A new energy project short-term production capacity prediction and driving diagnosis method

Abstract

The application discloses a new energy project short-term production capacity prediction and driving diagnosis method, relates to the technical field of production capacity prediction, and comprises the following steps: performing phase-in reliable solidification on an environment factor candidate time lag list, eliminating invalid time difference caused by accidental disturbance, generating an environment factor production capacity response time lag fingerprint, correcting a short-term production capacity prediction window based on the environment factor production capacity response time lag fingerprint, and performing operation boundary reliable gating to obtain a short-term production capacity prediction result; constructing a counterfactual production capacity diagnosis scene according to the short-term production capacity prediction result, separating independent influence by restoring the stable state of any environment factor, obtaining a counterfactual production capacity diagnosis result, and performing difference diagnosis on the counterfactual production capacity diagnosis result and the short-term production capacity prediction result to generate a short-term production capacity driving contribution record. The application realizes stable expression of the response delay relationship between the environment factor change segment and the production capacity time sequence change segment by generating the environment factor production capacity response time lag fingerprint.

Description

Technical Field

[0001] This invention relates to the field of capacity forecasting technology, and in particular to a method for short-term capacity forecasting and drive diagnosis of new energy projects. Background Technology

[0002] Short-term capacity forecasting for new energy projects has continued to develop as the scale of wind power and photovoltaic grid connection expands. Related research has gradually expanded from single historical power extrapolation to multi-source time series modeling that integrates meteorological forecasts, on-site measurements, satellite cloud images, ground-based cloud images, and equipment operation data. Forecasting methods have also evolved from statistical regression and physical modeling to machine learning, deep learning, and spatiotemporal feature fusion models to support short-term power reporting, scheduling plans, and operation analysis.

[0003] In the short-term capacity forecasting of new energy projects, the effects of environmental factors on capacity time series typically exhibit phase differences, response lags, and coupling disturbances. Simply using multi-source environmental data as forecast inputs makes it difficult to stably express the temporal relationship between changes in environmental state and changes in capacity, and also makes it difficult to explain which environmental factor triggers the changes in the forecast curve. Although time series counterfactual interpretation has been used in forecast interpretation research, phase constraints, time lag fingerprints, and driving contribution closed loops for new energy projects still need further refinement. Summary of the Invention

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

[0005] Therefore, this invention provides a method for short-term capacity forecasting and driver diagnosis of new energy projects to solve the problems that the time-delay response of environmental factors is difficult to solidify for short-term capacity forecasting and the source of capacity fluctuations is difficult to separate and diagnose.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for short-term capacity prediction and driver diagnosis of new energy projects. The method includes: calibrating the same time axis based on the new energy project's capacity time series and a set of multi-source environmental factors; implementing credible labeling for abnormal sampling to obtain capacity-environment co-window sampling records; isolating non-environmental interference and verifying the sampling credibility of the capacity-environment co-window sampling records; filtering out invalid driver segments to obtain a credible capacity-environment sample set; identifying environmental phases and binding capacity response features based on the credible capacity-environment sample set to form an environmental phase capacity response map; conducting multi-time-window response time delay scanning around the environmental phase capacity response map; comparing environmental factor change segments in the multi-source environmental factor set with capacity time series change segments to generate a candidate time delay list for environmental factors. The candidate time delay list of environmental factors is reliably solidified within the phase, and invalid time differences caused by occasional disturbances are eliminated to generate environmental factor capacity response time delay fingerprints. Based on the environmental factor capacity response time delay fingerprints, the short-term capacity prediction window is corrected and the operational boundary is reliably gated to obtain the short-term capacity prediction results. A counterfactual capacity diagnosis scenario is constructed based on the short-term capacity prediction results. By restoring the steady state of any environmental factor and separating its independent influence, the counterfactual capacity diagnosis results are obtained. The difference between the counterfactual capacity diagnosis results and the short-term capacity prediction results is diagnosed to generate short-term capacity driving contribution records. Based on the short-term capacity driving contribution records, closed-loop archiving and write-back correction are performed, and the prediction deviation is fed back to the next prediction period to form a closed-loop archive of short-term capacity prediction and driving diagnosis.

[0007] As a preferred embodiment of the short-term capacity forecasting and driving diagnosis method for new energy projects described in this invention, the steps for obtaining the capacity environment window sampling record are as follows: A unified sampling sequence is established using the grid connection point sampling time in the new energy project capacity time series as the master clock. The new energy project capacity time series is mapped to the unified sampling sequence. Each environmental factor in the multi-source environmental factor set is calibrated and written into the corresponding unified sampling time according to the actual sampling time to generate an initial time alignment sequence. The sampling confidence level of each unified sampling moment is calculated based on the initial time alignment sequence. An abnormal sampling confidence mark is added to the sampling moment when the sampling confidence level is lower than the preset confidence threshold. The corresponding production capacity time series value, environmental factor calibration value and sampling confidence level are bound together to obtain the production capacity and environment window sampling record.

[0008] As a preferred embodiment of the short-term capacity prediction and driving diagnosis method for new energy projects described in this invention, the steps for obtaining a reliable sample set of capacity environment are as follows: Read the limited-issue status, maintenance status, equipment offline status and communication abnormal status corresponding to each unified sampling time from the same sampling record of production capacity and environment. Classify the same sampling pieces with the matching status into non-environmental isolation segments, and form the same sampling pieces with the unmatched status into an environment-driven set of segments to be verified. The set of environmental-driven segments to be verified is subjected to sampling credibility verification and continuity verification. Segments whose sampling credibility meets the preset credibility threshold and whose continuous effective duration meets the preset minimum driving observation duration are identified as valid driving segments. The remaining invalid driving segments are screened out to obtain a reliable sample set of production capacity environment.

[0009] As a preferred embodiment of the short-term capacity prediction and drive diagnosis method for new energy projects described in this invention, the steps for forming the environmental phase capacity response map are as follows: Read continuous and effective driving segments from the set of reliable samples of production capacity and environment, and match the direction of production capacity change, the continuous state of production capacity, the state of production capacity recovery and the state of environmental factor change in the continuous and effective driving segments to generate a set of phase candidate segments. Based on the phase candidate fragment set, phase candidate fragments that continuously correspond to the changes in production capacity and the changes in environmental factors are grouped into the same environmental phase. The environmental phase is used as the phase node, the environmental factors in the multi-source environmental factor set are used as the factor node, and the production capacity response feature composed of the direction of production capacity change, the continuous state of production capacity, and the state of production capacity recovery is used as the response node. Phase factor correlation edges are established between phase nodes and factor nodes, factor response correlation edges are established between factor nodes and response nodes, and phase response correlation edges are established between phase nodes and response nodes to form an environmental phase production capacity response map.

[0010] As a preferred embodiment of the short-term capacity prediction and driving diagnosis method for new energy projects described in this invention, the steps for generating a candidate time-lag list of environmental factors are as follows: Based on the environmental phase energy production response map, environmental factor change segments under the same environmental phase are extracted as time-delay scanning objects by preset short windows, medium windows and long windows with increasing time lengths. For time-lag scanning objects, the environmental factor change segments are gradually shifted relative to the production capacity time-series change segments along the time axis forward direction, and the correspondence formed by each shift is written into the time-lag correspondence set to be evaluated. Phase-locked time delay matching is performed on the set of time delay correspondences to be evaluated to generate phase-locked time delay matching strength. The correspondences that meet the preset candidate time delay conditions are associated and encapsulated with environmental phase, environmental factors, candidate time delays, phase-locked time delay matching strength and production capacity response characteristics to generate a list of environmental factor candidate time delays.

[0011] As a preferred embodiment of the short-term capacity prediction and driving diagnosis method for new energy projects described in this invention, the steps for generating the environmental factor capacity response time-delay fingerprint are as follows: Read the candidate options corresponding to the same environmental phase, the same environmental factor, and the same candidate time delay from the list of candidate time delays for environmental factors, and retrieve the sampling credibility of the candidate options from the set of reliable samples for production capacity environment. Write the production capacity response characteristics, phase-locked time delay matching strength, and sampling credibility associated with the candidate options into the candidate time delay group within the phase. Repeated hit verification and response consistency verification are performed on candidate time delay groups within the phase. Candidate time delays that repeatedly hit in multiple consecutive effective drive segments in the same environmental phase and whose production capacity response characteristics remain consistent are confirmed as effective response time delays. Candidate time delays that cannot be repeatedly hit are eliminated as invalid time differences caused by occasional disturbances. The effective response time delays are associated and encapsulated with the corresponding environmental phase, environmental factor, production capacity response characteristics and phase locking time delay matching strength to generate environmental factor production capacity response time delay fingerprints.

