Multi-dimensional network performance index evaluation system and method based on measured data of new energy station

Abstract

The invention relates to the technical field of power systems, in particular to a multi-dimensional network performance index evaluation system and method based on measured data of a new energy station. Comprising a data acquisition preprocessing unit; a multi-dimensional index construction unit; the weight adjustment unit dynamically adjusts the index weight according to the operation condition based on a fusion algorithm of an improved analytic hierarchy process and an entropy weight method; the performance evaluation model unit adopts a comprehensive evaluation model fused with a random forest algorithm to generate multi-dimensional performance scores and risk early warning; and the visual interaction terminal unit is used for displaying the evaluation result in real time and generating a trend analysis report. According to the invention, the three-dimensional evaluation system covering the static electrical parameters, the dynamic stability indexes and the harmonic pollution indexes is constructed, so that the limitation of single-dimensional evaluation in the prior art is broken through, and the extension of the network-related performance of the new energy station from local detection to full-dimensional coverage is realized; and the multi-scene evaluation requirements of the power grid on the steady-state operation, the transient response and the electric energy quality of the station are accurately matched.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a multi-dimensional network-related performance index evaluation system and method based on measured data from new energy stations. Background Art

[0002] With the large-scale integration of new energy stations into the grid, grid-related performance evaluation has become a core component in ensuring grid stability. Currently, new energy stations operate under complex and changing conditions. Traditional evaluation methods, with their single indicator dimension and fixed weighting mechanism, struggle to meet the precise assessment requirements in dynamic scenarios. Therefore, a comprehensive evaluation system integrating multi-source data and intelligent algorithms is urgently needed.

[0003] For example, Chinese patent CN202410284904.5 discloses a remote testing method and system for the grid-related performance of a new energy station. The system includes: obtaining an associated sub-grid of the new energy station, and performing an active disturbance test on the associated sub-grid, monitoring and analyzing the disturbance degree evaluation index of the associated sub-grid; according to the disturbance degree evaluation index of the associated sub-grid, analyzing the disturbance degree level of the associated sub-grid, and matching to obtain an ideal value for the grid-related performance evaluation of the new energy station; based on the ideal value for the grid-related performance evaluation of the new energy station, evaluating the grid-related performance of the new energy station, and obtaining feedback on the grid-related performance evaluation results of the new energy station. The present invention evaluates the degree of grid disturbance, accurately simulates various grid conditions that may occur during the grid connection process, and at the same time evaluates the response performance and disturbance suppression performance of the new energy station, comprehensively analyzes the grid-related performance of the new energy station, and helps to maintain the stability of the grid. For example, Chinese patent CN202211392622.4 discloses a PMU-based new energy station performance indicator monitoring system and its application method, including a performance indicator online monitoring master station unit and a PMU unit. The PMU unit is used to collect the grid connection point voltage / current of the new energy station AGC system, AVC system, primary frequency modulation system, wind turbine EMS system, wind turbine main control system, and photovoltaic data acquisition system in real time and send it to the performance indicator online monitoring master station unit. The performance indicator online monitoring master station unit is used to analyze the data collected by the PMU unit in real time and in parallel and evaluate the performance indicators of each new energy station. This invention can realize the transformation of performance indicators from an offline intermittent detection method to an online full-time, full-process online monitoring and evaluation method, which is conducive to strengthening the power grid's real-time control of the network-related performance of new energy stations.

[0004] Although the above technologies all have their design advantages, the above technical solutions still have the following technical defects: First, the evaluation dimensions and mechanisms are limited: Chinese patent CN202410284904.5 relies on active disturbance testing to simulate the power grid conditions, and has not built a multi-dimensional evaluation system covering static electrical indicators, dynamic stability indicators and harmonic pollution indicators. The weight distribution is fixed and cannot adapt to dynamic working conditions such as power fluctuations and meteorological changes in new energy sites; Chinese patent CN202211392622.4 realizes online data monitoring based on PMU units, but only analyzes indicators through single data. It lacks the integration mechanism of expert experience and dynamic working conditions, lacks adaptability to the weight adjustment of dynamic indicators such as voltage sag recovery capability, and is difficult to coordinate the accurate evaluation of multi-dimensional indicators. Second, there is a lack of dynamic adaptability and correction: The preset disturbance test scenarios in CN202410284904.5 are unable to cope with the complex operating conditions that randomly arise in actual grid operation, resulting in biased evaluation results. The online monitoring in CN202211392622.4 lacks a model-adaptive correction mechanism. When grid disturbances or sudden changes in renewable energy output occur, the evaluation logic cannot be adjusted promptly, resulting in delayed grid-related performance evaluation. Third, there is insufficient integration and coordination of subjective and objective factors: Neither method effectively integrates expert experience with real-time operating data. CN202410284904.5 focuses on physical disturbance testing, ignoring industry experts' understanding of the importance of indicators. CN202211392622.4 focuses on data collection and fails to leverage expert experience to optimize weight distribution, making it difficult to balance the impact of subjective and objective factors on the evaluation results. In light of this, we propose a multidimensional grid-related performance indicator evaluation system and method based on measured data from renewable energy stations. Summary of the Invention

[0005] The purpose of the present invention is to provide a multi-dimensional network-related performance index evaluation system and method based on the measured data of new energy stations, so as to solve the problems raised in the above background technology, such as the limitations of evaluation dimensions and mechanisms, the lack of dynamic adaptability and correction, and the insufficient integration and coordination of subjective and objective factors.

