Grading evaluation method and system for new energy power prediction model

Abstract

The invention relates to a grading evaluation method and system for a new energy power prediction model, and belongs to the technical field of power prediction. The method comprises the steps of dividing historical operation data of a new energy station into multi-dimensional scenes according to meteorological conditions, equipment operation time periods and power fluctuation characteristics, and matching evaluation indexes for each scene; processing the evaluation index to obtain a first evaluation result; according to the first evaluation result, a second evaluation result of the single prediction model in the scene c is obtained by integrating the key evaluation indexes according to the first weight; calculating the weight of the scene c, and obtaining the third-level evaluation of each prediction model in the scene c according to the weight and the second evaluation result; based on the third-level evaluation, obtaining a second-level evaluation of each prediction model under the scene c dimension d; and according to the second-level evaluation, obtaining the first-level evaluation of each prediction model under the multi-scene dimension. According to the method, multi-dimensional diagnosis of the performance of the prediction model is realized, and the evaluation efficiency is improved.

Description

Technical Field

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

[0002] As the proportion of new energy sources in the power system continues to increase, new energy power forecasting plays a crucial role in power system dispatch. Accurate power forecasting can effectively improve the absorption capacity of high-proportion new energy sources and ensure the stable operation of the power system. However, existing evaluation systems for new energy power forecasting models have significant limitations.

[0003] On the one hand, the installed capacity of new energy sources such as wind power and photovoltaics continues to grow. Their output is affected by multiple factors, including sudden weather changes, time-of-day characteristics, and power fluctuations, exhibiting strong randomness and scenario dependence. Traditional "single-indicator, global evaluation" methods, such as relying solely on mean absolute error (MAE) or root mean square error (RMSE) as evaluation indicators, struggle to objectively reflect the adaptability differences of prediction models under special scenarios such as sandstorms, icing, and power fluctuations, leading to a disconnect between evaluation results and actual dispatch needs. On the other hand, the construction of new power systems places higher demands on the scenario-based performance of prediction models. Differentiated evaluations of models are needed under different meteorological conditions, peak and valley periods of grid load, and power output fluctuation characteristics to support refined management and control of dispatch decisions.

[0004] Furthermore, existing evaluation systems suffer from two main shortcomings: First, static indicator weights cannot adapt to the dynamic needs of multiple scenarios. Existing methods typically employ fixed combinations of evaluation indicators, failing to consider the varying sensitivity of prediction errors to special meteorological scenarios such as strong winds and low temperatures, and also failing to differentiate evaluation weights between peak and non-critical periods. This results in the prediction shortcomings of key scenarios being masked by global averaging. Second, fragmented scenario dimensions lead to one-sided evaluations. Most systems lack a hierarchical fusion mechanism of "single scenario - scenario dimension - global performance," making it impossible to quantify the model's balanced performance across different dimensions such as meteorology, volatility, and time periods, thus hindering targeted model optimization. Summary of the Invention

[0005] Therefore, the technical problem to be solved by the present invention is to overcome the problem that the existing technology fails to fully consider the impact of different meteorological conditions, power fluctuation characteristics and differences in the importance of different operating periods on the performance of the prediction model, resulting in a single and one-sided evaluation result that is difficult to fully reflect the real performance of the model in the actual complex and ever-changing operating environment, and cannot effectively support model optimization and scheduling decisions.

[0006] Firstly, to address the aforementioned technical problems, this invention provides a hierarchical evaluation method for new energy power prediction models, comprising: Acquire historical operating data of new energy power plants; divide the historical operating data into multi-dimensional scenarios based on meteorological conditions, equipment operating periods, and power fluctuation characteristics; For each scenario in the multi-dimensional scenario, key evaluation indicators are selected; the key evaluation indicators are normalized to obtain a normalized first evaluation result; based on the first evaluation result, the key evaluation indicators are combined with a first weight to obtain the scenario. c The second evaluation result of the next single prediction model; wherein the scenario c For any scene in the multi-dimensional scene; Calculate the scenario c The second weight is used to obtain the prediction model for each scenario based on the second weight and the second evaluation result. c The third level of evaluation below; Based on the aforementioned third-level evaluation, the dimensionality of each prediction model is calculated. d The original score is used to obtain the prediction model for each scenario based on the original score. c Dimension d The second level of evaluation below; among which d It is a positive integer greater than 1; Based on the second-level evaluation, the first-level evaluation of each prediction model in the multi-dimensional scenario is calculated; based on the first-level evaluation, the final score of each prediction model is obtained.