[0012] As a preferred embodiment of the short-term capacity forecasting and drive diagnosis method for new energy projects described in this invention, the steps for obtaining the short-term capacity forecasting result are as follows: Call the environmental factor capacity response time lag fingerprint, map the environmental factors in the multi-source environmental factor set used for future short-term prediction to the capacity prediction period according to the corresponding effective response time lag, and write the current continuous capacity time series, project available capacity and equipment operating status into the same prediction window to generate a time lag correction prediction window; Based on the time-delay correction prediction window and the capacity response characteristics in the capacity response time-delay fingerprint of environmental factors, the change in capacity time series values ​​with the same environmental phase, the same environmental factor, the same effective response time delay and consistent capacity response characteristics are extracted from the reliable sample set of capacity environment. The predicted increment for each capacity prediction period is determined, and the predicted capacity value is formed by the end capacity time series value of the current continuous capacity time series for each period, generating candidate short-term capacity prediction curves. The candidate short-term capacity forecast curves are subjected to operational boundary confidence gating. Capacity forecast segments that exceed the available capacity of the project are locked at the upper limit. Capacity forecast segments that do not meet the equipment ramp-up boundary are subject to smoothing constraints. Low confidence markers are added to capacity forecast segments with insufficient phase-locked time delay matching strength. The short-term capacity forecast results containing the forecast capacity value and the short-term capacity forecast curve are obtained.

[0013] As a preferred embodiment of the short-term capacity forecasting and driving diagnosis method for new energy projects described in this invention, the steps for obtaining the counterfactual capacity diagnosis result are as follows: The short-term capacity forecast curve in the short-term capacity forecast results is determined as the true forecast baseline, and the set of counterfactual diagnostic objects to be restored is determined from the time-lag correction forecast window according to the correspondence between environmental phase, effective response time lag and environmental factors. For any environmental factor in the set of objects to be restored in the counterfactual diagnosis, the stable state corresponding to the previous stable cycle within the same environmental phase is read from the set of reliable samples of production capacity environment. The effective response time delay fingerprint of the environmental factor is called to determine the corresponding effective response time delay and production capacity response characteristics. Other environmental factors, current continuous production capacity time series, project available capacity, equipment operating status and effective response time delay are kept unchanged. Production capacity trend extrapolation and reliable gating of operating boundary are performed again to obtain the counterfactual production capacity diagnosis results.

[0014] As a preferred embodiment of the short-term capacity forecasting and driving diagnosis method for new energy projects described in this invention, the steps for generating short-term capacity driving contribution records are as follows: The difference diagnosis between the counterfactual capacity diagnosis results and the actual prediction baseline is performed according to the same capacity prediction period. The difference in capacity change caused by restoring the steady state of any environmental factor is determined as the independent driving contribution of the corresponding environmental factor. The main driving factor and secondary driving factor are identified according to the magnitude, direction and duration of the independent driving contribution. The environmental factors corresponding to multiple independent driving contributions with the same direction and overlapping duration within the same capacity prediction period are identified as coupled driving factors. Based on the primary driving factor, secondary driving factor, and coupled driving factor, the corresponding capacity forecast period is reviewed without environmental driving factors. When the independent driving contribution of environmental factors does not meet the preset driving contribution threshold and there are abnormal capacity fluctuations in the capacity forecast period that are inconsistent with the capacity response characteristics, the abnormal source is supplemented and confirmed by reading the equipment operation status, power restriction status, maintenance status, and communication abnormal status. The true forecast baseline, counterfactual capacity diagnosis results, independent driving contribution, and abnormal source supplementary confirmation content are associated and encapsulated to generate a short-term capacity driving contribution record.

[0015] As a preferred embodiment of the short-term capacity forecasting and driving diagnosis method for new energy projects described in this invention, the steps for forming a closed-loop file for short-term capacity forecasting and driving diagnosis are as follows: Read the short-term capacity-driven contribution record, and receive the actual capacity feedback data after the corresponding capacity forecast period ends. Match the actual capacity feedback data with the short-term capacity forecast results in the same window according to the same capacity forecast period. Combine the short-term capacity-driven contribution record to confirm the environmental-driven deviation, time-delay drift deviation and non-environmental operation deviation, and obtain the hierarchical write-back object set. Based on the hierarchical write-back object set, the time-delay correction prediction window and environmental factor capacity response time-delay fingerprint of the next prediction period are corrected. The non-environmental isolation segments corresponding to non-environmental operation deviations, capacity and environment window sampling records, capacity and environment reliable sample sets, environmental phase capacity response maps, environmental factor capacity response time-delay fingerprints, short-term capacity prediction results, counterfactual capacity diagnosis results and short-term capacity driving contribution records are associated and encapsulated to form a short-term capacity prediction and driving diagnosis closed-loop archive.

[0016] The beneficial effects of this invention are as follows: by generating environmental factor capacity response time delay fingerprints, a stable expression of the response delay relationship between environmental factor change segments and capacity time-series change segments is realized, enabling short-term capacity prediction windows to be corrected according to the actual response time of environmental factors, and enabling subsequent prediction curves to have environmental phase basis, time delay basis and capacity response basis, thereby providing a traceable data foundation for counterfactual capacity diagnosis and driving contribution identification. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method for short-term capacity forecasting and driver diagnosis of new energy projects.

[0019] Figure 2 A flowchart for generating environmental phase capacity response maps and candidate time delay lists.

[0020] Figure 3 This is a flowchart illustrating the linkage between phase-based reliable solidification and short-term capacity prediction gating.

[0021] Figure 4 A flowchart for contributing closed-loop write-back to counterfactual capacity diagnosis and driving. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4This is one embodiment of the present invention, which provides a method for short-term capacity prediction and driving diagnosis of new energy projects, including the following steps: S1. Based on the time series of new energy project production capacity and the set of multi-source environmental factors, the same time axis is calibrated, and the abnormal sampling is marked with a credible label to obtain the production capacity and environment window sampling record. The production capacity and environment window sampling record is isolated from non-environmental interference and the sampling credibility is verified. Invalid driving segments are screened out to obtain the production capacity and environment credible sample set.

[0026] S1.1: Establish a unified sampling sequence using the grid connection point sampling time in the new energy project capacity time series as the master clock, map the new energy project capacity time series to the unified sampling sequence, and calibrate and write each environmental factor in the multi-source environmental factor set into the corresponding unified sampling time according to the actual sampling time to generate an initial time alignment sequence. It should be noted that the capacity time series of new energy projects refers to the capacity output change data of new energy projects arranged according to the sampling time within a continuous sampling period. The capacity time series of new energy projects includes the active power at the grid connection point, the power generation at the grid connection point, the output power on the equipment side, the grid-connected power, and the corresponding sampling time. The multi-source environmental factor set refers to a collection of multiple environmental factors obtained from on-site measured environmental data, meteorological forecast data, data from nearby meteorological stations, cloud image identification data, and environmental monitoring data. Each environmental factor is an environmental state data corresponding to a certain factor influencing the short-term production capacity change of a new energy project. Environmental factors include irradiance, cloud cover, ambient temperature, component temperature, humidity, wind speed, wind direction, air pressure, air density, rainfall status, snowfall status, dust concentration, and visibility.

[0027] Specifically, the active power, power generation, equipment output power, grid-connected power, and corresponding sampling time of the grid-connected point are read from the capacity time series of new energy projects. The sampling time corresponding to the active power of the grid-connected point is determined as the sampling time of the grid-connected point, and a unified sampling sequence is established according to the chronological order of the sampling times of the grid-connected points. The active power, power generation, equipment output power, and grid-connected power of the grid-connected point in the capacity time series of new energy projects are written into the corresponding unified sampling time in the unified sampling sequence according to the corresponding sampling time. Each environmental factor and its actual sampling time in the multi-source environmental factor set are read. The actual sampling time of each environmental factor is matched with the unified sampling time in the unified sampling sequence. The environmental factor calibration value with matching time position is written into the corresponding unified sampling time, and the actual sampling time corresponding to each environmental factor calibration value is retained in the initial time alignment sequence. When the same environmental factor corresponds to multiple sampling values ​​at the same unified sampling time, the sampling value with the actual sampling time closest to the unified sampling time is selected as the environmental factor calibration value. When the unified sampling time does not correspond to an environmental factor sampling value, an environmental factor vacancy mark is retained at the corresponding unified sampling time, and an initial time alignment sequence is generated.