[0006] To solve the above technical problems, one of the objectives of the present invention is to provide a multi-dimensional network-related performance index evaluation system based on measured data from new energy stations, including: A data acquisition and preprocessing unit, which is used to synchronously collect electrical operation data and environmental parameters of new energy stations and perform denoising and normalization preprocessing; A multi-dimensional indicator construction unit, which is used to construct a three-dimensional evaluation system including static electrical indicators, dynamic stability indicators and harmonic pollution indicators; A weight adjustment unit, which dynamically adjusts the indicator weights according to operating conditions based on a fusion algorithm of an improved hierarchical analysis method and an entropy weight method; A performance evaluation model unit, which uses a comprehensive evaluation model that integrates a random forest algorithm to generate multi-dimensional performance scores and risk warnings; A visual interactive terminal unit is used to display evaluation results in real time, generate trend analysis reports, and provide a human-computer interaction interface.

[0007] As a further improvement of the present technical solution, the data acquisition preprocessing unit includes a multi-source heterogeneous data acquisition module and a spatiotemporal synchronization preprocessing module, wherein: The multi-source heterogeneous data acquisition module uses a distributed array consisting of 16 synchronized phasor acquisition devices, combined with a meteorological sensor network, and is compatible with the IEEE C37.118.2 synchronization protocol to achieve sub-millisecond synchronous acquisition of electrical parameters (three-phase voltage amplitude, frequency deviation, active power, etc.) and environmental parameters (light intensity, ambient temperature, wind speed, wind direction angle, air pressure); The spatiotemporal synchronization preprocessing module (120) realizes multi-source data timestamp alignment based on the GPS timing system, constructs a three-level cache architecture (L1 cache SRAM stores the latest 10 seconds of data, L2 cache DRAM stores the latest 1 hour of data, and L3 cache SSD array stores historical data), and transmits the preprocessed data to the multi-dimensional indicator construction unit (200) through the OPCUA protocol to realize breakpoint resuming and CRC-32 verification.

[0008] Furthermore, the three-level cache architecture includes: The L1 cache uses high-speed SRAM to store the real-time data of the last 10 seconds to ensure high-frequency access needs; The L2 cache uses DRAM to store the latest hour's data and supports short-term historical data calls; The L3 cache uses an SSD array to store long-term historical data, facilitating subsequent analysis and model iteration.

[0009] As a further improvement of this technical solution, the spatiotemporal synchronization preprocessing module includes an adaptive noise suppression submodule, a dynamic dimension normalization submodule and a data quality assessment submodule, wherein: The adaptive noise suppression submodule adopts a three-stage filtering architecture, which sequentially passes through a hardware low-pass filter, a digital median filter, and an adaptive threshold filter, combined with spectrum analysis and automatic matching filtering strategy to suppress various types of noise interference and improve data reliability in complex electromagnetic environments; The dynamic dimension normalization submodule unifies the dimensions and scales the dynamic range of the collected data, and identifies and repairs outliers through sliding window statistical characteristics to adapt to the wide range of fluctuations in the operating conditions of new energy stations. The data quality assessment submodule performs multi-dimensional quantitative scoring on the pre-processed data based on preset integrity, consistency, and timeliness assessment standards, divides the data into different quality levels, and automatically isolates data that does not meet the quality standards.

[0010] As a further improvement of this technical solution, the multi-dimensional indicator construction unit includes a static electrical indicator module and a dynamic stability indicator module, wherein: The static electrical indicator module constructs a comprehensive voltage quality assessment system based on steady-state operation data. By quantifying voltage deviation, frequency stability, power factor, voltage amplitude fluctuation, and harmonic distortion rate parameters, it generates static evaluation indicators reflecting the basic power quality of the station. Among them, the voltage quality assessment integrates voltage amplitude fluctuation and harmonic distortion rate parameters to form a comprehensive quantitative value to characterize grid compatibility. The dynamic stability index module constructs a dynamic evaluation system including voltage sag recovery capability, frequency regulation rate and active power ramp rate based on the transient response characteristics of the new energy station; and quantifies the stability performance of the new energy station under load fluctuations or faults by capturing the parameter change rate and recovery time during the disturbance process.

[0011] As a further improvement of the present technical solution, the multidimensional index construction unit further includes a harmonic pollution index module, which includes a harmonic separation calculation submodule and a source end positioning analysis submodule, wherein: The harmonic separation calculation submodule uses fast Fourier transform technology to separate the 2-50 harmonic components based on the electrical parameters collected synchronously by multiple channels, calculates the content rate of each harmonic and the total distortion rate of the harmonic voltage, generates a harmonic content distribution matrix, and outputs the harmonic component parameters to the source end positioning analysis submodule; The source end positioning analysis submodule constructs the phase difference matrix of each harmonic ,By setting the phase difference threshold, the main harmonic source equipment is located, a harmonic source positioning map is generated, and the positioning results are output to the performance evaluation model unit as a multi-dimensional scoring parameter.

[0012] As a further improvement of the present technical solution, the weight adjustment unit includes an expert experience library construction module for constructing an expert experience library, and the construction of the expert experience library includes the following steps: S310.1. Select experts and select at least five experts in the fields of new energy grid connection and power system analysis; Expert selection criteria must meet: Years of experience ≥ 10 years, with extensive engineering practice experience; Covering core positions such as scheduling, operation and maintenance, and design, ensuring a multi-dimensional professional perspective; In the past three years, he has participated in ≥2 new energy station network performance evaluation projects and is familiar with industry standards and technical specifications.

