Multi-energy cooperative power generation prediction method and system based on multivariate data fusion

Abstract

The invention discloses a multi-energy cooperative power generation prediction method and system based on multivariate data fusion, and relates to the technical field of water conservancy and hydropower engineering, and the method comprises the steps: collecting meteorological, hydrological and load data, executing hour-by-hour alignment and abnormity correction, and building a wind-light-water multi-source data set; according to a low ebb, noon and peak division rule, calculating a time-sharing mean value of each energy source and performing seasonal correction to generate a power generation factor; and executing multi-energy collaborative prediction and scheduling based on the power generation factors, and updating power generation parameters by adopting a rolling verification and credibility mechanism. According to the method, the time sequence consistency and reliability of input data are improved, power generation factors are generated through time-sharing mean value calculation and seasonal correction, dynamic modeling and periodic correction of energy output characteristics are achieved, a dynamic prediction and self-adaptive updating mechanism of wind, light and water complementation is established, and the power generation efficiency is improved. Therefore, the accuracy and the scheduling response capability of multi-energy cooperative power generation prediction based on multivariate data fusion are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water conservancy and hydropower engineering, in particular to a multi-energy coordinated power generation prediction method and system based on multi-element data fusion. BACKGROUND

[0002] With the transformation of global energy structure to low carbonization and cleanization, renewable energy such as wind energy, solar energy and water energy gradually becomes an important part of the power system. New energy power generation has the characteristics of wide distribution, strong fluctuation, randomness and the like. The prediction accuracy of its output directly affects the safe operation and dispatching decision of the power system. In recent years, the prediction method has experienced the evolution process from statistical regression to machine learning, and then to multi-source fusion and intelligent dispatching. The traditional single-source prediction mode based on time series has been difficult to meet the requirements of new energy multi-scene operation characteristics. Research tends to use the collaborative analysis of meteorological, hydrological, geographical and load and other multi-source data to realize the dynamic prediction and optimization control of medium and short-term and intra-day power generation.

[0003] The existing multi-energy coordinated power generation prediction method based on multi-element data fusion mainly focuses on single energy or single factor modeling, such as constructing a wind power prediction model based only on meteorological parameters, or fitting photovoltaic power through light and temperature factors. This kind of method ignores the correlation of water energy regulation characteristics and load fluctuation, resulting in a significant decrease in prediction accuracy under multi-energy complementary operation conditions. At the same time, the existing data preprocessing methods generally have inconsistent time granularity, rough missing value processing and missing extreme weather event identification, etc. The input data set has deviations in time sequence consistency and spatial correlation. Traditional prediction models mostly use fixed time period average values or static weights, which are difficult to reflect the dynamic change law of new energy output in different operation periods and seasonal cycles. The existing dispatching system mostly takes independent energy prediction results as input, lacks cross-energy power factor fusion and rolling verification mechanism, and cannot realize real-time coordination and compensation regulation between wind, light and water. In summary, the existing technology has not established a unified data fusion framework and multi-energy coordinated prediction system, making it difficult to realize the continuous update and dispatching closed-loop control of power generation parameters, which limits the reliability and timeliness of new energy output prediction. SUMMARY

[0004] In view of the above problems, the present application is proposed.

[0005] Therefore, the technical problem solved by the present application is that the existing new energy power generation prediction technology has poor multi-source data coupling processing, low time period adaptability of prediction model, insufficient dispatching control feedback, and how to realize the fusion processing of meteorological, hydrological and load information under a unified data framework, and establish a multi-energy coordinated prediction and dispatching method with time rolling update and credibility identification mechanism.

[0006] To solve the above technical problems, the application provides the following technical scheme: a multi-energy collaborative power generation prediction method based on multi-element data fusion, comprising: collecting meteorological, hydrological and load data, performing hourly alignment and abnormal correction, and establishing a wind-solar-water multi-source data set; according to the low valley, noon and peak division rules, the average value of each energy is calculated and seasonally corrected to generate a power generation factor; based on the power generation factor, multi-energy collaborative prediction and scheduling are performed, and a rolling verification and credibility mechanism is used to update the power generation parameters.