[0028] S1.2: Calculate the sampling confidence of each unified sampling moment based on the initial time alignment sequence, add an abnormal sampling confidence mark for sampling moments with sampling confidence lower than the preset confidence threshold, and bind the corresponding production capacity time series value, environmental factor calibration value and sampling confidence in the same window to obtain the production capacity environment same window sampling record; Specifically, based on the completed corresponding production capacity time-series value, environmental factor calibration value, and actual sampling time of the environmental factor calibration value at each unified sampling moment in the initial time alignment sequence, the time alignment between the unified sampling moment and the actual sampling time of the environmental factor is verified, as well as the integrity of each environmental factor calibration value at the unified sampling moment and the fluctuation status of the production capacity time-series value and environmental factor calibration value between adjacent unified sampling moments. The sampling reliability of each unified sampling moment is calculated based on the time alignment, integrity, and fluctuation status. The sampling reliability is compared with a preset reliability threshold, and an abnormal sampling reliability mark is added to the unified sampling moment with a sampling reliability lower than the preset reliability threshold. The production capacity time-series value, environmental factor calibration value, sampling reliability, abnormal sampling reliability mark, limited generation status, maintenance status, equipment offline status, and communication abnormal status at the same unified sampling moment are bound together in the same window and arranged in the order of the unified sampling moments to obtain the production capacity and environment window sampling record.

[0029] It should be noted that time alignment refers to the degree of deviation between the unified sampling time and the actual sampling time of the environmental factors, which is used to determine whether the calibration value of the environmental factors can reliably correspond to the production capacity time series value at the same unified sampling time.

[0030] The complete state refers to whether all environmental factors in the multi-source environmental factor set have formed environmental factor calibration values ​​at the same unified sampling time. It is used to determine whether the data at the corresponding unified sampling time has the basis for subsequent peer analysis.

[0031] Fluctuation state refers to the change in the time series value of production capacity and the calibration value of environmental factors between adjacent unified sampling times after normalization processing, and is used to determine whether there is abrupt sampling or abnormal fluctuation at the corresponding unified sampling time.

[0032] The capacity time series value refers to at least one of the following capacity output values ​​of a new energy project: active power at the grid connection point, power generation at the grid connection point, output power on the equipment side, and grid-connected power, which are written into the capacity time series of the project at the corresponding unified sampling time.

[0033] Environmental factor calibration value refers to the environmental state value written into the corresponding unified sampling time after the environmental factors in the multi-source environmental factor set are matched with the actual sampling time.

[0034] The preset confidence threshold is set based on the sampling confidence distribution of historical effective driving segments, the hit rate of abnormal sampling confidence markers, and the allowable deviation of short-term capacity prediction. The example value is 0.80, and the value range is from 0.60 to 0.95. The basis for the value is that sampling points below the preset confidence threshold should not be included in the capacity environment confidence sample set, and sampling points above the preset confidence threshold can maintain the stability of continuous effective driving segments.

[0035] The sampling reliability at each uniform sampling moment is calculated based on the time-fitting state, the complete state, and the fluctuation state, and the expression is: ; in, For the first The sampling reliability at a unified sampling time; To unify the sequence number of sampling times; For the first The maximum time offset between a unified sampling time and the actual sampling time of environmental factors; To allow time offset, the allowed time offset is set based on the sampling period of each environmental factor in the multi-source environmental factor set and the sampling interval of the unified sampling sequence; For the first The number of missing environmental factor calibration values ​​in a uniform sampling time; For the first The number of environmental factor calibration values ​​that should be written in each uniform sampling time; For the first The maximum fluctuation of the time series value of production capacity and the calibration value of environmental factors at a unified sampling time relative to adjacent unified sampling times after normalization. To allow for normalized volatility, the normalized volatility is set based on the volatility distribution of time-series values ​​of production capacity and environmental factor calibration values ​​in historical effective driving segments.

[0036] The value of is greater than 0 and does not exceed 1. The closer the value is to 1, the higher the sampling reliability at the corresponding uniform sampling time. The closer to 0, the more likely it is that an abnormal sampling credibility marker needs to be added at the corresponding uniform sampling time.

[0037] S1.3: Read the limited-generation status, maintenance status, equipment offline status and communication abnormal status corresponding to each unified sampling time from the production capacity environment window sampling record, classify the window sampling pieces with the matching status into non-environment isolation segments, and form the window sampling pieces with the unmatched status into an environment-driven segment set to be verified. Specifically, based on the same-window sampling records of production capacity and environment arranged according to the unified sampling time, each sampling piece is verified for limited power generation status, maintenance status, equipment offline status, and communication anomaly status. When any of these states is present in a sampling piece, it is confirmed that the production capacity change at the corresponding unified sampling time is affected by non-environmental factors, and the corresponding sampling piece is classified into a non-environmental isolation segment. When none of these states are present in a sampling piece, it is confirmed that the production capacity change at the corresponding unified sampling time is not directly interfered with by non-environmental factors, and the corresponding sampling piece is retained as an environment-driven segment to be verified. All environment-driven segments to be verified are arranged according to the order of the unified sampling time to form an environment-driven segment to be verified set.

[0038] It should be noted that the "limited power generation status" refers to the operating status of new energy projects that are affected by dispatch power rationing, grid connection constraints, or power control commands at the corresponding unified sampling time, resulting in the actual production capacity not being output naturally according to environmental conditions.

[0039] Maintenance status refers to the operational status of equipment in new energy projects that is undergoing planned maintenance, temporary maintenance, or maintenance at the corresponding unified sampling time, resulting in the production capacity time series value not reflecting the true driving effect of environmental factors.

[0040] Equipment offline status refers to the operating state in which equipment in a new energy project is not in normal grid-connected power generation or normal data transmission at the corresponding unified sampling time, resulting in missing, stagnant or abnormally declining corresponding production capacity time series values.

[0041] Communication anomalies refer to situations where new energy projects experience collection delays, data interruptions, data duplication, or data loss at the corresponding unified sampling time, resulting in the inability of the time-series production capacity values ​​and environmental factor calibration values ​​to reliably correspond to each other during sampling.

[0042] S1.4: Perform sampling reliability verification and continuity verification on the set of environmental driven segments to be verified. Segments whose sampling reliability meets the preset reliability threshold and whose continuous effective duration meets the preset minimum driving observation duration are confirmed as valid driving segments. The remaining invalid driving segments are screened out to obtain a reliable sample set of production capacity environment.

[0043] Specifically, based on the set of environment-driven verification segments, the sampling confidence level of each environment-driven verification segment is compared with a preset confidence threshold. Environment-driven verification segments whose sampling confidence level does not meet the preset confidence threshold are identified as invalid driving segments. For environment-driven verification segments whose sampling confidence level meets the preset confidence threshold, a continuity verification is performed. The duration for which the sampling confidence level continuously meets the preset confidence threshold is determined according to the chronological order of the unified sampling time, and the continuous effective duration is compared with a preset minimum driving observation duration. When the continuous effective duration meets the preset minimum driving observation duration, the corresponding segment is identified as a valid driving segment. When the continuous effective duration does not meet the preset minimum driving observation duration, the corresponding segment is identified as an invalid driving segment. After filtering out invalid driving segments, a production capacity environment confidence sample set is obtained.

[0044] It should be noted that the preset minimum driving observation duration is based on the sampling period of the new energy project's production capacity time series, the typical response duration of environmental factors to changes in production capacity, and the short-term production capacity prediction window length. The example value range is from 3 to 12 consecutive uniform sampling times. The value is determined to exclude single-point abnormal sampling and instantaneous disturbances, and to retain effective driving segments with continuous environmental driving characteristics.