[0013] S310.2. Use the Delphi method to conduct multiple rounds of indicator importance ratings: First round: Scale questionnaires were distributed to experts using a 1-9 scale (1 represents an indicator and Equally important, 3 means Slightly more important than , 5 means Obviously more important than , 7 means Strongly important to , 9 means Extremely important , the reciprocal is the reverse importance), collect the original judgment matrix set ,in, , indicating the Experts on indicators and The importance scoring matrix, is the total number of experts, ,in, is the number of indicators; Second round: Summarize all expert scores from the first round and calculate the mean of each matrix element and standard deviation , feedback to experts and invite revision of scores; define coefficient of variation , when all elements When the score converges, the iteration is stopped; S310.3. Perform consistency check on each expert matrix: Calculate the maximum eigenvalue , through the formula Get consistency index ; Check the standard random consistency index table to obtain the corresponding standard random consistency index ; Calculating the consistency ratio ,filter <0.1 matrix, constitutes the candidate set of the experience library ; S310.4. Repeat S310.1, S310.2, and S310.3 every 4-8 months based on new expert opinions or new energy technology iteration data (such as new equipment access, control strategy upgrades), and iteratively optimize the expert experience database to ensure the timeliness and accuracy of expert experience.

[0014] As a further improvement of the present technical solution, the weight adjustment unit further includes a weight calculation module, which calculates the dynamic weight based on the expert experience database, including the following steps: S320.1. Call the candidate matrix set that has passed the consistency test from the expert experience database , according to the experts' years of experience Calculate weights: ; in, is the number of effective experts, satisfy , reflecting the weight of opinions of senior experts; Perform weighted averaging on the candidate matrices to generate the initial judgment matrix : ; in, Represents the initial judgment matrix Middle Rank The elements of the column are; is the number of candidate matrices participating in weighted averaging; For the The weights corresponding to the candidate matrices; For the The candidate matrix Rank Elements of the column; are the row and column indices of the matrix; The matrix integrates the knowledge of multiple experts and balances the professional perspectives of experts in different positions through weight distribution.

[0015] S320.2. Collect the operating data of new energy stations in real time and calculate the active power fluctuation rate and voltage deviation rate, including: Active power fluctuation rate ;in, for Active power at all times, is the active power at the previous moment, is the rated active power of the station; Voltage deviation rate ;in, for The effective value of voltage at the moment, is the rated voltage; S320.3, through the improved Gaussian mixture model Correct the initial matrix: ; in, is the correction coefficient (trained by historical data) to control the correction amplitude; Furthermore, the Gaussian mixture model in the improved hierarchical analysis method is The present invention adopts the optimization design of adaptive bandwidth kernel function and working condition classification correction strategy, which is as follows: To make the model more responsive to power fluctuations, The kernel function introduces a dynamic bandwidth parameter: is the base bandwidth, [0.1,0.3], represents the adjustment coefficient; According to the active power fluctuation rate of the new energy station and voltage deviation rate , the operating conditions are divided into the following three categories, and Set differentiated output ranges to avoid weight logic confusion caused by indifferent corrections: Stable working condition ( ): [-0.1,0.1]; Fluctuating working conditions (5 < 15 or 2 < 5 ): [-0.2,0.2]; Severe fluctuation conditions ( >15 or >5 ): [-0.3,0.3].

[0016] S320.4. Calculate the dynamic consistency ratio for the correction matrix: ; ; in, is the dynamic consistency ratio, which is used to measure the consistency of the correction matrix. is the consistency indicator, is the standard random consistency indicator; is the maximum eigenvalue of the matrix (obtained from the standard table), is the order of the correction matrix; when When >0.1, a secondary correction is triggered: ; in, is the correction intensity factor; is the correction matrix at time , No. Rank The elements of the column are the objects of the secondary correction operation; After the second correction, the matrix is , No. Rank New elements of the column; S320.5, indicator data matrix Perform extreme value normalization: Efficiency indicators (the larger the value, the better, such as power factor): ; Cost-based indicators (the smaller the value, the better, such as harmonic distortion rate): ; in, For the Samples, The standardized value of each indicator; For the Samples, The original measurement value of each indicator; and Respectively The minimum and maximum values ​​of an indicator in all samples; Constructing the probability matrix : ; in, Standardized value for benefit-based indicators Or standardized value of cost-type indicator ; is the number of samples; And calculate the information entropy (Reflects the value of indicator information, the smaller the value, the more important the information): ; in, For the The information entropy of each indicator; And calculate the objective weight: ; in, For the The objective weight of each indicator; For the The information entropy of an indicator; For all The sum of (1-information entropy) of the indicators; S320.6, calculate the working condition complexity index, through the dynamic fusion factor Fusion of subjective weights and objective weight .

[0017] As a further improvement of this technical solution, the dynamic fusion of subjective and objective weights in S320.6 includes the following steps: S330.1. Calculate the standard deviation for each indicator and coefficient of variation , weighted average to get the working condition complexity: ; in, Where, for Moment The first indicator values, is the mean value of this indicator, The number of samples for calculating the standard deviation; , used to reflect the relative fluctuation of indicator data; is the total number of network-related performance indicators; For the current moment The working condition complexity index; S330.2. Use Sigmoid function to define fusion factor , to achieve a smooth transition between subjective and objective weights: ; in, is the regulating factor; is the complexity benchmark value; S330.3. Subjective weights of integrated improved analytic hierarchy process Objective weights of the entropy weight method : ; in, time No. The final dynamic weight of each indicator.

