adjustment of the model due to slight fluctuations, but also timely identify systematic deviation to provide an objective basis for parameter optimization. This method not only improves the adaptive ability of the model, but also continuously improves the prediction accuracy through data driving, ultimately realizes the dual improvement of power generation plan compilation efficiency and new energy consumption rate, and provides more reliable decision support for power grid dispatching.

[0101] Specifically, whether to readjust the prediction model parameters is determined, and the specific determination process is as follows: if the power generation difference value belongs to the power generation difference value interval, the prediction power generation result is accurate, and it is determined not to adjust the prediction model parameters.

[0102] If the power generation difference value does not belong to the power generation difference value interval, the power generation prediction value output by the prediction model is stored in the difference value area of the database, and then the learning rate and Dropout rate of the prediction model are adjusted based on the power generation difference value, so as to update the prediction model parameters.

[0103] The learning rate and Dropout rate of the prediction model are adjusted based on the power generation difference value, so as to update the prediction model parameters. The power generation difference value is divided into different level intervals in the database, and each level interval corresponds to a different reduction value of the learning rate in the prediction model. The learning rate in the current prediction model is reduced by the reduction value of the learning rate in the prediction model to obtain the learning rate in the next step of the prediction model. The power generation difference value is divided into different level intervals in the database, and each level interval corresponds to a different reduction value of the Dropout rate in the prediction model. The Dropout rate in the current prediction model is reduced by the reduction value of the Dropout rate in the prediction model to obtain the Dropout rate in the next step of the prediction model.

[0104] The power generation difference value interval is a quantitative range for evaluating the accuracy of the prediction model. It is determined by combining historical data statistical analysis, business tolerance and model performance dynamic adjustment. If the prediction and actual power generation difference value is within this interval, the result is considered accurate, otherwise the model parameter adjustment is triggered to realize closed-loop optimization.

[0105] By setting a reasonable power generation difference interval, normal fluctuations and significant deviations can be effectively distinguished, unnecessary parameter adjustment of the prediction model can be avoided, and the stability of the model can be improved. When the power generation difference exceeds the reasonable range, the system automatically stores the abnormal prediction value in the difference area of the database, which is convenient for subsequent analysis and model optimization. At the same time, the learning rate and the Dropout rate are dynamically adjusted, so that the model can adaptively balance the training speed and the generalization ability, and reduce the risk of overfitting or underfitting. This intelligent adjustment mechanism not only improves the adaptability and accuracy of the prediction model, but also ensures that the new energy power generation prediction remains highly reliable for a long time, providing more accurate data support for power grid dispatching and energy management.

[0106] Referring to Figure 2 The second aspect of the present application provides a multi-source data fusion intelligent prediction system for wind and light power generation, which comprises a multi-source data fusion module, a prediction model module, an intelligent optimization module and a database.

[0107] The multi-source data fusion module is connected with the prediction model module, and the prediction model module is connected with the intelligent optimization module. The multi-source data fusion module, the prediction model module and the intelligent optimization module are all connected with the database.

[0108] The database is used to store the parameters involved in the multi-source data fusion intelligent prediction system for wind and light power generation.

[0109] The multi-source data fusion module is used to collect the power generation performance parameters, environmental data and equipment operation state data monitored by the regional new energy control center in real time, preprocess the power generation performance parameters, environmental data and equipment operation state data respectively, fuse the preprocessed power generation performance parameters, environmental data and equipment operation state data, collect and analyze the data fusion process parameters, and judge whether to adjust the data fusion process, so as to obtain a precise multi-source data set.

[0110] The prediction model module is used to construct the spatio-temporal correlation feature through the multi-source data set, and input it into the power generation prediction model. The prediction model outputs the power generation prediction value, collects and analyzes the model error index, and judges whether to adjust the parameters of the prediction model, so as to intelligently optimize the accuracy of the power generation prediction model.

[0111] The spatio-temporal correlation feature represents that the equipment layout of the wind farm and the photovoltaic power station is modeled as a graph structure. The node represents the wind turbine or the photovoltaic array, and the edge represents the spatial correlation between the devices.

[0112] The intelligent optimization module is used to collect the actual total power generation, and compare it with the power generation prediction value output by the prediction model. Based on the comparison result, it is judged whether to readjust the parameters of the prediction model, so as to optimize the result of intelligently predicting the wind and light power generation in the prediction model.

[0113] The above merely illustrates and explains the structure of the present application, and those skilled in the art can make various modifications or supplements to the described specific embodiments or replace them with similar ways, as long as they do not deviate from the structure of the present application or exceed the scope defined by the present application, which shall belong to the protection scope of the present application.