[0045] S2. Based on the reliable sample set of production capacity environment, identify the environmental phase and bind the production capacity response features to form an environmental phase production capacity response map. Conduct multi-time window response time delay scanning around the environmental phase production capacity response map, compare the environmental factor change segments in the multi-source environmental factor set with the production capacity time series change segments, and generate a candidate time delay list of environmental factors.

[0046] S2.1: Read continuous and effective driving segments from the set of reliable samples of production capacity and environment, and match the direction of production capacity change, production capacity continuity status, production capacity recovery status and environmental factor change status in the continuous and effective driving segments to generate a set of phase candidate segments; Specifically, based on the effective driving segments arranged continuously according to a unified sampling time in the reliable sample set of production capacity environment, the effective driving segments are identified as continuous effective driving segments. Within each continuous effective driving segment, the production capacity time series values ​​at adjacent unified sampling times are compared according to the chronological order of the unified sampling times to determine the production capacity change direction corresponding to production capacity increase, production capacity decrease, and production capacity maintenance. The production capacity continuity state is determined based on the consistency of the production capacity change direction within the continuous unified sampling times, and the production capacity recovery state is determined based on the situation of production capacity recovery after decrease, production capacity decline after increase, and production capacity maintenance followed by change. At the same time, the changes in environmental factor calibration values ​​at adjacent unified sampling times within the same continuous effective driving segment are compared to determine the environmental factor change state of each environmental factor. The production capacity change direction, production capacity continuity state, and production capacity recovery state within the same continuous effective driving segment are correlated with the environmental factor change states at the corresponding unified sampling times to generate a phase candidate segment set.

[0047] S2.2: Based on the phase candidate fragment set, phase candidate fragments that continuously correspond to the changes in production capacity and the changes in environmental factors are grouped into the same environmental phase. The environmental phase is used as the phase node, the environmental factors in the multi-source environmental factor set are used as the factor node, and the production capacity response feature composed of the direction of production capacity change, the continuous state of production capacity, and the state of production capacity recovery is used as the response node. Phase factor association edges are established between phase nodes and factor nodes, factor response association edges are established between factor nodes and response nodes, and phase response association edges are established between phase nodes and response nodes to form an environmental phase production capacity response map. Specifically, based on the phase candidate segment set, the capacity change state, characterized by the capacity change direction, capacity continuity state, and capacity recovery state, is continuously verified to correspond with the environmental factor change state at a unified sampling time. When multiple phase candidate segments show a continuous correspondence of the same capacity change direction, the same capacity continuity state, the same capacity recovery state, and the same environmental factor change state within adjacent unified sampling times, the corresponding phase candidate segments are classified into the same environmental phase. The phase candidate segments classified into the same environmental phase form an environmental phase, and the environmental phase is registered as a phase node. The environmental factors in the multi-source environmental factor set are registered as factor nodes, and the capacity response features composed of the capacity change direction, capacity continuity state, and capacity recovery state are registered as response nodes. According to the source of the environmental factor change state in the same environmental phase, a phase factor association edge is established between the phase node and the corresponding factor node. According to the correspondence between the environmental factor change state and the capacity response features in the same phase candidate segment, a factor response association edge is established between the factor node and the response node. According to the attribution relationship of the capacity response features in the same environmental phase, a phase response association edge is established between the phase node and the response node, forming an environmental phase capacity response map.

[0048] S2.3: Based on the environmental phase production capacity response map, preset short windows, medium windows and long windows with increasing time lengths are used to extract environmental factor change segments under the same environmental phase as time-delay scanning objects; Specifically, based on the environmental phase capacity response map, the environmental phase corresponding to the same phase node is determined as the current processing environmental phase, and the environmental factors associated with the current processing environmental phase are determined according to the phase factor association edge. Around the continuous effective driving segment and unified sampling time corresponding to the current processing environmental phase, the continuous change of the environmental factor calibration value of the environmental factor associated with the current processing environmental phase is sequentially extracted within the same environmental phase range according to the increasing time length of the preset short window, medium window and long window, forming environmental factor change segments corresponding to different time lengths. The same environmental phase, corresponding environmental factor, start and end unified sampling time, preset short window, medium window and long window and environmental factor change segment are associated and retained as time-delay scanning objects.

[0049] It should be noted that the preset short window, medium window, and long window are set based on the sampling period of the new energy project's production capacity time series, the duration of typical changes in environmental factors, and the length of the short-term production capacity prediction window. The example value range is 3 to 6 consecutive unified sampling times for the preset short window, 7 to 12 consecutive unified sampling times for the preset medium window, and 13 to 24 consecutive unified sampling times for the preset long window. The value is determined by the fact that the preset short window is used to capture rapid environmental changes, the preset medium window is used to cover the normal environmental driving process, and the preset long window is used to retain the lag effect of slow environmental changes on the production capacity time series changes.

[0050] S2.4: For time-delay scanning objects, the environmental factor change segments are gradually shifted relative to the production capacity time-series change segments along the forward direction of the time axis, and the correspondence formed by each shift is written into the time-delay correspondence set to be evaluated; Specifically, for time-delay scanning objects, within the same environmental phase and the same intercept range, the production capacity time series values ​​are formed into production capacity time series change segments according to the chronological order of the unified sampling times. Using the interval between adjacent unified sampling times in the unified sampling sequence as the translation interval, the environmental factor change segments are moved sequentially along the forward direction of the time axis relative to the production capacity time series change segments, so that the environmental factor change segments are earlier than the production capacity time series change segments in time. After each movement, the environmental factor change segments and the production capacity time series change segments are re-corresponded according to the unified sampling times to form a set of time delay correspondences to be evaluated.

[0051] S2.5: Perform phase-locked time delay matching on the set of time delay correspondences to be evaluated, generate phase-locked time delay matching strength, associate and encapsulate the correspondences that meet the preset candidate time delay conditions with environmental phase, environmental factors, candidate time delays, phase-locked time delay matching strength and production capacity response characteristics, and generate a list of environmental factor candidate time delays.

[0052] Specifically, based on the correspondences in the set of time-delay correspondences to be evaluated and the capacity response characteristics corresponding to the same environmental phase and the same environmental factor in the environmental phase capacity response map, the environmental phase is used as the locking range to limit the phase factor association edges, factor response association edges, and phase response association edges participating in time-delay matching. Time-delay matching is performed on each correspondence within the locked environmental phase range. The moved environmental factor change segments are compared with the corresponding capacity time-series change segments according to the correspondence of unified sampling time. When the capacity change direction of the capacity time-series change segment is consistent with the capacity change direction in the capacity response characteristics, it is recorded as capacity change direction matching. When the change process of the capacity time-series change segment within the continuous unified sampling time is consistent with the capacity continuity state in the capacity response characteristics, it is recorded as capacity continuity state matching. When the capacity time-series change segment experiences a rebound, fall, or re-emergence after the change, it is recorded as capacity continuity state matching. When a changing process aligns with the capacity recovery state in the capacity response characteristics, it is recorded as a capacity recovery state match. When the correspondence simultaneously falls within the range of environmental phase, environmental factor, and capacity response characteristics defined by the locked phase factor association edge, factor response association edge, and phase response association edge, it is recorded as a locked association edge hit. Based on the success of locked association edge hits, capacity change direction matching, capacity continuity state matching, and capacity recovery state matching, a phase lock time delay matching strength is generated according to preset assignment rules. Correspondences whose phase lock time delay matching strength satisfies preset candidate time delay conditions are confirmed as candidate time delay correspondences. The preceding interval in the candidate time delay correspondences is determined as candidate time delays. The candidate time delay correspondences, environmental phase, environmental factor, candidate time delay, phase lock time delay matching strength, and capacity response characteristics are associated and encapsulated to generate an environmental factor candidate time delay list.

[0053] It should be noted that the preset assignment rules are as follows: when the locked associated edge is not hit, the phase lock delay matching strength is assigned to 0%; when the locked associated edge is hit and the matching of capacity change direction, capacity continuity state, and capacity recovery state are all true, the phase lock delay matching strength is assigned to 100%; when the locked associated edge is hit and any two of the matching of capacity change direction, capacity continuity state, and capacity recovery state are true, the phase lock delay matching strength is assigned to 75%; when the locked associated edge is hit and any one of the matching of capacity change direction, capacity continuity state, and capacity recovery state is true, the phase lock delay matching strength is assigned to 50%; when the locked associated edge is hit and none of the matching of capacity change direction, capacity continuity state, and capacity recovery state are true, the phase lock delay matching strength is assigned to 0%.