[0018] As a further improvement of this technical solution, the performance evaluation model unit uses a comprehensive evaluation model that integrates a random forest algorithm to generate multidimensional performance scores and risk warnings, including the following steps: S410, extracting the pre-processed and weight-adjusted network-related indicator data and integrating them into a feature set, covering dimensional indicators such as power control, voltage regulation, and frequency response, as well as dynamic weight information; S420. Divide historical operation data into training sets and validation sets, build multiple decision trees based on the training set, randomly select samples and feature subsets for each tree, and learn the mapping relationship between indicators and performance; and optimize the number of decision trees, maximum depth, number of features per tree, and minimum number of samples for node splitting through the validation set to minimize prediction error and form a stable random forest model. S430: Input real-time or to-be-evaluated data into the trained random forest model, and perform predictions on multiple decision trees in parallel. The outputs of each decision tree are integrated with the dynamic weights of the indicators to calculate a comprehensive performance score for the new energy station network. This score reflects performance in multiple dimensions, including power control, voltage regulation, and frequency response. S440: Pre-set performance score warning thresholds based on historical fault data and abnormal operating condition data; compare the real-time calculated performance score with the warning threshold. If the score is lower than the threshold, trigger a risk warning; and output the warning level (such as general warning, severe warning) and the main related abnormal indicators to provide support for operation and maintenance decisions.

[0019] A second object of the present invention is to provide a multi-dimensional network-related performance index evaluation method based on measured data of new energy stations. The multi-dimensional network-related performance index evaluation system based on measured data of new energy stations comprises the following steps: S100, Multi-source data synchronous acquisition and preprocessing: Collect electrical parameters and environmental data through distributed arrays and meteorological sensor networks, and store and transmit them after spatiotemporal synchronization, noise suppression, and normalization processing; S200, three-dimensional indicator system construction: Based on steady-state and transient operation data, an evaluation system including static electrical indicators, dynamic stability indicators and harmonic pollution indicators is constructed; S300, dynamic weight fusion calculation: Use the Delphi method to build an expert experience database, and combine the operating conditions to generate dynamic indicator weights by integrating the improved hierarchical analysis method and the entropy weight method; S400, Comprehensive Performance Evaluation and Early Warning: Use a model that integrates the random forest algorithm to score indicator data and trigger risk warnings based on thresholds; S500, Visual Interaction and Model Iteration: Display evaluation results through the terminal and provide an interactive interface, and regularly iterate and optimize the evaluation model based on new data.

[0020] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention builds a three-dimensional evaluation system covering static electrical parameters (voltage deviation, frequency fluctuation), dynamic stability indicators (active power ramp rate, voltage sag recovery capability), and harmonic pollution indicators (total harmonic distortion, interharmonic content). This system breaks through the limitations of single-dimensional evaluation in existing technologies and extends the grid-related performance of new energy stations from "local detection" to "full-dimensional coverage." It accurately matches the grid's multi-scenario assessment requirements for station steady-state operation, transient response, and power quality. 2. This invention builds a dynamic weight calculation model based on an improved fusion mechanism of the Analytic Hierarchy Process (AHP) and the Entropy Weight Method, combining an expert experience database (Delphi scoring matrix) with real-time operating data (power fluctuation rate, voltage deviation rate). This addresses the drawback of traditional fixed weights that are unable to adapt to complex operating conditions. The importance of indicators is adjusted in real time based on the station's operating status, ensuring the adaptability of the evaluation logic to dynamic operating conditions and improving the accuracy of network-related performance evaluation in multiple scenarios. 3. This invention introduces a Gaussian mixture model to cluster and analyze real-time operating condition data, and uses a Sigmoid fusion factor to dynamically adjust the subjective and objective weights, establishing an adaptive correction mechanism for the evaluation model. In scenarios such as power grid disturbances and sudden weather changes, it can calibrate assessment deviations in real time, addressing the lag issues of traditional solutions, enhancing the model's responsiveness to complex operating conditions, and improving the timeliness and accuracy of abnormal operating condition warnings. 4. This invention integrates expert experience with measured data from new energy stations to build a collaborative evaluation framework. This framework balances the influence of industry expert knowledge and physical operational data, filling the gaps in existing technologies where subjective and objective integration is insufficient. It also promotes the evolution of network-related performance evaluation from a "single data-driven" approach to a "multi-source information collaboration" approach, providing more comprehensive decision-making support for the stable interaction between new energy stations and the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the system framework of the present invention; Figure 2 Schematic diagram of the method steps of the present invention; The meaning of each number in the figure is: 100, data acquisition preprocessing unit; 110, multi-source heterogeneous data acquisition module; 120, spatiotemporal synchronization preprocessing module; 121, adaptive noise suppression submodule; 122, dynamic dimension normalization submodule; 123, data quality assessment submodule; 200, multi-dimensional index construction unit; 210, static electrical index module; 220, dynamic stability index module; 230, harmonic pollution index module; 231, harmonic separation calculation submodule; 232, source end positioning analysis submodule; 300, weight adjustment unit; 310, expert experience database construction module; 320, weight calculation module; 400, performance evaluation model unit; 500. Visual interactive terminal unit. DETAILED DESCRIPTION

[0022] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention. Example 1

[0023] like Figure 1 As shown, this embodiment provides a multi-dimensional network-related performance index evaluation system based on measured data from new energy stations, including: The data acquisition and preprocessing unit 100 is used to synchronously collect electrical operation data and environmental parameters of the new energy station and perform denoising and normalization preprocessing; In this embodiment, the data acquisition and preprocessing unit 100 includes a multi-source heterogeneous data acquisition module 110 and a spatiotemporal synchronization preprocessing module 120, wherein: The multi-source heterogeneous data acquisition module 110 uses a distributed array consisting of 16 synchronized phasor acquisition devices, combined with a meteorological sensor network, and is compatible with the IEEE C37.118.2 synchronization protocol to achieve sub-millisecond synchronous acquisition of electrical parameters (three-phase voltage amplitude, frequency deviation, active power, etc.) and environmental parameters (light intensity, ambient temperature, wind speed, wind direction angle, air pressure); As a further illustration of this embodiment, the synchronized phasor acquisition equipment in this embodiment is deployed in a distributed manner, with 16 devices forming an array. It supports the IEEE C37.118.2 synchronization protocol and implements synchronized acquisition of electrical parameters (such as three-phase voltage amplitude, frequency deviation, and active power). Furthermore, the meteorological sensor network includes sensors for light intensity, ambient temperature, wind speed, wind direction, and air pressure, and these sensors are triggered by a unified clock with the synchronized phasor acquisition equipment for acquisition. Furthermore, this embodiment uses the time synchronization mechanism based on the IEEE C37.118.2 protocol to achieve sub-millisecond synchronization of multi-source data through PTP (Precision Time Protocol), ensuring the time consistency of electrical parameters and environmental parameters; the acquisition equipment synchronizes the GPS clock in real time, and the uniformity of the sampling time is guaranteed through the hardware trigger mechanism.