[0007] In one embodiment of the present invention, the key evaluation indicators are normalized to obtain a normalized first evaluation result; based on the first evaluation result, and by combining the key evaluation indicators with a first weight, a scenario is obtained. c The steps for the second evaluation result of a single prediction model are as follows: The key evaluation indicators include positive indicators and negative indicators; the positive indicators and negative indicators are aligned in direction; the method for aligning the direction is as follows: ; in, For the scene c Down k The first evaluation result after indicator normalization. For the scene c Down k The evaluation results of the indicators This represents the function that takes the maximum value. This represents the function that takes the minimum value. Indicates the mean absolute error. This represents the root mean square error. Represents the correlation coefficient. Indicates average directional accuracy; The positive indicators include average directional accuracy and correlation coefficient; the negative indicators include mean absolute error and root mean square error; based on the first evaluation result, and by combining the mean absolute error, root mean square error, correlation coefficient, and average directional accuracy with the same first weight, a scenario is obtained. c The second evaluation result for the next single prediction model is expressed as follows: ; in, This indicates the second evaluation result.

[0008] In one embodiment of the present invention, the scenario is calculated. c The second weight is used to obtain the prediction model for each scenario based on the second weight and the second evaluation result. c The steps for the third-level evaluation are as follows: Calculate the scenario c The second weight, based on the second weight, assigns the scenario. c Total score; Rankings are assigned based on the second evaluation results, and the total score is used to calculate the prediction model for each scenario. c The score is given below, and the expression for the score is: ; in, Indicates the first A prediction model in the scenario c The score below, Representing the scenario c Total score, i Indicates the predictive model in the scene c The ranking of performance below.

[0009] In one embodiment of the present invention, the scenario c The expression for the second weight is: ; in, Indicates the second weight. Representing a scene c Subjective experience adjustment coefficient, Representing a scene c Historical frequency of occurrence; Representing a scene c The resulting profit loss; γ represents the objective data weighting coefficient. This indicates the total number of multi-dimensional scenarios.

[0010] In one embodiment of the present invention, each prediction model is calculated in dimensionality based on the third-level evaluation. d The original score is used to obtain the prediction model for each scenario based on the original score. c Dimension d The second-level evaluation method is as follows: ; in, Indicates the first Each prediction model in the scenario dimension d The original score below, Representing dimensions d The collection of all scenes below, Indicates the first A prediction model in the scenario c The score below.

[0011] In one embodiment of the present invention, the method for calculating the first-level evaluation of each prediction model in a multi-dimensional scenario based on the second-level evaluation is as follows: ; in, Indicates the first The final score of each prediction model, Representing dimensions d The weighting coefficients, Indicates the first Each prediction model in the scenario dimension d The original score below, D Indicates the total number of dimensions.

[0012] In one embodiment of the present invention, the method for dividing the historical operating data into multi-dimensional scenarios based on meteorological conditions, equipment operating periods, and power fluctuation characteristics is as follows: the historical operating data is classified according to meteorological dimension, power fluctuation dimension, and time period dimension; wherein, the meteorological dimension represents the scenario of dividing the prediction period according to meteorological monitoring data; the power fluctuation dimension represents the scenario of dividing the prediction period according to the power change rate; and the time period dimension represents the scenario of dividing the prediction period according to the importance of the equipment operating period.

[0013] In one embodiment of the present invention, the scenarios for dividing the prediction period according to meteorological monitoring data include strong wind scenarios, sandstorm scenarios, cold wave scenarios, and low temperature scenarios; the scenarios for dividing the prediction period according to the power change rate include stable output scenarios and fluctuating output scenarios; and the scenarios for dividing the prediction period according to the importance of the equipment operating period include morning peak period scenarios, midday peak period scenarios, evening peak period scenarios, and non-key period scenarios.

[0014] In one embodiment of the present invention, the criteria for classifying the scenarios as strong wind, sandstorm, cold wave, and low temperature include: The criteria for classifying a windy scene as described above are instantaneous wind speeds reaching or exceeding 15 meters per second and lasting for more than 30 minutes. The criteria for classifying a sandstorm scene are: horizontal visibility of less than 1 kilometer and PM10 concentration of 500 micrograms per cubic meter or more. The criteria for classifying a cold wave scenario are a temperature drop of 10 degrees Celsius or more within 24 hours, with the lowest temperature not exceeding 0 degrees Celsius. The standard for classifying a low-temperature scenario is a daily average temperature not exceeding -20 degrees Celsius.