[0054] The preset candidate time delay conditions are based on the distribution of the preceding interval of environmental factor change segments relative to the time series change segments of production capacity in the historical effective driving segments, the distribution of phase-locked time delay matching strength, and the allowable deviation of short-term production capacity prediction. The example value is that the candidate time delay does not exceed the preset long window and the phase-locked time delay matching strength is not less than 75%. The basis for the value is that the candidate time delay can cover the common lag response time of environmental factors to production capacity changes, while eliminating the correspondence that only forms a single match and has unstable phase correlation.

[0055] S3. Perform intra-phase reliable solidification on the candidate time delay list of environmental factors, eliminate invalid time differences caused by occasional disturbances, generate environmental factor capacity response time delay fingerprints, correct the short-term capacity prediction window based on the environmental factor capacity response time delay fingerprints and perform reliable gating of the operating boundary to obtain the short-term capacity prediction results.

[0056] S3.1: Read the candidate options corresponding to the same environmental phase, the same environmental factor and the same candidate time delay from the environmental factor candidate time delay list, and retrieve the sampling credibility corresponding to the candidate options from the production capacity environment credibility sample set. Write the production capacity response characteristics, phase-locked time delay matching strength and sampling credibility associated with the candidate options into the phase-in-phase candidate time delay group. Specifically, based on the environmental phase, environmental factor, candidate time delay, phase-locked time delay matching strength, production capacity response characteristics, and candidate time delay correspondence already associated and encapsulated in the environmental factor candidate time delay list, candidate time delay correspondences with the same environmental phase, the same environmental factor, and the same candidate time delay are identified as the same group of candidates. According to the unified sampling time correspondence retained in each candidate, the sampling credibility under the unified sampling time is matched in the production capacity environment credibility sample set, and the sampling credibility is correlated with the production capacity response characteristics and phase-locked time delay matching strength in the candidate. All candidates under the same environmental phase, the same environmental factor, and the same candidate time delay, as well as the production capacity response characteristics, phase-locked time delay matching strength, and sampling credibility associated with the candidate, are merged in the order of unified sampling time to form a candidate time delay group within the phase.

[0057] S3.2: Perform repeated hit verification and response consistency verification on the candidate time delay group within the phase. The candidate time delay that repeatedly hits in multiple consecutive effective driving segments in the same environmental phase and whose production capacity response characteristics remain consistent is confirmed as an effective response time delay. The candidate time delay that cannot be repeatedly hit is eliminated as an invalid time difference caused by occasional disturbance. The effective response time delay is associated and encapsulated with the corresponding environmental phase, environmental factor, production capacity response characteristics and phase locking time delay matching strength to generate an environmental factor production capacity response time delay fingerprint. Specifically, the candidate options corresponding to the same environmental phase, environmental factor, and candidate time delay within the candidate time delay group within the phase are used as verification objects. Repeated hit verification and response consistency verification are performed. Repeated hit verification confirms whether the same candidate time delay forms candidates in different consecutive effective driving segments under the same environmental phase, and combines the sampling reliability at the unified sampling time corresponding to the candidate options to confirm whether the source of repeated hits comes from a reliable sample set of production capacity environment. Response consistency verification confirms whether the production capacity response characteristics associated with repeatedly hit candidates maintain the same direction of production capacity change, production capacity continuity, and production capacity recovery. When the same candidate time delay repeatedly hits in different consecutive effective driving segments under the same environmental phase, and the production capacity response characteristics associated with the repeatedly hit candidates remain consistent, the corresponding candidate time delay is confirmed as a valid response time delay. When the candidate time delay only hits in a single consecutive effective driving segment, or the production capacity response characteristics associated with repeatedly hit candidates are inconsistent, the corresponding candidate time delay is removed as an invalid time difference caused by occasional disturbances, generating an environmental factor production capacity response time delay fingerprint.

[0058] S3.3: Call the environmental factor capacity response time lag fingerprint, map the environmental factors used for future short-term prediction in the multi-source environmental factor set to the capacity prediction period according to the corresponding effective response time lag, and write the current continuous capacity time series, project available capacity and equipment operating status into the same prediction window to generate a time lag correction prediction window; Specifically, based on the environmental phase, environmental factors, effective response time delay, capacity response characteristics, and phase-locked time delay matching strength already associated and encapsulated in the environmental factor capacity response time delay fingerprint, environmental factors in the multi-source environmental factor set that are used for future short-term prediction and have formed effective response time delays are identified. Taking the capacity prediction period as the time correspondence target, the environmental factor calibration values ​​of the corresponding environmental factors are shifted backward on the time axis according to the effective response time delay, so that the time when the environmental factor changes occurs is mapped to the capacity prediction period after the effective response time delay. The continuous effective capacity time series values ​​in the capacity time series of new energy projects up to the current prediction time are organized into the current continuous capacity time series, and the available capacity of the project and the equipment operating status are aligned and written according to the same capacity prediction period to generate a time delay correction prediction window.

[0059] It should be noted that the available capacity of a project refers to the capacity boundary that a new energy project can participate in capacity output after deducting the impact of power rationing, maintenance, equipment offline, and capacity constraints within the corresponding capacity forecast period.

[0060] Equipment operating status refers to the grid connection, shutdown, maintenance, fault, offline and restricted operation of inverters, wind turbines, combiner boxes, step-up equipment and grid-connected equipment in new energy projects during the corresponding capacity forecast period.

[0061] S3.4: Based on the time-delay correction prediction window and the capacity response characteristics in the capacity response time-delay fingerprint of environmental factors, extract the capacity time series value change of the same environmental phase, the same environmental factor, the same effective response time delay and the same capacity response characteristics from the capacity environment reliable sample set, determine the prediction increment of each capacity prediction period, and form the prediction capacity value from the end capacity time series value of the current continuous capacity time series for each period, and generate candidate short-term capacity prediction curves. Specifically, based on environmental factors and effective response time delays arranged by capacity prediction period in the time-delay correction prediction window, and combined with the capacity response characteristics in the capacity response time delay fingerprint of environmental factors, continuous effective driving segments with the same environmental phase, the same environmental factor, the same effective response time delay, and consistent capacity response characteristics are searched in the reliable sample set of capacity environment. The changes in capacity time series values ​​at adjacent unified sampling times are statistically analyzed from the found continuous effective driving segments, and the average change in capacity time series values ​​is determined as the prediction increment benchmark for the corresponding environmental factor under the corresponding capacity response characteristics. When the capacity change direction in the capacity response characteristics is an increase in capacity, the prediction increment benchmark is used as the positive prediction increment; when the capacity change direction in the capacity response characteristics is a decrease in capacity, the prediction increment benchmark is used as the negative prediction increment. The quantity; when the capacity recovery status in the capacity response characteristics indicates one of the processes of capacity decline followed by recovery, capacity rise followed by decline, and capacity stabilization followed by change, the recovery prediction increment is determined according to the recovery direction corresponding to the capacity recovery status; taking the end capacity time series value of the current continuous capacity time series as the starting prediction capacity value, the prediction increment corresponding to each capacity prediction period is updated segment by segment according to the order of the capacity prediction period to form the prediction capacity value; when multiple environmental factors are mapped to the same capacity prediction period, the prediction increments with the same direction are merged, and the prediction increments with opposite directions are canceled out to form the synthetic prediction increment of the corresponding capacity prediction period, and the prediction capacity value of the corresponding capacity prediction period is formed based on the synthetic prediction increment; the prediction capacity values ​​are connected according to the order of the capacity prediction period to generate candidate short-term capacity prediction curves.

[0062] S3.5: Perform operational boundary confidence gating on the candidate short-term capacity prediction curves, implement upper limit blocking for capacity prediction segments that exceed the available capacity of the project, implement smoothing constraints for capacity prediction segments that do not meet the equipment ramp-up boundary, and add low confidence markers to capacity prediction segments with insufficient phase-locking delay matching strength, to obtain short-term capacity prediction results that include predicted capacity values ​​and short-term capacity prediction curves.