[0024] The spatiotemporal synchronization preprocessing module 120 realizes multi-source data timestamp alignment based on the GPS timing system, builds a three-level cache architecture (L1 cache SRAM stores the last 10 seconds of data, L2 cache DRAM stores the last hour of data, and L3 cache SSD array stores historical data), and transmits the preprocessed data to the multi-dimensional indicator construction unit 200 through the OPCUA protocol to realize breakpoint resumption and CRC-32 verification.

[0025] As a further illustration of this embodiment, the GPS timing system is used in this embodiment to add a unified timestamp to all collected data to eliminate clock deviations between different devices. At the same time, the three-level cache architecture in this embodiment includes: The L1 cache uses high-speed SRAM to store the real-time data of the last 10 seconds to ensure high-frequency access needs; The L2 cache uses DRAM to store the latest hour's data and supports short-term historical data calls; The L3 cache uses an SSD array to store long-term historical data, facilitating subsequent analysis and model iteration.

[0026] In this embodiment, the spatiotemporal synchronization preprocessing module 120 includes an adaptive noise suppression submodule 121, a dynamic dimension normalization submodule 122, and a data quality assessment submodule 123, wherein: The adaptive noise suppression submodule 121 adopts a three-stage filtering architecture, which sequentially passes through a hardware low-pass filter, a digital median filter, and an adaptive threshold filter, combined with spectrum analysis and automatic matching filtering strategies to suppress various types of noise interference and improve data reliability in complex electromagnetic environments; As a further illustration of this embodiment, this embodiment adopts a three-stage filtering architecture to suppress noise, specifically including: The hardware low-pass filter filters out high-frequency interference and automatically matches the cutoff frequency in combination with spectrum analysis; A digital median filter removes impulse noise and smoothes the signal using a sliding window; The adaptive threshold filter dynamically adjusts the filtering threshold based on the statistical characteristics of the signal to suppress random noise.

[0027] The dynamic dimension normalization submodule 122 performs dimension unification and dynamic range scaling on the collected data, and identifies and repairs outliers through sliding window statistical characteristics to adapt to the wide range of fluctuation characteristics of the operating conditions of the new energy station; The data quality assessment submodule 123 performs multi-dimensional quantitative scoring on the pre-processed data based on the preset integrity, consistency, and timeliness assessment standards, divides the data into different quality levels, and automatically isolates data that does not meet the quality standards.

[0028] As a further illustration of this embodiment, this embodiment evaluates data quality from three dimensions: completeness, consistency, and timeliness, specifically including: Completeness assessment is based on the number of data points to detect whether there is any missing data; Consistency assessment uses statistical analysis to determine whether data fluctuations are logical; Timeliness assessment determines whether the real-time requirements are met based on the data delay time; This embodiment divides the data into different quality levels according to the evaluation results, automatically isolates unqualified data, and prevents poor-quality data from affecting the evaluation results.

[0029] It should be added that the multi-source heterogeneous data acquisition module 110 synchronously collects electrical and environmental parameters to generate raw data with timestamps; the spatiotemporal synchronization preprocessing module 120 aligns the timestamps of the raw data, stores it in a three-level cache, and sequentially processes it through noise suppression and dimensional normalization; the data quality assessment submodule 123 grades the preprocessed data, and the qualified data is transmitted to the multidimensional indicator construction unit 200 via the OPCUA protocol, and the unqualified data is stored in isolation.

[0030] The multi-dimensional index construction unit 200 is used to construct a three-dimensional evaluation system including static electrical indicators, dynamic stability indicators and harmonic pollution indicators; In this embodiment, the multi-dimensional index construction unit 200 includes a static electrical index module 210 and a dynamic stability index module 220, wherein: The static electrical indicator module 210 constructs a comprehensive voltage quality assessment system based on steady-state operating data. By quantifying voltage deviation, frequency stability, power factor, voltage amplitude fluctuation, and harmonic distortion rate parameters, it generates static evaluation indicators reflecting the basic power quality of the station. The voltage quality assessment integrates voltage amplitude fluctuation and harmonic distortion rate parameters to form a comprehensive quantitative value to represent grid compatibility. As a further illustration of this embodiment, the static electrical indicator module 210 in this embodiment constructs a comprehensive voltage quality evaluation system based on steady-state operation data, and generates static evaluation indicators by quantifying parameters such as voltage deviation, frequency stability, and power factor. Voltage deviation is calculated using the following formula: ,in is the voltage deviation, for The real-time voltage effective value at the moment (unit: V), is the rated voltage (unit: V); the frequency stability is evaluated by the following formula: ,in for The real-time frequency at the moment (unit: Hz), is the rated frequency (usually 50Hz), is the frequency deviation threshold (unit: Hz); the power factor is , determine, where for Active power at the moment (unit: W), is the reactive power at time t (unit: var). , fused to form a comprehensive quantitative value of voltage quality, where 、 is the weight coefficient, which is used to characterize the influence of each parameter on the grid compatibility.