[0007] As a preferred scheme of the multi-energy collaborative power generation prediction method based on multi-element data fusion, wherein: the step of performing hourly alignment and abnormal correction comprises: obtaining wind speed, wind direction, light intensity, solar radiation, environmental temperature, air humidity and precipitation from ground meteorological observation stations, satellite remote sensing platforms and power grid dispatching monitoring systems; when the sampling periods of different data sources are inconsistent, time aggregation is performed with the whole point time as the boundary to unify the hourly resolution; when short-time data is missing, linear interpolation compensation is performed using the observation values of the previous and next two valid time points; when the missing time length exceeds the interval of two hours, it is marked as an invalid section; a dual judgment rule based on physical rationality and statistical distribution threshold is used to identify abnormal points, and observation data exceeding the reasonable range are marked as invalid samples; when a typhoon, rainstorm or other extreme weather event is monitored, the current period data is retained and an extreme weather label is added; a standardized multi-source fusion data set containing original values, corrected values, time stamps and quality identifiers is generated as the input for the calculation of time average and power generation factor.

[0008] As a preferred scheme of the multi-energy collaborative power generation prediction method based on multi-element data fusion, wherein: the step of establishing a wind-solar-water multi-source data set comprises: establishing a double-layer spatial zoning rule according to terrain features and regional energy use attributes, dividing the target region into multiple spatial units, and determining each unit as an independent site cluster according to terrain type and energy use characteristics; each site cluster saves installed capacity, wind and light rejection rate, power grid upper limit value and historical data quality level; based on the wind and light output curves and load variation characteristics of the same month in the past ten years, the time boundaries of the low valley, noon and peak period are adaptively adjusted, and the boundary adjustment is automatically completed according to the matching criterion of minimizing the difference between the time load and the renewable energy output curve; when the fitting improvement degree brought by adaptive adjustment is lower than the set threshold, the original time period division is kept unchanged; after completing the time period optimization, a time-periodized regional site cluster list containing the identification, geographical location, installed capacity, boundary time and data quality level of each site cluster is output as the structured input for the calculation of time average and power generation factor.

[0009] As a preferred scheme of the multi-energy collaborative power generation prediction method based on multi-element data fusion, the calculation of the hourly average of each energy and the seasonal correction comprises: in the low valley, lunch and peak periods, the average output power of wind power generation, photovoltaic power generation and hydraulic power generation is respectively counted; the representative hourly average is obtained by using the same month samples of multiple years to perform moving average smoothing and abnormal sample elimination; the seasonal correction coefficient is calculated according to the relative deviation of the climatic periodic change and the main meteorological factor, and the correction coefficient is applied to the hourly average of each period to eliminate the seasonal influence; when the data missing ratio of any one of the low valley period, the lunch period or the peak period is more than 20%, the hourly average of the corresponding period is suspended, and the effective average result of the same period in the last statistical period is reserved as a substitute input; the corrected hourly average is integrated according to the installed capacity weight and the data quality weight to obtain the average output parameter of the regional station as the input reference of the power generation factor.

[0010] As a preferred scheme of the multi-energy collaborative power generation prediction method based on multi-element data fusion, the generation of the power generation factor comprises: comparing the hourly average after the seasonal correction with the average output of the regional station in the same period to calculate a ratio parameter reflecting the output proportion of different power generation periods; the extreme sample elimination and numerical normalization are performed on the ratio parameter; in each station cluster, the weighted average is calculated according to the installed capacity and the data quality score to obtain the intra-cluster power generation factor; the power grid receiving capacity weight is introduced at the cross-regional level, and the power generation factor of each region is corrected and integrated according to the historical power limiting rate and the wind and light curtailment rate to obtain the whole regional power generation factor; the change of the power generation factor of each station cluster is continuously monitored, and when the fluctuation amplitude in the last two statistical periods exceeds a preset threshold, the time period boundary re-estimation and data backtracking checking operation are automatically triggered.