[0063] Specifically, based on the sequence of predicted capacity changes arranged by prediction period in the candidate short-term capacity prediction curve, each capacity prediction segment is sequentially subjected to operational boundary confidence gating, consisting of project available capacity gating, equipment ramp-up boundary gating, and phase-locked time delay matching strength gating. In the project available capacity gating, the predicted capacity value is compared with the project available capacity within the same capacity prediction period. When the predicted capacity value exceeds the project available capacity, the predicted capacity value of the corresponding capacity prediction segment is limited to within the project available capacity. In the equipment ramp-up boundary gating, the permitted equipment within the corresponding capacity prediction period is determined based on the equipment operating status. The ramp-up boundary compares the predicted capacity change process between adjacent capacity prediction segments with the equipment ramp-up boundary. When the predicted capacity change process does not meet the equipment ramp-up boundary, the change process between adjacent capacity prediction segments is continuously smoothed according to the equipment ramp-up boundary. In the phase-locked delay matching strength gating, the phase-locked delay matching strength associated with each capacity prediction segment is compared with the matching strength requirement in the preset candidate delay conditions. When the phase-locked delay matching strength does not meet the matching strength requirement, a low confidence mark is added to the corresponding capacity prediction segment to obtain the short-term capacity prediction result containing the short-term capacity prediction curve.

[0064] S4. Construct a counterfactual capacity diagnosis scenario based on the short-term capacity forecast results. Separate the independent effects by restoring the steady state of any environmental factor to obtain the counterfactual capacity diagnosis results. Perform a difference diagnosis between the counterfactual capacity diagnosis results and the short-term capacity forecast results to generate a short-term capacity driving contribution record.

[0065] S4.1: Determine the short-term capacity forecast curve in the short-term capacity forecast results as the true forecast baseline, and determine the set of counterfactual diagnostic objects to be restored from the time-delay correction forecast window according to the correspondence between environmental phase, effective response time delay and environmental factors; Specifically, based on the short-term capacity forecast results, the predicted capacity values, capacity forecast segments, and corresponding capacity change trends arranged by capacity forecast period are retained as the true forecast baseline. Around the environmental phases, environmental factors, effective response delays, capacity response characteristics, and phase-locked delay matching strength already written for the same capacity forecast period in the time-lag correction forecast window, the correspondence between environmental phases, effective response delays, and environmental factors is checked one by one. When an environmental factor has been mapped to the corresponding capacity forecast period according to the effective response delay, and the capacity response characteristics corresponding to the environmental phase are consistent with the capacity change trend of the corresponding capacity forecast segment in the true forecast baseline, the corresponding environmental factor is identified as a counterfactual diagnostic object to be restored, forming a counterfactual diagnostic object set to be restored.

[0066] S4.2: For any environmental factor in the set of objects to be restored in the counterfactual diagnosis, read the stable state corresponding to the previous stable cycle within the same environmental phase from the set of reliable samples of production capacity environment, and call the environmental factor production capacity response time delay fingerprint to determine the corresponding effective response time delay and production capacity response characteristics. Keep other environmental factors, current continuous production capacity time series, project available capacity, equipment operating status and effective response time delay unchanged, and re-perform production capacity trend extrapolation and reliable gating of operating boundary to obtain the counterfactual production capacity diagnosis results. Specifically, for any environmental factor in the set of objects to be restored in the counterfactual diagnosis, the system searches the production capacity environment reliable sample set for periods earlier than the current prediction time within the same environmental phase, where the environmental factor change state remains unchanged within consecutive uniform sampling times, and the sampling reliability meets a preset reliability threshold. The period that meets the aforementioned conditions and is closest to the current prediction time is identified as the previous stable period, and the environmental factor calibration value of the corresponding environmental factor in the previous stable period is determined as the stable state. Based on the environmental factor production capacity response time delay fingerprint, the system confirms the effective response time delay and production capacity response characteristics under the corresponding environmental phase and environmental factor, and identifies the corresponding period in the time delay correction prediction window. The environmental factor calibration values ​​of environmental factors are replaced with stable states, while keeping the environmental factor calibration values ​​of other environmental factors, the current continuous capacity time series, the project's available capacity, the equipment operating status, and the effective response time delay unchanged. Within the time delay correction prediction window after the replacement, the capacity trend is re-performed, and the operation boundary credibility gating consisting of project available capacity gating, equipment ramp-up boundary gating, and phase-locked time delay matching strength gating is applied to the newly formed capacity prediction segment. The resulting counterfactual capacity prediction curve is associated with and retained with the corresponding environmental factors, environmental phase, effective response time delay, capacity response characteristics, and capacity prediction period to obtain the counterfactual capacity diagnosis result.

[0067] S4.3: The difference diagnosis between the counterfactual capacity diagnosis results and the actual prediction baseline is performed according to the same capacity prediction period. The difference in capacity change caused by restoring the stable state of any environmental factor is determined as the independent driving contribution of the corresponding environmental factor. The main driving factor and secondary driving factor are confirmed according to the magnitude, direction and duration of the independent driving contribution. The environmental factors corresponding to multiple independent driving contributions with the same direction and overlapping duration within the same capacity prediction period are confirmed as coupled driving factors. Specifically, based on the counterfactual capacity prediction curve formed after any environmental factor in the counterfactual capacity diagnosis results returns to a stable state, and the predicted capacity value corresponding to the same capacity prediction period in the true prediction baseline, difference diagnosis is performed segment by segment according to the capacity prediction period. The difference between the predicted capacity value of the true prediction baseline and the predicted capacity value in the counterfactual capacity diagnosis results is determined as the independent driving contribution of the corresponding environmental factor. The magnitude of the difference is determined as the size of the independent driving contribution. The direction of capacity increase or decrease corresponding to the difference is determined as the direction of the independent driving contribution. The time range in which the difference exists within the continuous capacity prediction period is determined as the persistence state of the independent driving contribution. Within the same capacity prediction period, each environmental factor is ranked according to the size of its independent driving contribution. The environmental factor with the largest independent driving contribution and whose persistence state covers the corresponding capacity prediction period is identified as the primary driving factor. Environmental factors that still form independent driving contributions other than the primary driving factor are identified as secondary driving factors. When the independent driving contribution directions of multiple environmental factors within the same capacity prediction period are consistent and their persistence states overlap, the environmental factors corresponding to the multiple independent driving contributions are identified as coupled driving factors.

[0068] S4.4: Based on the main driving factor, secondary driving factor and coupled driving factor, perform non-environmental driving verification on the corresponding capacity forecast period. When the independent driving contribution of the environmental factor does not meet the preset driving contribution threshold and there is abnormal capacity fluctuation in the capacity forecast period that is inconsistent with the capacity response characteristics, read the equipment operation status, power restriction status, maintenance status and communication abnormal status to supplement and confirm the source of the abnormality. Then, associate and encapsulate the true forecast baseline, counterfactual capacity diagnosis results, independent driving contribution and supplementary confirmation of the source of the abnormality to generate a short-term capacity driving contribution record.

[0069] Specifically, based on the capacity prediction period corresponding to the primary driving factor, secondary driving factor, and coupled driving factor, the magnitude, direction, duration, and capacity response characteristics of the independent driving contribution, a non-environmental driving verification is performed on the corresponding capacity prediction period. When the independent driving contribution of the environmental factor meets the preset driving contribution threshold, and the direction of capacity change, duration, and recovery status within the capacity prediction period are consistent with the capacity response characteristics, the corresponding capacity prediction period is confirmed as having an environmental driving interpretation. When the independent driving contribution of the environmental factor does not meet the preset driving contribution threshold, and there are capacity anomalies in the capacity prediction period that are inconsistent with the capacity response characteristics, the corresponding capacity prediction period is confirmed as having an environmental driving interpretation. During periods of fluctuation, the system retrieves the equipment operating status, power generation restriction status, maintenance status, and communication anomaly status for the corresponding capacity forecast period, and confirms the source of abnormal capacity fluctuations based on the status match. When the equipment operating status exhibits at least one of the following conditions: fault, offline, shutdown, or restricted operation, the equipment operating status is added as supplementary confirmation of the anomaly source. When at least one of the following conditions is matched: power generation restriction status, maintenance status, or communication anomaly status, the corresponding matched status is added as supplementary confirmation of the anomaly source. When none of the following conditions are matched, the unmatched status is added as supplementary confirmation of the anomaly source, and a short-term capacity-driven contribution record is generated.