[0031] The dynamic stability index module 220 constructs a dynamic evaluation system based on the transient response characteristics of the new energy station, including voltage sag recovery capability, frequency regulation rate, and active power ramp rate; and quantifies the stability performance of the new energy station under load fluctuations or faults by capturing the parameter change rate and recovery time during the disturbance process.

[0032] As a further illustration of this embodiment, the dynamic stability index module 220 in this embodiment constructs a dynamic evaluation system including voltage sag recovery capability, frequency regulation rate and active power ramp rate based on the transient response characteristics of the new energy station. The voltage sag depth is calculated by: ,in is the temporary drop depth, is the minimum voltage during the sag (unit: V); the recovery time is defined as ,in is the recovery time (unit: ms), is the time when the sag occurs (unit: s). The active power ramp rate is quantified by ,in is the calculation period (unit: s), is the rated active power of the station (unit: W). This design achieves a quantitative evaluation of stability performance by capturing the rate of change and recovery time of parameters during disturbances.

[0033] In this embodiment, the multi-dimensional index construction unit 200 further includes a harmonic pollution index module 230, which includes a harmonic separation calculation submodule 231 and a source end positioning analysis submodule 232, wherein: The harmonic separation calculation submodule 231 uses the fast Fourier transform technology to separate the 2-50 harmonic components based on the electrical parameters collected synchronously by multiple channels, calculates the content rate of each harmonic and the total distortion rate of the harmonic voltage, generates a harmonic content distribution matrix, and outputs the harmonic component parameters to the source end positioning analysis submodule 232; The source end positioning analysis submodule 232 constructs the phase difference matrix of each harmonic , by setting the phase difference threshold, the main harmonic source equipment is located, a harmonic source positioning map is generated, and the positioning result is output to the performance evaluation model unit 400 as a multi-dimensional scoring parameter.

[0034] As a further explanation of this embodiment, this embodiment uses fast Fourier transform technology to separate the 2nd to 50th harmonic components. The calculation of each harmonic content is done by ,in For the Subharmonic content rate, For the Subharmonic voltage amplitude (unit: V), is the fundamental voltage amplitude (unit: V); the total harmonic voltage distortion rate is calculated by ,in The source end location analysis submodule 232 constructs a phase difference matrix ,in For the Branch Road No. The phase difference of the subharmonics, For the Branch Road No. Subharmonic phasors, is the reference point phasor; when Locate the main harmonic source equipment in real time, including Here, the generated harmonic source localization map is directly output to the performance evaluation model unit 400 as a multi-dimensional scoring parameter.

[0035] The weight adjustment unit 300 dynamically adjusts the indicator weights according to the operating conditions based on the expert scoring judgment matrix and consistency test results of the expert experience database and the fusion algorithm of the improved hierarchical analysis method and the entropy weight method; In this embodiment, the weight adjustment unit 300 includes an expert experience database construction module 310 for constructing an expert experience database. The construction of the expert experience database includes the following steps: S310.1. Select experts and select at least five experts in the fields of new energy grid connection and power system analysis; Expert selection criteria must meet: Years of experience ≥ 10 years, with extensive engineering practice experience; Covering core positions such as scheduling, operation and maintenance, and design, ensuring a multi-dimensional professional perspective; In the past three years, he has participated in ≥2 new energy station network performance evaluation projects and is familiar with industry standards and technical specifications.

[0036] S310.2. Use the Delphi method to conduct multiple rounds of indicator importance ratings: First round: Scale questionnaires were distributed to experts using a 1-9 scale (1 represents an indicator and Equally important, 3 means Slightly more important than , 5 means Obviously more important than , 7 means Strongly important to , 9 means Extremely important , the reciprocal is the reverse importance), collect the original judgment matrix set ,in, , indicating the Experts on indicators and The importance scoring matrix, is the total number of experts, ,in, is the number of indicators; Second round: Summarize all expert scores from the first round and calculate the mean of each matrix element and standard deviation , feedback to experts and invite revision of scores; define coefficient of variation , when all elements When the score converges, the iteration is stopped; S310.3. Perform consistency check on each expert matrix: Calculate the maximum eigenvalue , through the formula Get consistency index ; Check the standard random consistency index table to obtain the corresponding ; Calculating the consistency ratio ,filter <0.1 matrix, constitutes the candidate set of the experience library ; S310.4. Repeat S310.1, S310.2, and S310.3 every 4-8 months based on new expert opinions or new energy technology iteration data (such as new equipment access, control strategy upgrades), and iteratively optimize the expert experience database to ensure the timeliness and accuracy of expert experience.

[0037] As a further explanation of this embodiment, this embodiment builds a reliable experience database through multiple rounds of expert scoring. The specific steps are as follows: First, select experts and select at least 5 experts in the fields of new energy grid connection and power system analysis. They must have been in the industry for ≥10 years, covering core positions such as dispatching, operation and maintenance, and design, and have participated in ≥2 new energy station grid-related performance evaluation projects in the past 3 years. The Delphi method is used to perform multiple rounds of indicator importance scoring: in the first round, experts score the indicator importance using a 1-9 scale system to generate an original judgment matrix set. After the second round of summary scoring, the mean and standard deviation of each matrix element are calculated, and the convergence is judged by the coefficient of variation. When all When the value is less than 0.15, the iteration stops, otherwise it returns to the expert for correction. A consistency test is performed on each expert matrix, and the consistency index is obtained by calculating the maximum eigenvalue. , combined with the standard random consistency index Calculate ratios ,filter Matrices with a value < 0.1 constitute the candidate set of the experience database. To ensure timeliness, the expert experience database is iteratively optimized every 4-8 months based on new expert opinions or technical iteration data.