[0015] Secondly, to solve the above-mentioned technical problems, the present invention provides a hierarchical evaluation system for a new energy power prediction model, comprising: A segmentation module is used to acquire historical operating data of new energy power plants; based on meteorological conditions, equipment operating periods, and power fluctuation characteristics, the historical operating data is divided into multi-dimensional scenarios. The evaluation module is used to select key evaluation indicators for each scenario in the multi-dimensional scenario; normalize the key evaluation indicators to obtain a normalized first evaluation result; and based on the first evaluation result, synthesize the key evaluation indicators with a first weight to obtain the scenario. c The second evaluation result of the next single prediction model; wherein the scenario c For any scene in the multi-dimensional scene; The third-level evaluation module is used to calculate the scenario. c The second weight is used to obtain the prediction model for each scenario based on the second weight and the second evaluation result. c The third level of evaluation below; The second-level evaluation module is used to calculate the dimensionality of each prediction model based on the third-level evaluation. d The original score is used to obtain the prediction model for each scenario based on the original score. c Dimension d The second level of evaluation below; among which d It is a positive integer greater than 1; The first-level evaluation module is used to calculate the first-level evaluation of each prediction model in the multi-dimensional scenario based on the second-level evaluation; and to obtain the final score of each prediction model based on the first-level evaluation.

[0016] Compared with the prior art, the above-described technical solution of the present invention has the following advantages: (1) The hierarchical evaluation method and system for a new energy power prediction model described in this invention accurately quantifies the characteristics of key scenarios such as strong winds, icing, dust storms, power fluctuations, and peak grid periods through a three-dimensional dynamic scenario division mechanism of meteorology-fluctuation-time period. This solves the problem that the traditional "single index, global evaluation" method is difficult to reflect the adaptability of the model under special scenarios, and makes the evaluation results closely follow the actual laws of strong randomness and scenario dependence of new energy output. In particular, it provides accurate basis for model selection in high-penetration distributed photovoltaic areas.

[0017] (2) This invention effectively overcomes the one-sidedness caused by the fragmentation of dimensions in traditional evaluation by adopting a three-level evaluation architecture of "single scenario scoring - scenario dimension aggregation - multi-dimensional fusion". This architecture not only identifies the failure risk of the model in high-impact risk scenarios such as icing and cold waves, but also reveals the balanced performance of the model in different dimensions such as meteorological adaptability, power fluctuation suppression ability, and time-period response characteristics, pointing out a clear path for the targeted optimization of the prediction model.

[0018] (3) This invention constructs a dynamic weighting adaptation mechanism that integrates objective data and expert experience. The scenario weights combine historical occurrence frequency, potential profit loss quantification value and expert experience adjustment coefficient to ensure that the evaluation weights of peak power grid periods and high loss scenarios are significantly improved; the configurable dimension weights allow users to flexibly adjust the evaluation orientation according to actual dispatching needs (such as focusing on weather change response or fluctuation mitigation capabilities), promoting the transformation of the evaluation system from "static weighting" to "demand-driven", and enhancing the supporting value of evaluation results for the refined dispatching decision-making of the new power system. Attached Figure Description

[0019] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart of a graded evaluation method for a new energy power prediction model in a preferred embodiment of the present invention; Figure 2 This is a flowchart of obtaining the total score of the prediction model in a multi-dimensional scenario in a preferred embodiment of the present invention; Figure 3 This is a structural diagram of a graded evaluation system for a new energy power prediction model in a preferred embodiment of the present invention. Detailed Implementation

[0020] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0021] Example 1: Reference Figure 1As shown, this embodiment of the invention provides a hierarchical evaluation method for a new energy power prediction model, including but not limited to the following steps: S1. Obtain historical operating data of new energy power plants; divide the historical operating data into multi-dimensional scenarios based on meteorological conditions, equipment operating periods, and power fluctuation characteristics; S2. Select key evaluation indicators for each scenario in the multi-dimensional scenario; normalize the key evaluation indicators to obtain the normalized first evaluation result; based on the first evaluation result, and with the first weight, synthesize the key evaluation indicators to obtain the scenario. c The second evaluation result of the next single prediction model; where the scenario c For any scene in a multi-dimensional scenario; S3, Computing Scenarios c The second weight is used to obtain the prediction model's performance in the scenario based on the second weight and the second evaluation result. c The third level of evaluation below; S4. Based on the third-level evaluation, calculate the dimensionality of each prediction model. d The original score is used to obtain the prediction model's performance in the scene. c Dimension d The second level of evaluation below; among which d It is a positive integer greater than 1; S5. Based on the second-level evaluation, calculate the first-level evaluation of each prediction model in multiple scenario dimensions; based on the first-level evaluation, obtain the final score of each prediction model.