[0070] It should be noted that the preset driving contribution threshold is set based on the distribution of independent driving contributions in historical effective driving segments, the allowable deviation of short-term capacity forecast, and the available capacity of the project. An exemplary value range is 2% to 8% of the available capacity of the project. The basis for this value is that independent driving contributions below the preset driving contribution threshold are insufficient to stably explain abnormal capacity fluctuations during the capacity forecast period, while independent driving contributions above the preset driving contribution threshold can support the distinction between primary driving factors, secondary driving factors, and coupled driving factors.

[0071] S5. Based on the short-term capacity-driven contribution record, a closed-loop archive and write-back correction are performed, and the prediction deviation is fed back to the next prediction cycle to form a closed-loop archive of short-term capacity prediction and driver diagnosis.

[0072] S5.1: Read the short-term capacity-driven contribution record, and receive the actual capacity feedback data after the corresponding capacity forecast period ends. Match the actual capacity feedback data with the short-term capacity forecast results in the same window according to the same capacity forecast period. Combine the short-term capacity-driven contribution record to confirm the environmental-driven deviation, time-delay drift deviation and non-environmental operation deviation, and obtain the hierarchical write-back object set. Specifically, based on short-term capacity-driven contribution records, actual capacity feedback data is received after the corresponding capacity forecast period ends. This data is then correlated with the short-term capacity forecast curve in the short-term capacity forecast results to confirm the predicted-actual difference between the actual and predicted capacity values ​​within the same forecast period. When the direction of the predicted-actual difference aligns with the independent driving contribution directions corresponding to the primary, secondary, and coupled driving factors, and the difference between the magnitude of the predicted-actual difference and the total magnitude of the independent driving contributions in the same direction within the same capacity forecast period reaches a preset environmental driving deviation threshold, then... Furthermore, if the supplementary confirmation of the anomaly source indicates that the equipment operating status, limited power generation status, maintenance status, and communication anomaly status are all not matched, the corresponding capacity prediction period is identified as an environmentally driven deviation. If the actual prediction difference is earlier or later than the effective response time lag, and the environmental factor change status and capacity change direction still maintain a corresponding relationship, the corresponding capacity prediction period is identified as a time lag drift deviation. If the equipment operating status, limited power generation status, maintenance status, and communication anomaly status are matched in the supplementary confirmation of the anomaly source, the corresponding capacity prediction period is identified as a non-environmental operating deviation, thus obtaining a hierarchical write-back object set.

[0073] It should be noted that the preset environmental-driven deviation threshold is set based on the allowable deviation of short-term capacity forecast, the distribution of actual differences in historical forecasts, and the available capacity of the project. An exemplary value range is 1% to 5% of the available capacity of the project. The basis for the value is that when the difference between the magnitude of the actual difference in forecast and the total size of the independent driving contributions in the same direction within the same capacity forecast period is lower than the preset environmental-driven deviation threshold, it is considered an acceptable forecast fluctuation. When the preset environmental-driven deviation threshold is reached, it indicates that the existing independent driving contributions have not fully covered the actual capacity deviation.

[0074] S5.2: Based on the hierarchical write-back object set, the time-delay correction prediction window and environmental factor capacity response time-delay fingerprint of the next prediction period are corrected. The non-environmental isolation segment corresponding to non-environmental operation deviation, capacity and environment window sampling record, capacity and environment reliable sample set, environmental phase capacity response map, environmental factor capacity response time-delay fingerprint, short-term capacity prediction results, counterfactual capacity diagnosis results and short-term capacity driving contribution record are associated and encapsulated to form a short-term capacity prediction and driving diagnosis closed-loop archive.

[0075] Specifically, based on the hierarchical write-back object set, the time lag correction prediction window and environmental factor capacity response time lag fingerprint for the next prediction period are hierarchically corrected. When the hierarchical write-back object set records environmental driving deviations, the corresponding environmental phase, environmental factor, capacity prediction period, and actual capacity feedback data are written into the time lag correction prediction window for the next prediction period. When the hierarchical write-back object set records time lag drift deviations, the time interval between the unified sampling time of the environmental factor change state and the unified sampling time of the same-direction capacity change in the actual capacity feedback data is used as the actual response time lag. Depending on whether the actual response time lag is shorter or longer than the effective response time lag, the effective response time lag in the environmental factor capacity response time lag fingerprint is adjusted downwards or upwards, respectively. When time delay drift deviations occur multiple times under the same environmental phase and the same environmental factor, the median value of the actual response time delay with sampling confidence meeting a preset confidence threshold is used to update the effective response time delay. When the hierarchical write-back object set records non-environmental operation deviations, the corresponding window sampling slices within the corresponding capacity prediction period are classified into non-environmental isolation segments, and the correction effect of the corresponding capacity prediction period on the environmental factor capacity response time delay fingerprint is excluded. The non-environmental isolation segments, capacity and environment window sampling records, capacity and environment confidence sample sets, environmental phase capacity response map, corrected environmental factor capacity response time delay fingerprint, short-term capacity prediction results, counterfactual capacity diagnosis results, and short-term capacity driving contribution records are associated and encapsulated to form a short-term capacity prediction and driving diagnosis closed-loop archive.

[0076] In summary, this invention achieves a stable expression of the response delay relationship between environmental factor change segments and production capacity time-series change segments by generating environmental factor production capacity response time-delay fingerprints. This enables short-term production capacity prediction windows to be corrected according to the actual response time of environmental factors, and provides subsequent prediction curves with environmental phase basis, time-delay basis, and production capacity response basis, thereby providing a traceable data foundation for counterfactual production capacity diagnosis and driving contribution identification.

[0077] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for short-term capacity forecasting and driver diagnosis of new energy projects, characterized in that, include: Based on the time series of new energy project capacity and the set of multi-source environmental factors, the same time axis is calibrated, and the abnormal sampling is marked with a credible label to obtain the capacity and environment window sampling record. The capacity and environment window sampling record is isolated from non-environmental interference and the sampling credibility is verified. Invalid driving segments are screened out to obtain a credible sample set of capacity and environment. Based on a reliable sample set of production capacity environment, environmental phases are identified and bound to production capacity response features to form an environmental phase production capacity response map. Multi-time window response time delay scanning is carried out around the environmental phase production capacity response map. Before and after comparison of environmental factor change segments and production capacity time series change segments in the multi-source environmental factor set is performed to generate a candidate time delay list of environmental factors. The candidate time delay list of environmental factors is reliably solidified within the phase, and invalid time differences caused by occasional disturbances are eliminated. The environmental factor capacity response time delay fingerprint is generated. Based on the environmental factor capacity response time delay fingerprint, the short-term capacity prediction window is corrected and the operating boundary is reliably gated to obtain the short-term capacity prediction result. Based on the short-term capacity forecast results, a counterfactual capacity diagnosis scenario is constructed. By restoring the steady state of any environmental factor to separate the independent effects, the counterfactual capacity diagnosis results are obtained. The difference between the counterfactual capacity diagnosis results and the short-term capacity forecast results is diagnosed, and a short-term capacity driving contribution record is generated. Based on short-term capacity-driven contribution records, a closed-loop archiving and write-back correction is performed, and the forecast deviation is fed back to the next forecast period, forming a closed-loop archive of short-term capacity forecasting and driver diagnosis.

2. The method for short-term capacity forecasting and driver diagnosis of new energy projects as described in claim 1, characterized in that, The steps to obtain the production capacity environment peer sampling record are as follows: A unified sampling sequence is established using the grid connection point sampling time in the new energy project capacity time series as the master clock. The new energy project capacity time series is mapped to the unified sampling sequence. Each environmental factor in the multi-source environmental factor set is calibrated and written into the corresponding unified sampling time according to the actual sampling time to generate an initial time alignment sequence. The sampling confidence level of each unified sampling moment is calculated based on the initial time alignment sequence. An abnormal sampling confidence mark is added to the sampling moment when the sampling confidence level is lower than the preset confidence threshold. The corresponding production capacity time series value, environmental factor calibration value and sampling confidence level are bound together to obtain the production capacity and environment window sampling record.