[0038] Furthermore, the iteration cycle of the expert experience database in this embodiment must match the speed of new energy technology development and the frequency of changes in site operating conditions. When technology iterations are rapid (e.g., the introduction of new inverters) or site operating conditions fluctuate frequently, the iteration cycle can be shortened to four months to update expert experience. When technology is relatively stable, the iteration cycle can be extended to eight months to reduce iteration costs. This interval covers the typical technology update cycle in the industry (refer to the recommended standard revision cycle in the "Technical Specifications for Network-Related Performance Evaluation of New Energy Sites").

[0039] In this embodiment, the weight adjustment unit 300 further includes a weight calculation module 320. The weight calculation module 320 calculates the dynamic weight based on the expert experience database, including the following steps: S320.1. Call the candidate matrix set that has passed the consistency test from the expert experience database , according to the experts' years of experience Calculate weights: ; in, is the number of effective experts, satisfy , reflecting the weight of opinions of senior experts; Perform weighted averaging on the candidate matrices to generate the initial judgment matrix : ; in, Represents the initial judgment matrix Middle Rank The elements of the column are; is the number of candidate matrices participating in weighted averaging; For the The weights corresponding to the candidate matrices; For the The candidate matrix Rank Elements of the column; are the row and column indices of the matrix; The matrix integrates the knowledge of multiple experts and balances the professional perspectives of experts in different positions through weight distribution.

[0040] S320.2. Collect the operating data of new energy stations in real time and calculate the active power fluctuation rate and voltage deviation rate, including: Active power fluctuation rate ;in, for Active power at all times, is the active power at the previous moment, is the rated active power of the station; Voltage deviation rate ;in, for The effective value of voltage at the moment, is the rated voltage; S320.3, through the improved Gaussian mixture model Correct the initial matrix: ; in, is the correction coefficient (trained by historical data) to control the correction amplitude; Furthermore, the Gaussian mixture model in the improved hierarchical analysis method is The present invention adopts the optimization design of adaptive bandwidth kernel function and working condition classification correction strategy, which is as follows: To make the model more responsive to power fluctuations, The kernel function introduces a dynamic bandwidth parameter: is the base bandwidth, [0.1,0.3], represents the adjustment coefficient; According to the active power fluctuation rate of the new energy station and voltage deviation rate , the operating conditions are divided into the following three categories, and Set differentiated output ranges to avoid weight logic confusion caused by indifferent corrections: Stable working condition ( ): [-0.1,0.1]; Fluctuating working conditions (5 < 15 or 2 < 5 ): [-0.2,0.2]; Severe fluctuation conditions ( >15 or >5 ): [-0.3,0.3].

[0041] S320.4. Calculate the dynamic consistency ratio for the correction matrix: ; ; in, is the dynamic consistency ratio, which is used to measure the consistency of the correction matrix. is the consistency indicator, is the standard random consistency indicator; is the maximum eigenvalue of the matrix (obtained from the standard table), is the order of the correction matrix; when At 0.1, the second correction is triggered: ; in, is the correction intensity factor; is the correction matrix at time , No. Rank The elements of the column are the objects of the secondary correction operation; After the second correction, the matrix is , No. Rank New elements of the column; S320.5, indicator data matrix Perform extreme value normalization: Efficiency indicators (the larger the value, the better, such as power factor): ; Cost-based indicators (the smaller the value, the better, such as harmonic distortion rate): ; in, For the Samples, The standardized value of each indicator; For the Samples, The original measurement value of each indicator; and Respectively The minimum and maximum values ​​of an indicator in all samples; Constructing the probability matrix : ; in, Standardized value for benefit-based indicators Or standardized value of cost-type indicator ; is the number of samples; And calculate the information entropy (Reflects the value of indicator information, the smaller the value, the more important the information): ; in, For the The information entropy of an indicator; And calculate the objective weight: ; in, For the The objective weight of each indicator; For the The information entropy of an indicator; For all The sum of (1-information entropy) of the indicators; S320.6, calculate the working condition complexity index, through the dynamic fusion factor Fusion of subjective weights and objective weight .

[0042] In this embodiment, the dynamic fusion of subjective and objective weights in S320.6 includes the following steps: S330.1. Calculate the standard deviation for each indicator and coefficient of variation , weighted average to get the working condition complexity: ; in, Where, for Moment The first sample indicator values, is the mean value of this indicator, The number of samples for calculating the standard deviation; , used to reflect the relative fluctuation of indicator data; is the total number of network-related performance indicators; For the current moment The working condition complexity index; S330.2. Use Sigmoid function to define fusion factor , to achieve a smooth transition between subjective and objective weights: ; in, is the regulating factor; is the complexity benchmark value; S330.3. Subjective weights of integrated improved analytic hierarchy process Objective weights of the entropy weight method : ; in, time No. The final dynamic weight of each indicator.

[0043] As a further explanation of this embodiment, the dynamic weight generation process in this embodiment specifically includes: first, calling the candidate matrix that has passed the consistency test from the experience database, and Calculating weights , weighted to generate the initial judgment matrix Collect new energy station operation data in real time and calculate active power fluctuation rate and voltage deviation rate ; Then, based on the operating parameters, the improved Gaussian mixture model Modify the initial matrix. The model is divided into three types of working conditions: stable, fluctuating and violently fluctuating according to the active power fluctuation rate and voltage deviation rate, and corresponds to different output ranges. The modified matrix needs to be tested for dynamic consistency. When the consistency ratio When the probability matrix is ​​>0.1, a secondary correction is triggered to ensure the reliability of the matrix. Finally, when the entropy weight method is used to calculate the objective weight, the indicator data is first normalized to the extreme value, and different normalization formulas are used for benefit-type indicators and cost-type indicators. After constructing the probability matrix, the information entropy is calculated. , and then get the objective weight By calculating the working condition complexity index , use the Sigmoid function to define the fusion factor , the subjective weight obtained by improving the hierarchical analysis method and objective weight Dynamic fusion to generate final weights .