[0011] As a preferred scheme of the multi-energy collaborative power generation prediction method based on multi-element data fusion, the multi-energy collaborative prediction and scheduling based on the power generation factor comprises: the correlation degree between the prediction output and the actual power generation is evaluated in a rolling time window manner, and the window length is not less than 24 months and is updated monthly; the prediction output and the measured output of each period are analyzed, the goodness of fit index is calculated, and the index value is divided into qualified, corrected and recalculated; when the goodness of fit is lower than the correction threshold, the model weight is automatically adjusted or the power generation factor parameter of the current period is corrected; the confidence interval range of the power generation factor is generated based on the repeated sampling algorithm, and the reliability label of each prediction result output includes the goodness of fit index, the sample size, the data quality score and the confidence boundary; when the proportion of extreme weather samples exceeds a preset proportion, an extreme influence identifier is automatically added in the label to trigger the derating operation strategy for the scheduling module.

[0012] As a preferred scheme of the multi-energy collaborative power generation prediction method based on multi-element data fusion, the rolling verification and credibility mechanism is used to update the power generation parameters, the available output upper limit of wind power and photovoltaic power in the low valley, midday and peak period is determined by using the full-area power generation factor, and a multi-energy scheduling model is established by combining the reservoir capacity, power grid capacity and ecological discharge constraint.

[0013] Another object of the present application is to provide a multi-energy collaborative power generation prediction system based on multi-element data fusion, which can generate a power generation factor by calculating the time-sharing average of each energy according to the low valley, midday and peak division rules and performing seasonal correction, thereby solving the problem of low time-sharing modeling accuracy and insufficient correction of seasonal changes in the current multi-energy collaborative power generation prediction technology based on multi-element data fusion.

[0014] As a preferred scheme of the multi-energy collaborative power generation prediction system based on multi-element data fusion, the system comprises a multi-source fusion module, a factor generation module and a scheduling optimization module, the multi-source fusion module is used to collect meteorological, hydrological and load data, perform hourly alignment and abnormal correction, and establish a wind-solar-water multi-source data set, the factor generation module is used to calculate the time-sharing average of each energy according to the low valley, midday and peak division rules and perform seasonal correction to generate a power generation factor, and the scheduling optimization module is used to perform multi-energy collaborative prediction and scheduling based on the power generation factor and update the power generation parameters by using the rolling verification and credibility mechanism.

[0015] A computer device comprises a memory and a processor, the memory stores a computer program, and the processor executes the computer program to realize the steps of the multi-energy collaborative power generation prediction method based on multi-element data fusion.

[0016] A computer readable storage medium stores a computer program, and the computer program is executed by a processor to realize the steps of the multi-energy collaborative power generation prediction method based on multi-element data fusion.

[0017] The beneficial effects of the present application: the multi-energy collaborative power generation prediction method based on multi-element data fusion provided by the present application collects meteorological, hydrological and load data, performs hourly alignment and abnormal correction, establishes a wind-solar-water multi-source data set, realizes unified fusion and abnormal correction of multi-source data, guarantees the time sequence consistency and reliability of the prediction input, calculates the time-sharing average of each energy according to the low valley, noon and peak division rules and performs seasonal correction to generate a power factor, constructs a time-sharing seasonal correction model, generates a power factor, quantifies the multi-energy output characteristics, performs multi-energy collaborative prediction and scheduling based on the power factor, updates the power generation parameters using a rolling verification and credibility mechanism, establishes a rolling verification and dynamic scheduling mechanism, realizes multi-energy collaborative prediction and parameter self-updating, and the present application achieves better results in multi-source data fusion accuracy, time-sharing power generation feature modeling accuracy and multi-energy collaborative prediction and scheduling stability. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0019] Figure 1 The overall flowchart of a multi-energy collaborative power generation prediction method based on multi-element data fusion provided by the present application embodiment 1.

[0020] Figure 2 The overall schematic diagram of a multi-energy collaborative power generation prediction system based on multi-element data fusion provided by the present application embodiment 2. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0022] Embodiment 1, refer to Figure 1 For an embodiment of the present application, a multi-energy collaborative power generation prediction method based on multi-element data fusion is provided, comprising: S1: Collect meteorological, hydrological and load data, perform hourly alignment and abnormal correction, and establish a wind-solar-water multi-source data set.