3. The method for short-term capacity forecasting and driver diagnosis of new energy projects as described in claim 2, characterized in that, The steps to obtain the reliable sample set of production capacity environment are as follows: Read the limited-issue status, maintenance status, equipment offline status and communication abnormal status corresponding to each unified sampling time from the same sampling record of production capacity and environment. Classify the same sampling pieces with the matching status into non-environmental isolation segments, and form the same sampling pieces with the unmatched status into an environment-driven set of segments to be verified. The set of environmental-driven segments to be verified is subjected to sampling credibility verification and continuity verification. Segments whose sampling credibility meets the preset credibility threshold and whose continuous effective duration meets the preset minimum driving observation duration are identified as valid driving segments. The remaining invalid driving segments are screened out to obtain a reliable sample set of production capacity environment.

4. The method for short-term capacity forecasting and driver diagnosis of new energy projects as described in claim 1, characterized in that, The steps for forming the environmental phase energy production response map are as follows: Read continuous and effective driving segments from the set of reliable samples of production capacity and environment, and match the direction of production capacity change, the continuous state of production capacity, the state of production capacity recovery and the state of environmental factor change in the continuous and effective driving segments to generate a set of phase candidate segments. Based on the phase candidate fragment set, phase candidate fragments that continuously correspond to the changes in production capacity and the changes in environmental factors are grouped into the same environmental phase. The environmental phase is used as the phase node, the environmental factors in the multi-source environmental factor set are used as the factor node, and the production capacity response feature composed of the direction of production capacity change, the continuous state of production capacity, and the state of production capacity recovery is used as the response node. Phase factor correlation edges are established between phase nodes and factor nodes, factor response correlation edges are established between factor nodes and response nodes, and phase response correlation edges are established between phase nodes and response nodes to form an environmental phase production capacity response map.

5. The method for short-term capacity forecasting and drive diagnosis of new energy projects as described in claim 4, characterized in that, The steps for generating the candidate time-delay list of environmental factors are as follows: Based on the environmental phase energy production response map, environmental factor change segments under the same environmental phase are extracted as time-delay scanning objects by preset short windows, medium windows and long windows with increasing time lengths. For time-lag scanning objects, the environmental factor change segments are gradually shifted relative to the production capacity time-series change segments along the time axis forward direction, and the correspondence formed by each shift is written into the time-lag correspondence set to be evaluated. Phase-locked time delay matching is performed on the set of time delay correspondences to be evaluated to generate phase-locked time delay matching strength. The correspondences that meet the preset candidate time delay conditions are associated and encapsulated with environmental phase, environmental factors, candidate time delays, phase-locked time delay matching strength and production capacity response characteristics to generate a list of environmental factor candidate time delays.

6. The method for short-term capacity forecasting and driver diagnosis of new energy projects as described in claim 5, characterized in that, The steps for generating the environmental factor productivity response time-delay fingerprint are as follows: Read the candidate options corresponding to the same environmental phase, the same environmental factor, and the same candidate time delay from the list of candidate time delays for environmental factors, and retrieve the sampling credibility of the candidate options from the set of reliable samples for production capacity environment. Write the production capacity response characteristics, phase-locked time delay matching strength, and sampling credibility associated with the candidate options into the candidate time delay group within the phase. Repeated hit verification and response consistency verification are performed on candidate time delay groups within the phase. Candidate time delays that repeatedly hit in multiple consecutive effective drive segments in the same environmental phase and whose production capacity response characteristics remain consistent are confirmed as effective response time delays. Candidate time delays that cannot be repeatedly hit are eliminated as invalid time differences caused by occasional disturbances. The effective response time delays are associated and encapsulated with the corresponding environmental phase, environmental factor, production capacity response characteristics and phase locking time delay matching strength to generate environmental factor production capacity response time delay fingerprints.

7. The method for short-term capacity forecasting and drive diagnosis of new energy projects as described in claim 6, characterized in that, The steps to obtain the short-term capacity forecast results are as follows: Call the environmental factor capacity response time lag fingerprint, map the environmental factors in the multi-source environmental factor set used for future short-term prediction to the capacity prediction period according to the corresponding effective response time lag, and write the current continuous capacity time series, project available capacity and equipment operating status into the same prediction window to generate a time lag correction prediction window; Based on the time-delay correction prediction window and the capacity response characteristics in the capacity response time-delay fingerprint of environmental factors, the change in capacity time series values ​​with the same environmental phase, the same environmental factor, the same effective response time delay and consistent capacity response characteristics are extracted from the reliable sample set of capacity environment. The predicted increment for each capacity prediction period is determined, and the predicted capacity value is formed by the end capacity time series value of the current continuous capacity time series for each period, generating candidate short-term capacity prediction curves. The candidate short-term capacity forecast curves are subjected to operational boundary confidence gating. Capacity forecast segments that exceed the available capacity of the project are locked at the upper limit. Capacity forecast segments that do not meet the equipment ramp-up boundary are subject to smoothing constraints. Low confidence markers are added to capacity forecast segments with insufficient phase-locked time delay matching strength. The short-term capacity forecast results containing the forecast capacity value and the short-term capacity forecast curve are obtained.

8. The method for short-term capacity forecasting and drive diagnosis of new energy projects as described in claim 7, characterized in that, The steps to obtain the counterfactual capacity diagnosis results are as follows: The short-term capacity forecast curve in the short-term capacity forecast results is determined as the true forecast baseline, and the set of counterfactual diagnostic objects to be restored is determined from the time-lag correction forecast window according to the correspondence between environmental phase, effective response time lag and environmental factors. For any environmental factor in the set of objects to be restored in the counterfactual diagnosis, the stable state corresponding to the previous stable cycle within the same environmental phase is read from the set of reliable samples of production capacity environment. The effective response time delay fingerprint of the environmental factor is called to determine the corresponding effective response time delay and production capacity response characteristics. Other environmental factors, current continuous production capacity time series, project available capacity, equipment operating status and effective response time delay are kept unchanged. Production capacity trend extrapolation and reliable gating of operating boundary are performed again to obtain the counterfactual production capacity diagnosis results.

9. The method for short-term capacity forecasting and drive diagnosis of new energy projects as described in claim 8, characterized in that, The steps for generating short-term capacity-driven contribution records are as follows: The difference diagnosis between the counterfactual capacity diagnosis results and the actual prediction baseline is performed according to the same capacity prediction period. The difference in capacity change caused by restoring the steady state of any environmental factor is determined as the independent driving contribution of the corresponding environmental factor. The main driving factor and secondary driving factor are identified according to the magnitude, direction and duration of the independent driving contribution. The environmental factors corresponding to multiple independent driving contributions with the same direction and overlapping duration within the same capacity prediction period are identified as coupled driving factors. Based on the primary driving factor, secondary driving factor, and coupled driving factor, the corresponding capacity forecast period is reviewed without environmental driving factors. When the independent driving contribution of environmental factors does not meet the preset driving contribution threshold and there are abnormal capacity fluctuations in the capacity forecast period that are inconsistent with the capacity response characteristics, the abnormal source is supplemented and confirmed by reading the equipment operation status, power restriction status, maintenance status, and communication abnormal status. The true forecast baseline, counterfactual capacity diagnosis results, independent driving contribution, and abnormal source supplementary confirmation content are associated and encapsulated to generate a short-term capacity driving contribution record.

10. The method for short-term capacity forecasting and drive diagnosis of new energy projects as described in claim 9, characterized in that, The steps for forming a closed-loop file for short-term capacity forecasting and driver diagnosis are as follows: Read the short-term capacity-driven contribution record, and receive the actual capacity feedback data after the corresponding capacity forecast period ends. Match the actual capacity feedback data with the short-term capacity forecast results in the same window according to the same capacity forecast period. Combine the short-term capacity-driven contribution record to confirm the environmental-driven deviation, time-delay drift deviation and non-environmental operation deviation, and obtain the hierarchical write-back object set. Based on the hierarchical write-back object set, the time-delay correction prediction window and environmental factor capacity response time-delay fingerprint of the next prediction period are corrected. The non-environmental isolation segments corresponding to non-environmental operation deviations, capacity and environment window sampling records, capacity and environment reliable sample sets, environmental phase capacity response maps, environmental factor capacity response time-delay fingerprints, short-term capacity prediction results, counterfactual capacity diagnosis results and short-term capacity driving contribution records are associated and encapsulated to form a short-term capacity prediction and driving diagnosis closed-loop archive.

Images