[0022] This invention provides a hierarchical evaluation method for new energy power prediction models, constructing a three-level evaluation architecture of scenario scoring, dimensional scoring, and model total score through first-level evaluation, second-level evaluation, and third-level evaluation. Specifically, the third-level evaluation allocates individual scenario scores based on scenario weights; the second-level evaluation aggregates scenario scores to output dimensional scores; and the first-level evaluation integrates the scores from each dimension to generate the total prediction model score. This evaluation structure effectively overcomes the one-sidedness caused by the fragmented dimensions in traditional evaluation methods. Based on meteorological conditions, equipment operating periods, and power fluctuation characteristics, this method performs refined division and hierarchical fusion of three-dimensional scenarios, accurately quantifying the adaptability differences of prediction models in key scenarios such as icing, dust storms, power fluctuations, and peak grid conditions, thus solving the problem of the evaluation results being disconnected from actual dispatching needs. Furthermore, this method normalizes key evaluation indicators and, through a multi-level weighted comprehensive evaluation system, can scientifically quantify and objectively compare the performance of different prediction models. Normalization eliminates the influence of differences in dimensions and orders of magnitude between different indicators, making the comparison and synthesis between different indicators more scientific and reasonable. The multi-level weighted comprehensive evaluation system can fully consider the differences in importance of different indicators and scenarios. Through flexible weight allocation, it makes the evaluation results more objectively reflect the comprehensive performance of the prediction model in multi-dimensional scenarios, avoiding the one-sidedness caused by evaluation of a single indicator or scenario. The evaluation method described in this embodiment of the invention only needs to use historical power and meteorological data to achieve multi-dimensional diagnosis of the performance of the prediction model, significantly improving the evaluation efficiency. This method provides a scientific basis for the selection of prediction models, algorithm optimization, and scheduling strategy formulation in high-penetration distributed photovoltaic areas, effectively supporting the safe and stable operation of the new power system.

[0023] Specifically, in step S1, historical operating data of the new energy power station is acquired; the specific steps for dividing the historical operating data into multi-dimensional scenarios based on meteorological conditions, equipment operating periods, and power fluctuation characteristics are as follows: S110. Obtain historical operating data of new energy power stations, including wind speed, solar intensity, temperature, power generation, and equipment status.

[0024] S120. Based on meteorological conditions, power fluctuation characteristics, and equipment operating periods (also known as grid periods), historical operating data is divided into multi-dimensional dynamic scenarios. The method of division is as follows: historical operating data is classified according to three dimensions: meteorological dimension, power fluctuation dimension, and time period dimension. In this embodiment of the invention, the preferred time resolution of the operating data is 15 minutes. The meteorological dimension is based on meteorological monitoring data to divide the predicted time period into scenarios such as general meteorological scenarios, strong wind scenarios, sandstorm scenarios, cold wave scenarios, low temperature scenarios, snow cover scenarios, and icing scenarios. The power fluctuation dimension is based on the power change rate to divide the predicted time period into scenarios such as stable output scenarios and fluctuating output scenarios. The time period dimension is based on the importance of the equipment operating period into scenarios such as morning peak time scenarios, midday peak time scenarios, evening peak time scenarios, and non-critical time period scenarios.

[0025] S130. Establish a specific classification criterion for each scenario, based on the numerical or temporal characteristics of meteorology and power.

[0026] Specifically, in the meteorological dimension, the classification criteria include: S131. Criteria for classifying strong wind scenarios: Instantaneous wind speeds reaching or exceeding 15 meters per second and lasting for more than 30 minutes. The specific expression is: ; in, for Wind speed at all times; Indicates time; This is an indicator function; its value is 1 when the condition is true and 0 otherwise.

[0027] S132. Sandstorm Scene Classification Standard: Horizontal visibility is less than 1 kilometer and PM10 concentration reaches or exceeds 500 micrograms per cubic meter. The specific expression is: ; in, For a moment Horizontal visibility, For a moment PM10 concentration.

[0028] S133. If the assessment object is a photovoltaic power station, then the snow-covered scenario shall be considered, and the criteria for classification are a snow thickness of 2 cm or more and an irradiance of less than 300 watts per square meter. The specific expression is: ; in, For a moment Irradiance, For a moment The thickness of the snow cover.