[0023] Further, the performing of the hourly alignment and abnormal correction comprises: obtaining wind speed, wind direction, light intensity, solar radiation, ambient temperature, air humidity, precipitation, and hourly divided power load data from ground meteorological observation stations, satellite remote sensing platforms, and power grid dispatching monitoring systems; when different data sources have inconsistent sampling periods, time aggregation is performed with the whole point time as the boundary to unify the hourly resolution; when short time data is missing, linear interpolation compensation is performed using the observation values of the previous and subsequent two valid time points; when the missing time length exceeds the interval of two hours, it is marked as an invalid section; a dual determination rule based on physical rationality and statistical distribution threshold is used to identify abnormal points, and observation data that exceeds the reasonable range is marked as an invalid sample; when extreme weather events such as typhoons and heavy rains are monitored, the current period data is retained and an extreme weather label is added; a standardized multi-source fusion data set containing original values, corrected values, timestamps, and quality identifiers is generated as the input for the calculation of the hourly average and the generation factor.

[0024] It should be noted that one preferred solution of performing the hourly alignment and abnormal correction specifically comprises calculating the linear interpolation compensation , which is expressed as: wherein, the interpolated sample value at the th whole point time in the missing report interval, is the interpolation node, is the value of the index before the start of the missing report section, is the value of the index after the end of the missing report section, is the th whole point timestamp in the missing report interval, is the left end point timestamp (corresponding to the time of ), is the right end point timestamp (corresponding to the time of ).

[0025] It should be noted that the establishment of the wind, light and water multi-source data set includes establishing a double-layer spatial zoning rule according to the terrain characteristics and regional energy consumption attributes, dividing the target region into a plurality of spatial units, and determining each unit as an independent site cluster according to the terrain type and energy consumption characteristics; each site cluster saves the installed capacity, wind and light rejection rate, power grid upper limit value and historical data quality level; based on the wind and light output curve and load change characteristics of the same month in the past ten years, the time boundaries of the low valley, noon and peak period are adaptively adjusted, and the boundary adjustment is automatically completed according to the matching criterion of minimizing the difference between the time load and the renewable energy output curve; when the fitting improvement degree brought by adaptive adjustment is lower than the set threshold, the original time period division is kept unchanged; after completing the time period optimization, the time periodized regional site cluster list containing the site cluster identification, geographical position, installed capacity, boundary time and data quality level is output, which is used as the structured input of time-sharing average calculation and power generation factor calculation.

[0026] It should be noted that by collecting meteorological, hydrological and load data, performing hourly alignment and anomaly correction, establishing a wind, light and water multi-source data set, unified time sequencing and standardized processing of data of different sources and different sampling frequencies are realized; by setting the integral time aggregation and missing report linear interpolation mechanism, the consistency of meteorological, hydrological and load data in the time dimension is ensured; and by physical constraint and statistical threshold double judgment to remove abnormal samples, the input data is more stable and reliable in statistical distribution. At the same time, special labels are set for typhoon, rainstorm and other extreme weather periods to retain special working condition characteristics and avoid excessive smoothing of key disturbance information in the data cleaning process, providing high consistency and high quality input samples for subsequent power generation factor calculation and multi-energy collaborative prediction, so that the wind, light and water multi-energy data can be related in a unified time framework. Modeling, thereby realizing effective guarantee of the time sequence consistency and data integrity of the power generation prediction model.

[0027] S2: According to the low valley, noon and peak division rule, the time-sharing average of each energy is calculated and seasonally corrected to generate the power generation factor.

[0028] Further, the calculation of the time-sharing average of each energy and the seasonal correction includes, in the low valley, lunch and peak period, the average output power of wind power generation, photovoltaic power generation and hydroelectric power generation is counted respectively; using multiple years of the same month samples to carry out moving average smoothing and abnormal sample elimination, to obtain the representative time-sharing average; according to the relative deviation of the climatic periodic change and the main meteorological factor, the seasonal correction coefficient is calculated, the seasonal correction coefficient is calculated based on the multiple years of the same month median ratio method, and is limited in the range of [0.85, 1.15], when the corrected time-sharing average deviates from the average value of the same month in the past three years by more than 20%, the uncorrected value is used as the replacement input; when the data missing rate of any one of the low valley period, the lunch period or the peak period is more than 20%, the time-sharing average of the corresponding period is suspended, and the effective average value of the same period in the last statistical period is retained as the replacement input; the corrected time-sharing average is integrated according to the installed capacity weight and the data quality weight to obtain the average output parameter of the regional site, which is used as the input reference of the power generation factor.