Claims

1. A method for intelligent prediction of wind and solar power generation based on multi-source data fusion, characterized in that, The application relates to a method for intelligent optimization of a power generation prediction model. Step one: collecting real-time power generation performance parameters, environmental data and equipment operation state data monitored by a regional new energy centralized control center, respectively preprocessing the power generation performance parameters, the environmental data and the equipment operation state data, fusing the preprocessed power generation performance parameters, the environmental data and the equipment operation state data, collecting and analyzing data fusion process parameters, judging whether the data fusion process needs to be adjusted, and thus obtaining a precise multi-source data set; Step two: constructing a space-time correlation feature through the multi-source data set and inputting the space-time correlation feature into a power generation prediction model, the power generation prediction model outputting a power generation prediction value, collecting and analyzing model error indicators, judging whether the parameters of the prediction model need to be adjusted, and thus intelligently optimizing the precision of the power generation prediction model; The space-time correlation feature indicates that equipment layout of a wind farm and a photovoltaic power station is modeled as a graph structure, a node represents a fan or a photovoltaic array, and an edge represents spatial correlation between devices; Step three: collecting an actual total power generation and comparing the actual total power generation with the power generation prediction value output by the prediction model, judging whether the parameters of the prediction model need to be readjusted based on the comparison result, and thus optimizing the result of intelligently predicting wind and light power generation in the prediction model. 2.The method of claim 1, wherein: The judging whether the data fusion process needs to be adjusted includes the following steps: Collecting and analyzing data fusion process parameters to obtain a data fusion process anomaly index; The power generation performance parameters reflect real-time power generation capacity and efficiency of power generation equipment; The environmental data reflect external environmental factors that affect power generation performance and operation of power generation equipment; The equipment operation state data reflect the operation state of the power generation equipment; Comparing the data fusion process anomaly index with a data fusion process anomaly threshold value, if the data fusion process anomaly index is less than or equal to the data fusion process anomaly threshold value, storing the multi-source data set in a normal area of a database, and constructing a space-time correlation feature through the multi-source data set; The multi-source data set is a comprehensive data set formed after preprocessing, fusing and optimizing power generation performance parameters, environmental data and equipment operation state data monitored by a regional new energy centralized control center; If the data fusion process anomaly index is greater than the data fusion process anomaly threshold value, discarding this fusion, storing the preprocessed power generation performance parameters, the environmental data and the equipment operation state data in an abnormal area of a database, increasing an observation noise covariance matrix in a data fusion algorithm based on the data fusion process anomaly index, fusing the preprocessed power generation performance parameters, the environmental data and the equipment operation state data stored in the abnormal area of the database, collecting and analyzing data fusion process parameters again to obtain an adjusted data fusion process anomaly index, and thus judging whether the data fusion process needs to be adjusted again. 3.The method of claim 2, wherein: The judging whether the data fusion process needs to be adjusted again includes the following steps: If the adjusted data fusion process anomaly index is less than or equal to the data fusion process anomaly threshold value, storing the multi-source data set in a normal area of a database, and constructing a space-time correlation feature through the multi-source data set; If the adjusted data fusion process anomaly index is greater than the data fusion process anomaly threshold, the preprocessed power generation performance parameters, environmental data and equipment operating state data are stored again in the abnormal area of the database, the observation noise covariance matrix in the increased data fusion algorithm and the buffer size in the increased data fusion algorithm are increased based on the adjusted data fusion process anomaly index, the power generation performance parameters, environmental data and equipment operating state data in the abnormal area of the database are fused again to obtain a re-adjusted data fusion process anomaly index; If the re-adjusted data fusion process anomaly index is less than or equal to the data fusion process anomaly threshold, the multi-source data set is stored in the normal area of the database, and the spatiotemporal correlation features are constructed based on the multi-source data set; If the re-adjusted data fusion process anomaly index is greater than the data fusion process anomaly threshold, a warning is triggered. 4.The method of claim 1, wherein: The data fusion process parameters are collected and analyzed, and the specific analysis process is as follows: The data fusion process parameters include noise covariance of the data fusion process, coverage of the data fusion process and error covariance of the data fusion process; The calculation of the data fusion process anomaly index is based on the proportional relationship between the noise covariance of the data fusion process and the defined noise covariance, the proportional relationship between the defined coverage and the coverage of the data fusion process, and the proportional relationship between the error covariance of the data fusion process and the defined error covariance. The specific process is as follows: the defined noise covariance, the defined coverage and the defined error covariance contrast are used as the reference, the proportional relationship between the noise covariance of the data fusion process, the coverage of the data fusion process and the error covariance of the data fusion process and the corresponding defined values is calculated respectively, different measurement proportions are given according to the influence degree of each proportional relationship on the data fusion process anomaly index, and the data fusion process anomaly index is finally obtained by summarizing. The data fusion process anomaly index is used to quantify the comprehensive deviation degree of the data fusion result from the expected normal state. 5.The method of claim 1, wherein: The judgment of whether to adjust the parameters of the prediction model is as follows: The model error index is collected and analyzed to obtain the prediction model evaluation index abnormal value; The prediction model evaluation index abnormal value is compared with the evaluation index abnormal threshold, if the prediction model evaluation index abnormal value is less than or equal to the evaluation index abnormal threshold, the parameters of the prediction model are retained, and the power generation prediction value output by the prediction model is obtained; If the prediction model evaluation index abnormal value is greater than the evaluation index abnormal threshold, the spatiotemporal correlation features constructed based on the multi-source data set are stored in the abnormal area of the database, the learning rate in the prediction model is reduced based on the prediction model evaluation index abnormal value, the prediction model is retrained and verified based on the spatiotemporal correlation features constructed based on the multi-source data set, and the adjusted prediction model evaluation index abnormal value is obtained to determine whether to adjust the parameters of the prediction model again. 6.The method of claim 5, wherein: The judgment of whether to adjust the parameters of the prediction model again is as follows: If the adjusted prediction model evaluation index abnormal value is less than or equal to the evaluation index abnormal threshold, the parameters of the prediction model are retained, and the power generation prediction value output by the prediction model is obtained; If the adjusted prediction model evaluation index abnormal value is greater than the evaluation index abnormal threshold, the spatiotemporal correlation features of the multi-source data set are stored in the abnormal area of the database again, and then the learning rate and the Dropout rate in the prediction model are reduced based on the adjusted prediction model evaluation index abnormal value, so as to obtain a re-adjusted prediction model evaluation index abnormal value; If the re-adjusted prediction model evaluation index abnormal value is less than or equal to the evaluation index abnormal threshold, the parameters of the prediction model are retained, and the prediction model outputs the power generation prediction value; If the re-adjusted prediction model evaluation index abnormal value is greater than the evaluation index abnormal threshold, a warning signal is generated. 7.The method of claim 1, wherein: The model error index is collected and analyzed, and the specific analysis process is as follows: The model error index is collected and analyzed, and the model error index includes the power generation prediction model average absolute error, the power generation prediction model root mean square error, and the power generation prediction model mean square error; An abnormal terminal value of the data fusion process is obtained, and an abnormal increment of the data fusion process is matched from the database; The calculation of the prediction model evaluation index abnormal value is based on the proportional relationship between the power generation prediction model average absolute error and the defined average absolute error, the proportional relationship between the power generation prediction model root mean square error and the defined root mean square error, and the proportional relationship between the power generation prediction model mean square error and the defined mean square error. The specific process is as follows: the defined average absolute error, the defined root mean square error, and the defined mean square error are taken as the reference, the proportional relationship between the power generation prediction model average absolute error, the power generation prediction model root mean square error, and the power generation prediction model mean square error and the corresponding defined value is calculated, different measurement proportions are given according to the influence degree of each proportional relationship on the prediction model evaluation index abnormal value, and then the data fusion process abnormal increment is added to obtain the prediction model evaluation index abnormal value; The prediction model evaluation index abnormal value is used to quantify the abnormal degree of the overall performance of the prediction model. 8.The method of claim 1, wherein: The specific comparison process is as follows: The actual total power generation is collected from the regional new energy control center; The actual total power generation and the power generation prediction value output by the prediction model are analyzed to obtain a power generation difference value; The power generation difference value is compared with the power generation difference value interval to determine whether the prediction model parameters need to be adjusted. 9.The method of claim 8, wherein: The specific determination process is as follows: If the power generation difference value belongs to the power generation difference value interval, the prediction power generation result is accurate, and it is determined that the prediction model parameters do not need to be adjusted; If the power generation difference value does not belong to the power generation difference value interval, the power generation prediction value output by the prediction model is stored in the difference area of the database, and then the learning rate and the Dropout rate of the prediction model are adjusted based on the power generation difference value, so as to update the prediction model parameters.