[0029] S134. If the assessment object is a wind farm, then the icing scenario is considered. The criteria for classification are an icing thickness of 3 mm or more and a wind speed not exceeding the wind turbine's cut-in wind speed. The specific expression is: ; in, For a moment Wind speed, To cut into wind speed, For a moment The thickness of the ice cover.

[0030] It should be noted that the time span for the judgment criteria in the above scenarios S131 to S134 is in units of 15 minutes.

[0031] S135. Criteria for classifying cold wave scenarios: A temperature drop of 10 degrees Celsius or more within 24 hours, with the lowest temperature not exceeding 0 degrees Celsius. The specific expression is: ; in, For a moment temperature.

[0032] S136. Criteria for classifying low-temperature scenarios: Daily average temperature not exceeding -20 degrees Celsius. The specific expression is: ; in, This represents the serial number, with an initial value of 1.

[0033] It should be noted that the time span for the above scenario criteria is in days.

[0034] S137, and the remaining time periods are classified as general weather scenarios.

[0035] Specifically, the criteria for classifying power fluctuations include: stable output scenarios are defined as power fluctuations not exceeding 10% of the rated capacity within one hour; fluctuating output scenarios are defined as power fluctuations exceeding 10% of the rated capacity within one hour. The time span for these scenario criteria is in hours.

[0036] Furthermore, the specific expression for classifying fluctuating power output scenarios is as follows: ; ; in, For time In power scenarios, For a moment Power value, This refers to the rated capacity of the station.

[0037] Specifically, the criteria for classifying time periods within the device runtime dimension include: 10:00 AM to 2:00 PM as the midday peak period; 7:00 AM to 9:00 AM as the morning peak period; 6:00 PM to 7:30 PM as the evening peak period; and the remaining time periods as non-peak periods. The time span for these scenario criteria is measured in hours. The segmentation function for the time period dimension is as follows. The expression is: .

[0038] Compared with existing technologies, the embodiments of the present invention break through the limitations of traditional static global evaluation. Through the fine division and hierarchical fusion mechanism of three-dimensional scenarios of meteorology-fluctuation-time period, the adaptability differences of the prediction model in key scenarios such as icing, dust storms, power fluctuations and power grid peaks are accurately quantified, thus solving the problem of the disconnect between evaluation results and actual scheduling needs.

[0039] Specifically, refer to Figure 2 As shown, the specific steps for step S2 are as follows: S210. For each scenario in the multi-dimensional scenario, select a normalized set of key evaluation indicators. In this embodiment of the invention, the set of key evaluation indicators includes mean absolute error (MAE), root mean square error (RMSE), mean orientation accuracy (MDA), and correlation coefficient (R). The functions and specific expressions of the four evaluation indicators are as follows: The mean absolute error (MAE) reflects the absolute level of prediction error and is insensitive to outliers. Its specific expression is as follows: (1).

[0040] The root mean square error (RMSE) is used to emphasize the penalty for large errors and requires higher prediction stability. Its specific expression is: (2).

[0041] Mean Directional Accuracy (MDA) is used to evaluate the predictive ability of power variation trends, and its specific expression is: (3); (4).

[0042] The correlation coefficient (R) measures the similarity in shape between the predicted curve and the actual curve, reflecting the ability to grasp the fluctuation pattern. Its specific expression is as follows: (5); in, It is the total number of samples. yes Actual power value at any time yes Predict power values ​​at all times. The symbol indicates the objective value to be obtained. It is the average actual power. It is the predicted average power.

[0043] Reference Figure 2 As shown, in the meteorological dimension, scores for scenarios such as strong winds, sandstorms, cold waves, low temperatures, snow cover, and ice cover can be obtained through evaluation indicators; in the power fluctuation dimension, scores for stable output and fluctuating output scenarios can be obtained; and in the equipment operating time dimension, scores for morning, noon, evening peak hours, and non-peak hours scenarios can be obtained.

[0044] S220. The four evaluation indicators are normalized using the extreme value normalization method, and the directions of the positive indicators MDA and R are aligned with those of the negative indicators MAE and RMSE. The method for unifying the directions is as follows: (6); in, For the scene c Down k The first evaluation result after indicator normalization. For the scene c Down k The evaluation results of the indicators This represents the function that takes the maximum value. This represents the function that takes the minimum value.

[0045] S230, Based on the first evaluation result The scene is obtained by combining the mean absolute error, root mean square error, correlation coefficient, and mean orientation accuracy calculation results with the same first weight. c The second evaluation result of the next single prediction model, the second evaluation result The expression is: (7); Among them, the second evaluation result The larger the value, the better the model performs in the scene. c The better the performance.