[0029] It should be pointed out that the generation of the power generation factor includes comparing the time-sharing average after seasonal correction with the average output of the regional site in the same period, calculating the ratio parameter reflecting the output proportion of different power generation periods; the extreme sample elimination and numerical normalization are performed on the ratio parameter; in each site cluster, the weighted average is calculated according to the installed capacity and the data quality score to obtain the intra-cluster power generation factor; the grid accommodation capacity weight is introduced at the cross-regional level, and the power generation factor of each region is corrected and integrated according to the historical power limiting rate and the wind and light curtailment rate to obtain the regional power generation factor; the change of the power generation factor of each site cluster is continuously monitored, and when the fluctuation amplitude in the last two statistical periods exceeds the preset threshold, the time period boundary reestimation and data backtracking verification operation are automatically triggered.

[0030] It should also be pointed out that one preferred scheme of calculating the time-sharing average of each energy and the seasonal correction specifically includes, according to the power generation prediction, the low valley period, the lunch period and the peak period are divided into three periods, and the average values of the three periods are calculated, which are represented as: Wherein, is the low valley period power generation prediction value, is the total number of valid samples in the low valley period, which represents the th valid hour sample in a certain fixed period, is the discrete sample number, is the number of valid samples participating in the average operation in the period, represents the low valley period th cleaned valid sample power value.

[0031] wherein, is the power generation prediction value of the midday period, is the total number of valid samples in the midday period, represents the cleaned valid sample power value of the th in the midday period.

[0032] wherein, is the power generation prediction value of the peak period, is the total number of valid samples in the peak period, represents the cleaned valid sample power value of the th in the peak period.

[0033] It should also be noted that one preferred scheme for generating the power generation factor specifically includes calculating a ratio parameter reflecting the output proportion of different power generation periods, denoted as: wherein, is the power generation factor, is the time-of-use average of the corresponding period, is the regional site average output parameter, serving as the reference quantity of the normalization denominator.

[0034] The total regional power generation factor is calculated and denoted as: wherein, is the power generation factor of a single site cluster, represents the number of site clusters participating in the average.

[0035] It should also be noted that by calculating the time-of-use average of each energy source and performing seasonal correction according to the low valley, midday, and peak division rules, the power generation factor is generated, achieving periodized modeling of power output and seasonal periodic correction. First, the three typical power generation periods are determined according to the regional load characteristics and historical output rules, and the average output of wind power, photovoltaic power, and hydroelectric power is statistically divided by time to form representative power distribution curves. Subsequently, the correction coefficient is calculated through the climate cycle factor, and the seasonal deviations of temperature, illumination, wind speed, and other meteorological parameters are mapped to the power generation correction amount, so that the time-of-use average can reflect long-term climate change and operation cycle characteristics. On this basis, the power generation factor is established as a comprehensive index of different energy output levels, enabling comparison and weighting of multiple energy sources in the same dimension, effectively associating meteorological characteristics of multiple time scales with energy output statistics, providing a quantitative parameter basis for multi-energy complementary dispatching, and providing dynamic correction basis for subsequent rolling prediction, enabling the prediction model to maintain stability and traceability in seasonal changes and period switching.

[0036] S3: Perform multi-energy collaborative prediction and dispatching based on the power generation factor, and update the power generation parameters using a rolling verification and credibility mechanism.

[0037] Furthermore, the multi-energy collaborative prediction and scheduling based on power generation factors includes: using a rolling time window to evaluate the correlation between predicted output and actual power generation, with a window length of no less than 24 months and updated monthly; performing error analysis on the predicted and measured outputs for each period, calculating the goodness-of-fit index, and classifying the index values ​​as qualified, requiring correction, or requiring recalculation; automatically adjusting model weights or correcting the power generation factor parameters for the current period when the goodness-of-fit is below the correction threshold; generating confidence intervals for power generation factors based on a repeated sampling algorithm, and outputting a confidence label for each prediction result that includes the goodness-of-fit index, sample size, data quality score, and confidence boundary; and automatically adding an extreme impact indicator to the label when the proportion of extreme weather samples exceeds a preset ratio, allowing the scheduling module to trigger a reduced-rate operation strategy.