10. A multi-source data fusion based intelligent prediction system for wind and solar power generation, characterized in that: The specific determination process is as follows: The multi-source data fusion module is used for collecting, in real time, power generation performance parameters, environmental data and equipment operation state data monitored by the regional new energy centralized control center, pre-processing the power generation performance parameters, the environmental data and the equipment operation state data respectively, fusing the pre-processed power generation performance parameters, the environmental data and the equipment operation state data, collecting and analyzing data fusion process parameters, judging whether to adjust the data fusion process, and thus obtaining a precise multi-source data set; The prediction model module is used for constructing a time-space correlation feature through the multi-source data set and inputting the time-space correlation feature into a power generation prediction model, the prediction model outputting a power generation prediction value, collecting and analyzing model error indexes, judging whether to adjust parameters of the prediction model, and thus intelligently optimizing the precision of the power generation prediction model; The time-space correlation feature is used for modeling equipment layout of a wind farm and a photovoltaic power station as a graph structure, a node representing a fan or a photovoltaic array, and an edge representing spatial correlation between equipment; The intelligent optimization module is used for collecting an actual total power generation and comparing the actual total power generation with the power generation prediction value output by the prediction model, judging whether to readjust the prediction model parameters based on a comparison result, and thus optimizing a result of intelligently predicting wind and light power generation in the prediction model.

Images