[0046] Specifically, in step S3, the third-level evaluation is used to represent the performance assessment of each model in a certain scenario. The specific steps for defining the third-level evaluation in the comprehensive evaluation system are as follows: S310, Computing Scenarios c The second weight is used to assign scenarios. c The total score. The second weighting... The specific expression is: (8); in, This represents the second weight, for the scenario. c The weights before normalization; Representing a scene c The subjective experience adjustment coefficient is set by experts, with a default value of 1. Representing a scene c Historical frequency of occurrence; Representing a scene c The resulting profit loss; γ represents the objective data weighting coefficient, ranging from 0 to 1; This indicates the total number of multi-dimensional scenarios.

[0047] Furthermore, allocate scenarios c The total score is expressed as follows: (9); in, Representing a scene c Total score.

[0048] S320. Assign scenario scores according to the ranking of the prediction model. Specifically, based on the second evaluation results... Rank and assign based on total score. The calculation yields the prediction model for each scenario. c The score is given below, where the expression for the score is: (10); in, Indicates the first A prediction model in the scenario c The score below; i This indicates that the prediction model is in this scenario (i.e., scenario). c The performance ranking under ) when i= A value of 1 indicates optimal performance.

[0049] Specifically, in step S4, the second-level evaluation in the comprehensive evaluation system is defined, representing the overall performance assessment of each model in a certain scenario dimension. This is based on the scores in the third-level evaluation. The computational model in dimensionality d The original score below: (11); in, Indicates the first Each prediction model in the scenario dimension d The original score below, Representing dimensions dThe set of all scenes below. The calculated raw scores... As a second-level evaluation.

[0050] Specifically, in step S5, the first-level evaluation in the comprehensive evaluation system is defined, representing the comprehensive performance evaluation result of each model across multiple scenario dimensions. This is based on the raw scores in the second-level evaluation. The total score of the prediction model is calculated using the following expression: (12); in, Indicates the first The final score of each prediction model, Representing dimensions d The weight coefficients must be set, and the sum of the weights of each dimension must be 1. Indicates the first Each prediction model in the scenario dimension d The original score below, D This represents the total number of dimensions. The calculated total score is used as the first-level evaluation.

[0051] In step S3, an innovative dynamic weighting adaptation mechanism integrating objective data and expert experience is constructed. Scenario weights are determined by a combination of historical occurrence frequency, quantified potential profit loss, and expert-experienced adjustment coefficients, ensuring a significant increase in evaluation weights for peak power grid periods and high-loss scenarios. The configurable dimensional weights allow users to flexibly adjust the evaluation orientation according to actual dispatching needs (such as emphasizing responses to sudden weather changes or fluctuation mitigation capabilities), driving the evaluation system from static weighting to demand-driven approaches and enhancing the supporting value of evaluation results for refined dispatching decisions in the new power system.

[0052] The three-level comprehensive evaluation system for renewable energy power prediction constructed in this invention provides a precise quantitative tool for the scenario-based performance of renewable energy power prediction models through multi-dimensional dynamic scenario division and hierarchical fusion scoring mechanism. This system addresses the blind spots in the adaptive assessment of traditional global evaluation methods in key scenarios such as extreme weather, power fluctuations, and grid peak-valley conditions, improving the alignment between evaluation results and dispatch decision requirements. It provides a scientific basis for the selection, optimization, and dispatch strategy formulation of prediction models in high-penetration distributed photovoltaic areas. Simultaneously, the dynamic weight adaptation mechanism enhances the evaluation system's responsiveness to actual grid risks, effectively supporting the safe and stable operation of new power systems and promoting the evolution of renewable energy prediction technology towards refinement and scenario-based application.

[0053] Furthermore, this invention constructs an evaluation system that integrates dynamic scenario segmentation, multi-dimensional indicator matching, and hierarchical aggregation. By quantifying the adaptability of the model in subdivided scenarios, it effectively overcomes the one-sidedness caused by the fragmented dimensions of traditional evaluations. This system achieves three core breakthroughs: first, it provides refined multi-dimensional scenario classification based on meteorological monitoring, power fluctuation rate, and equipment operating time; second, it establishes a dynamic mapping mechanism between scenario characteristics and evaluation indicators; and third, it achieves deep coupling between evaluation results and power grid dispatching needs through a three-level scoring architecture (scenario score - dimension score - model total score), providing precise guidance for the iterative optimization of new energy prediction models.

[0054] Example 2: Based on the same inventive concept, this embodiment provides a hierarchical evaluation system for a new energy power prediction model. The principle of solving the problem is similar to the hierarchical evaluation method for a new energy power prediction model provided in Embodiment 1, and the repeated parts will not be described again.