[0038] It should be noted that the rolling verification and credibility mechanism for updating power generation parameters includes: using the regional power generation factor to determine the available output limits of wind and solar power during off-peak, midday, and peak periods; and establishing a multi-energy dispatch model in conjunction with reservoir capacity, grid capacity, and ecological discharge constraints. When real-time load demand exceeds the available output limits of wind and solar power and exceeds the regional peak load threshold, the hydropower compensation module is triggered to gradually increase hydropower output based on the unit's adjustable capacity and the current water level margin. Daily rolling updates are performed on a three-hour cycle, with each update recalculating the wind, solar, and hydropower output allocation based on the latest power generation factor and load forecast data. When the extreme weather tag is detected to be active, the wind power output limit is automatically reduced proportionally to 70% of the normal value, and the solar power output limit is reduced to 60% of the normal value, with priority given to hydropower output allocation. After the extreme weather condition is resolved, the regular dispatch plan is restored based on the power generation factor parameters of the most recent stable cycle. When operating across regions, the grid acceptance capacity coefficient is used as a soft constraint to dynamically adjust the output allocation in each region, and the time, magnitude, and parameters of each dispatch adjustment are recorded in a log file.

[0039] It should also be noted that a preferred scheme for updating power generation parameters using rolling verification and a reliability mechanism specifically includes setting up a rolling verification mechanism and calculating the degree of linear correlation between the prediction results and the observed data, expressed as: in, express The correlation coefficient at time t is used to characterize the degree of linear correlation between the predicted result and the observed data. The value ranges from [-1, 1]. The closer the value is to 1, the higher the degree of agreement between the predicted value and the actual value. For time indexing, Indicates at time The predicted power generation value, Indicates at time The actual observed power generation value.

[0040] Define window length , is represented as: in, In time index The above is a set of indices for the rolling time window used for modeling and evaluation. Indicates the window length. The specified window length must cover at least 24 hours, ensuring that each window contains time-series information for at least one complete day and night.

[0041] The goodness-of-fit index is calculated and expressed as: in, Indicates the recalculation threshold, when If the value falls below this threshold, it indicates that the prediction significantly deviates from reality, triggering a recalculation of the "power generation factor" (an indicator based on time-period averages and site averages) to update the key quantitative factors for long-term forecasting. This indicates the fine-tuning threshold, when Between and The value between 0 and 1 indicates that the model has a certain level of accuracy but can still be improved, triggering fine-tuning of the model weights (such as minor adjustments to the output ratio of wind / solar / hydropower sources or feature weighting) to improve the fit.

[0042] Calculate the mean of the power generation factor , is represented as: in, Indicates the first Each power generation factor sample value, This represents the number of samples for the power generation factor.

[0043] Calculate the standard deviation of the power generation factor , is represented as: Calculate the variable confidence level , is represented as: in, Represents the overall mean of the power generation factor The confidence interval is used to measure the uncertainty of the estimated generation factor, serving as a basis for forecast reliability assessment and threshold trigger adjustment. The significance level range is specified (usually 0.1, 0.05, or 0.01).

[0044] Calculate the current time of hydraulic power generation , expressed as wherein, is the adjustable online power of the hydroelectric generating set, indicates actual demand, indicates the current wind power output, indicates the current photovoltaic power output.

[0045] Calculate the water level and ecological constraints, expressed as: wherein, indicates the current reservoir water level, is the water flow, is the water density, is the acceleration of gravity, is the unit efficiency coefficient.

[0046] It should also be noted that by performing multi-energy collaborative prediction and scheduling based on generation factors, updating generation parameters using rolling verification and credibility mechanisms, dynamic collaborative prediction and real-time scheduling update of multi-source new energy output are realized. In the prediction link, a rolling time window is introduced to continuously test the correlation between the predicted value and the measured value, and the model weight and factor parameters are automatically adjusted according to the goodness of fit result. At the scheduling level, the upper limit of wind and light output, water energy compensation ability and grid acceptance constraints are unified into a multi-energy scheduling model to realize dynamic coupling and complementary regulation of wind, light and water energy. Through the credibility identification mechanism, the system can independently evaluate extreme weather samples and trigger the derating operation strategy to ensure that the power generation plan still has robustness under uncertain conditions. At the same time, the three-hour period of intra-day rolling update enables the scheduling model to have adaptive correction ability, forming a "prediction-verification-update" closed loop, realizing the leap from static prediction to dynamic optimization, and enabling the new energy power generation system to have continuous self-correction and scheduling coordination ability in the multi-energy complementary scene.

[0047] Embodiment 2, refer to Figure 2 As an embodiment of the present application, a multi-energy collaborative power generation prediction system based on multi-element data fusion is provided, comprising a multi-source fusion module, a factor generation module, and a scheduling optimization module.

[0048] The multi-source fusion module is used to collect meteorological, hydrological and load data, perform hourly alignment and abnormal correction, and establish a wind, light and water multi-source data set.

[0049] The factor generation module is used to calculate the time-sharing average of each energy according to the low valley, midday and peak division rules and perform seasonal correction to generate the generation factor.

[0050] The scheduling optimization module is used for performing multi-energy collaborative prediction and scheduling based on the power generation factor, and updating the power generation parameter by using a rolling verification and credibility mechanism.

Claims

1. A multi-energy coordinated power generation prediction method based on multi-element data fusion, characterized in that, Comprise: By collecting meteorological, hydrological and load data, hourly alignment and abnormal correction are performed to establish a wind-solar-water multi-source dataset; According to the valley, noon and peak division rules, the average of each energy is calculated and seasonally corrected to generate the generation factor; Based on the generation factor, multi-energy collaborative prediction and scheduling are performed, and the rolling verification and credibility mechanism is used to update the generation parameters.

2. The multi-element data fusion based multi-energy coordinated power generation prediction method of claim 1, wherein: The execution of hourly alignment and abnormal correction includes, Obtain wind speed, wind direction, light intensity, solar radiation, environmental temperature, air humidity, and precipitation from ground meteorological observation stations, satellite remote sensing platforms, and power grid dispatching monitoring systems; When the sampling periods of different data sources are inconsistent, time aggregation is performed with the whole point time as the boundary to unify the hourly resolution; When short-time data is missing, linear interpolation compensation is used with the observation values of the two effective time points before and after it. When the missing duration exceeds two hours, it is marked as an invalid section; A dual-determination rule based on physical rationality and statistical distribution threshold is used to identify abnormal points, and observation data that exceed the reasonable range are marked as invalid samples; When an extreme weather event is monitored, the current period data is retained and an extreme weather label is added; A standardized multi-source fusion dataset containing original values, corrected values, timestamps, and quality identifiers is generated as input for the calculation of time-average and generation factor. 3.The multi-element data fusion based multi-energy cooperative power generation prediction method of claim 2, wherein: The establishment of the wind-solar-water multi-source dataset includes, According to the terrain characteristics and regional energy use attributes, a double-layer spatial zoning rule is established to divide the target area into multiple spatial units, and each unit is determined as an independent station cluster according to the terrain type and energy use characteristics; Each station cluster saves installed capacity, wind and light rejection rate, grid acceptance upper limit value, and historical data quality level; Based on the wind and light output curves and load variation characteristics of the same month in the past ten years, the time boundaries of the valley, noon, and peak periods are adaptively adjusted. The boundary adjustment is automatically completed according to the matching criterion of minimizing the difference between the time load and the renewable energy output curve; When the fitting improvement degree brought by adaptive adjustment is lower than the set threshold, the original time period division remains unchanged; After completing the time period optimization, a time-periodized regional station cluster list containing the identification, geographical location, installed capacity, boundary time, and data quality level of each station cluster is output as a structured input for time-average calculation and generation factor calculation.

4. The multi-element data fusion based multi-energy coordinated power generation prediction method of claim 3, wherein: The calculation of the average of each energy and the seasonal correction includes, In the valley, noon and peak periods, the average output power of wind power generation, photovoltaic power generation and hydroelectric power generation is respectively calculated; Use multiple years of the same month samples to perform moving average smoothing and abnormal sample removal to obtain representative time-average; According to the relative deviation of the climate periodicity and the main meteorological factors, the seasonal correction coefficient is calculated, and the correction coefficient is applied to the time-average of each period to eliminate the seasonal influence; When the data missing rate of any of the valley period, noon period or peak period exceeds 20%, the time-average update of the corresponding period is suspended, and the valid average result of the same period in the last statistical period is retained as the alternative input; The time-sharing average is modified and integrated according to the installed capacity weight and the data quality weight to obtain the average output parameter of the regional station as the input reference of the power generation factor.