[0055] Reference Figure 3 As shown, this embodiment provides a hierarchical evaluation system for a new energy power prediction model, including: The module is used to acquire historical operating data of new energy power plants; based on meteorological conditions, equipment operating periods, and power fluctuation characteristics, the historical operating data is divided into multi-dimensional scenarios. The evaluation module is used to select key evaluation indicators for each scenario in the multi-dimensional scenario; normalize the key evaluation indicators to obtain a normalized first evaluation result; and based on the first evaluation result, integrate the key evaluation indicators with a first weight to obtain the scenario. c The second evaluation result of the next single prediction model; where the scenario c For any scene in a multi-dimensional scenario; The third-level evaluation module is used to calculate the scenario. c The second weight is used to obtain the prediction model's performance in the scenario based on the second weight and the second evaluation result. c The third level of evaluation below; The second-level evaluation module is used to calculate the performance of each prediction model in dimensionality based on the third-level evaluation. d The original score is used to obtain the prediction model's performance in the scene. c Dimension d The second level of evaluation below; among which d It is a positive integer greater than 1; The first-level evaluation module is used to calculate the first-level evaluation of each prediction model in the multi-dimensional scenario based on the second-level evaluation; and to obtain the final score of each prediction model based on the first-level evaluation.

[0056] Example 3: This embodiment provides a storage medium for storing a computer program. When the computer program is executed by a processor, it implements a graded evaluation method for a new energy power prediction model as described above.

[0057] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0058] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0059] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0060] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0061] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A hierarchical evaluation method for a new energy power prediction model, characterized in that, include: Obtain historical operation data of new energy power plants; Based on meteorological conditions, equipment operating periods, and power fluctuation characteristics, the historical operating data is divided into multi-dimensional scenarios; Select key evaluation indicators for each scenario in the multi-dimensional scenario; The key evaluation indicators are normalized to obtain the normalized first evaluation result; Based on the first evaluation result, and by comprehensively considering the key evaluation indicators with a first weight, the scenario is obtained. c The second evaluation result of the next single prediction model; wherein the scenario c For any scene in the multi-dimensional scene; Calculate the scenario c The second weight is used to obtain the prediction model for each scenario based on the second weight and the second evaluation result. c The third level of evaluation below; Based on the aforementioned third-level evaluation, the dimensionality of each prediction model is calculated. d The original score is used to obtain the prediction model for each scenario based on the original score. c Dimension d The second level of evaluation below; in d It is a positive integer greater than 1; Based on the second-level evaluation, the first-level evaluation of each prediction model in the multi-dimensional scenario is calculated. The final score for each prediction model is obtained based on the first-level evaluation.

2. The hierarchical evaluation method for a new energy power prediction model according to claim 1, characterized in that, The key evaluation indicators are normalized to obtain a normalized first evaluation result; based on the first evaluation result, and by combining the key evaluation indicators with a first weight, a scenario is obtained. c The steps for the second evaluation result of a single prediction model are as follows: The key evaluation indicators include positive indicators and negative indicators; the positive indicators and negative indicators are aligned in direction; the method for aligning the direction is as follows: ; in, For the scene c Down k The first evaluation result after indicator normalization. For the scene c Down k The evaluation results of the indicators This represents the function that takes the maximum value. This represents the function that takes the minimum value. Indicates the mean absolute error. This represents the root mean square error. Represents the correlation coefficient. Indicates average directional accuracy; The positive indicators include average directional accuracy and correlation coefficient; the negative indicators include mean absolute error and root mean square error; based on the first evaluation result, and by combining the mean absolute error, root mean square error, correlation coefficient, and average directional accuracy with the same first weight, a scenario is obtained. c The second evaluation result for the next single prediction model is expressed as follows: ； in, This indicates the second evaluation result.

3. The hierarchical evaluation method for a new energy power prediction model according to claim 1, characterized in that, Calculate the scenario c The second weight is used to obtain the prediction model for each scenario based on the second weight and the second evaluation result. c The steps for the third-level evaluation are as follows: Calculate the scenario c The second weight, based on the second weight, assigns the scenario. c Total score; Rankings are assigned based on the second evaluation results, and the total score is used to calculate the prediction model for each scenario. c The score is given below, and the expression for the score is: ； in, Indicates the first A prediction model in the scenario c The score below, Representing the scenario c Total score, i Indicates the predictive model in the scene c The ranking of performance below.