5. The multi-element data fusion based multi-energy coordinated power generation prediction method of claim 4, wherein: The generating power generation factor comprises, The time-sharing average modified by the season is compared with the average output of the regional station in the same period to calculate a ratio parameter reflecting the output proportion in different power generation periods; The ratio parameter is subjected to extreme sample elimination and numerical normalization; In each station cluster, a weighted average is calculated according to the installed capacity and the data quality score to obtain the power generation factor in the cluster; The power grid accommodation capacity weight is introduced at the cross-regional level, and the power generation factor of each region is modified and integrated according to the historical power limiting rate and the wind and light curtailment rate to obtain the power generation factor of the whole region; The change of the power generation factor of each station cluster is continuously monitored, and when the fluctuation amplitude exceeds a preset threshold in two consecutive statistical periods, the time period boundary re-estimation and data backtracking verification operation are automatically triggered. 6.The multi-element data fusion based multi-energy cooperative power generation prediction method according to claim 5, characterized in that: The multi-energy collaborative prediction and scheduling based on the power generation factor comprises, The correlation degree between the prediction output and the actual power generation is evaluated in a rolling time window manner, and the window length is not less than 24 months and is updated monthly; The prediction output and the measured output of each period are subjected to error analysis, a goodness-of-fit index is calculated, and the index value is divided into qualified, needing correction and needing recalculation; When the goodness-of-fit is lower than the correction threshold, the model weight is automatically adjusted or the power generation factor parameter of the current period is modified; Based on the repeated sampling algorithm, a confidence interval range of the power generation factor is generated, and a reliability label including the goodness-of-fit index, the sample size, the data quality score and the confidence boundary is output for each prediction result; When the proportion of extreme weather samples exceeds a preset proportion, an extreme influence identifier is automatically added in the label to trigger the derating operation strategy for the scheduling module.

7. The multi-element data fusion based multi-energy coordinated power generation prediction method of claim 6, wherein: The rolling verification and reliability mechanism is used to update the power generation parameter, comprising, The available output upper limit of wind power and photovoltaic power in the valley, noon and peak periods is determined by using the regional power generation factor, and a multi-energy scheduling model is established by combining the reservoir capacity, the power grid capacity and the ecological discharge constraint; When the real-time load demand exceeds the available output upper limit of wind and light and exceeds the regional peak load threshold, the water energy compensation module is triggered, and the water energy power generation output is gradually increased according to the adjustable capacity of the unit and the current water level surplus; The intra-day rolling update is performed in a three-hour cycle, and the wind, light and water output distribution is recalculated based on the latest power generation factor and load prediction data each time; When the extreme weather label is activated, the available output upper limit of wind power is automatically reduced to 70% of the normal value, the available output upper limit of photovoltaic power is reduced to 60% of the normal value, and the water energy output is preferentially allocated; When the extreme weather state is removed, the normal scheduling plan is restored according to the power generation factor parameters in the last stable period; When running across regions, the power grid accommodation capacity coefficient is used as a soft constraint to dynamically adjust the output distribution of each region, and the time, amplitude and parameters of each scheduling adjustment are recorded in a log file.

8. A multi-energy coordinated power generation prediction system based on multi-element data fusion, adopting the multi-energy coordinated power generation prediction method based on multi-element data fusion according to any one of claims 1-7. It comprises a multi-source fusion module, a factor generation module and a scheduling optimization module; The multi-source fusion module is used to collect meteorological, hydrological and load data, perform hourly alignment and abnormal correction, and establish a wind, light and water multi-source data set; The factor generation module is configured to calculate time-sharing average of each energy and generate generation factor by performing seasonal correction according to low valley, midday and peak division rules; The scheduling optimization module is configured to perform multi-energy collaborative prediction and scheduling based on the generation factor, and update generation parameters by using rolling verification and credibility mechanism. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor, when executing the computer program, implements the steps of the multi-energy collaborative generation prediction method based on multi-element data fusion in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the multi-energy collaborative generation prediction method based on multi-element data fusion in any one of claims 1 to 7.

Images