4. A hierarchical evaluation method for a new energy power prediction model according to claim 1 or 3, characterized in that, The scenario c The expression for the second weight is: ； in, Indicates the second weight. Representing a scene c Subjective experience adjustment coefficient, Representing a scene c Historical frequency of occurrence; Representing a scene c The resulting profit loss; γ represents the objective data weighting coefficient. This indicates the total number of multi-dimensional scenarios.

5. The hierarchical evaluation method for a new energy power prediction model according to claim 1, characterized in that, Based on the aforementioned third-level evaluation, the dimensionality of each prediction model is calculated. d The original score is used to obtain the prediction model for each scenario based on the original score. c Dimension d The second-level evaluation method is as follows: Based on the scores in the third-level evaluation, calculate the dimensionality of each prediction model. d The original score is given below, and the expression for the original score is: ； in, Indicates the first Each prediction model in the scenario dimension d The original score below, Representing dimensions d The collection of all scenes below, Indicates the first A prediction model in the scenario c The score below.

6. The hierarchical evaluation method for a new energy power prediction model according to claim 1, characterized in that, Based on the second-level evaluation, the method for calculating the first-level evaluation of each prediction model in a multi-dimensional scenario is as follows: ； in, Indicates the first The final score of each prediction model, Representing dimensions d The weighting coefficients, Indicates the first Each prediction model in the scenario dimension d The original score below, D Indicates the total number of dimensions.

7. The hierarchical evaluation method for a new energy power prediction model according to claim 1, characterized in that, The method for dividing the historical operating data into multi-dimensional scenarios based on meteorological conditions, equipment operating periods, and power fluctuation characteristics is as follows: the historical operating data is classified according to meteorological dimension, power fluctuation dimension, and time period dimension; wherein, the meteorological dimension represents the scenario of dividing the prediction period according to meteorological monitoring data; the power fluctuation dimension represents the scenario of dividing the prediction period according to the power change rate; and the time period dimension represents the scenario of dividing the prediction period according to the importance of the equipment operating period.

8. The hierarchical evaluation method for a new energy power prediction model according to claim 7, characterized in that, The scenarios for dividing prediction periods based on meteorological monitoring data include strong wind scenarios, sandstorm scenarios, cold wave scenarios, and low temperature scenarios; the scenarios for dividing prediction periods based on power change rate include stable output scenarios and fluctuating output scenarios; the scenarios for dividing prediction periods based on the importance of equipment operating time include morning peak time scenarios, midday peak time scenarios, evening peak time scenarios, and non-key time period scenarios.

9. The hierarchical evaluation method for a new energy power prediction model according to claim 8, characterized in that, The criteria for classifying scenarios into high wind, sandstorm, cold wave, and low temperature scenarios include: The criteria for classifying a windy scene as described above are instantaneous wind speeds reaching or exceeding 15 meters per second and lasting for more than 30 minutes. The criteria for classifying a sandstorm scene are: horizontal visibility of less than 1 kilometer and PM10 concentration of 500 micrograms per cubic meter or more. The criteria for classifying a cold wave scenario are a temperature drop of 10 degrees Celsius or more within 24 hours, with the lowest temperature not exceeding 0 degrees Celsius. The standard for classifying a low-temperature scenario is a daily average temperature not exceeding -20 degrees Celsius.

10. A hierarchical evaluation system for a new energy power prediction model, characterized in that, include: The module is divided to obtain historical operating data of new energy power stations; Based on meteorological conditions, equipment operating periods, and power fluctuation characteristics, the historical operating data is divided into multi-dimensional scenarios; The evaluation module is used to select key evaluation indicators for each scenario in the multi-dimensional scenario. The key evaluation indicators are normalized to obtain the normalized first evaluation result; Based on the first evaluation result, and by comprehensively considering the key evaluation indicators with a first weight, the scenario is obtained. c The second evaluation result of the next single prediction model; wherein the scenario c For any scene in the multi-dimensional scene; The third-level evaluation module is used to calculate the scenario. c The second weight is used to obtain the prediction model for each scenario based on the second weight and the second evaluation result. c The third level of evaluation below; The second-level evaluation module is used to calculate the dimensionality of each prediction model based on the third-level evaluation. d The original score is used to obtain the prediction model for each scenario based on the original score. c Dimension d The second level of evaluation below; in d It is a positive integer greater than 1; The first-level evaluation module is used to calculate the first-level evaluation of each prediction model in the multi-dimensional scenario based on the second-level evaluation. The final score for each prediction model is obtained based on the first-level evaluation.

Images