[0044] Furthermore, the expert experience library construction module 310 provides a reliable initial judgment matrix for weight calculation through multiple rounds of expert scoring and consistency verification. The weight calculation module 320 combines real-time operating data, dynamically corrects the matrix, and integrates subjective and objective weights to generate a set of indicator weights adapted to different operating conditions. The weight adjustment unit 300, through a full-process design of "expert knowledge initialization - dynamic correction of operating condition data - fusion of subjective and objective weights," achieves adaptive weight adjustment, providing dynamic and precise weight support for the multi-dimensional evaluation of the grid-related performance of new energy stations.

[0045] The performance evaluation model unit 400 uses a comprehensive evaluation model that integrates a random forest algorithm to generate multi-dimensional performance scores and risk warnings; In this embodiment, the performance evaluation model unit 400 uses a comprehensive evaluation model that integrates a random forest algorithm to generate multi-dimensional performance scores and risk warnings, including the following steps: S410, extracting the pre-processed and weight-adjusted network-related indicator data and integrating them into a feature set, covering dimensional indicators such as power control, voltage regulation, and frequency response, as well as dynamic weight information; S420: Divide the historical operation data into a training set and a validation set, construct multiple decision trees based on the training set, randomly select samples and feature subsets for each tree, and learn the mapping relationship between indicators and performance; and optimize the number of decision trees, maximum depth, number of features per tree, and minimum number of samples for node splitting using the validation set to minimize prediction error and form a stable random forest model. S430: Input real-time or to-be-evaluated data into the trained random forest model, and perform predictions on multiple decision trees in parallel. The outputs of each decision tree are integrated with the dynamic weights of the indicators to calculate a comprehensive performance score for the new energy station network. This score reflects performance in multiple dimensions, including power control, voltage regulation, and frequency response. S440: Pre-set performance score warning thresholds based on historical fault data and abnormal operating condition data; compare the real-time calculated performance score with the warning threshold. If the score is lower than the threshold, trigger a risk warning; and output the warning level (such as general warning, severe warning) and the main related abnormal indicators to provide support for operation and maintenance decisions.

[0046] As a further illustration of this embodiment, the performance evaluation model unit 400 in this embodiment adopts a comprehensive evaluation model that integrates the random forest algorithm to achieve multi-dimensional scoring and risk warning of the network-related performance of new energy stations. The performance evaluation model unit 400 first extracts the network-related index data after preprocessing and weight adjustment, and integrates it into a feature set covering dimensions such as power control, voltage regulation, and frequency response. Each feature is associated with a dynamic weight to reflect the importance of the indicator under different working conditions. Specifically, the construction process of the comprehensive evaluation model in this embodiment is as follows: When building the comprehensive evaluation model, the historical operation data is divided into a training set and a validation set in a ratio of 7:3 (5-fold cross validation is used when the data volume is less than 10,000). 60% of the samples are randomly selected from the training set using the bootstrap sampling method, and N decision trees are repeatedly generated (N is 50-100). Each tree is randomly selected. features (d is the total number of features), and the node splitting threshold is determined based on the Gini impurity criterion, where the Gini impurity calculation formula is: , where: For nodes The number of samples, For nodes Belong to the The number of samples of the class, is the number of performance level categories; the number and depth of decision trees are optimized through the validation set to ensure that the model prediction error is minimized.

[0047] When real-time data is input into the trained model, N decision trees output a single score in parallel. , and generate a comprehensive score by weighted average: , where: For the The weight of each tree (equal weight by default, which can be optimized to adaptive weight based on the validation set error), For the Tree pairs sample The score ranges from 0 to 100. The score corresponds to the performance level, which includes: excellent (≥90 points), good (80-89 points), fair (70-79 points), and poor (<70 points).

[0048] The risk warning mechanism sets thresholds based on the quantiles of historical failure data: ,in, is the score set at the time of failure, Take 0.2 (general warning, score < 70 points) or 0.1 (serious warning, score < 60 points). When the score is lower than the threshold, the top three abnormal indicators are located by feature importance sorting (based on the decrease in Gini impurity). The feature importance calculation formula is: Where: Characterized by is the number of samples of node t, is the total number of samples, and are the left and right child nodes of node t.

[0049] The visual interaction terminal unit 500 is used to display the evaluation results in real time, generate trend analysis reports, and provide a human-computer interaction interface.

[0050] As a further illustration of this embodiment, the visual interactive terminal unit 500 employs an industrial-grade all-in-one touchscreen (e.g., equipped with a 15-inch or larger display panel, suitable for outdoor strong light environments). It establishes a real-time connection with the performance evaluation model unit 400 via the WebSocket protocol, synchronizing comprehensive scores, warning events, and dynamic weighting data within seconds. The backend relies on a server cluster to store historical evaluation results, supporting at least one year of operational data backlog. The terminal is linked to the data acquisition and preprocessing unit 100 via a data interface, enabling bidirectional traceability between evaluation data and original sample values ​​(e.g., clicking on a report indicator redirects to the original acquisition record).

[0051] As a further illustration of this embodiment, the terminal interface integrates functions according to the "result display - trend analysis - early warning and response" logic. A circular chart intuitively displays real-time comprehensive scores and performance levels (with differentiated color schemes for excellent, good, fair, and poor). Clicking the circular chart expands the "score-operating condition correlation view" (e.g., comparing score distributions under different operating conditions). A timeline-driven line chart displays the dynamic changes of key network-related indicators such as voltage deviation and power fluctuation rate. When an indicator exceeds a preset threshold, the corresponding data point is clearly marked. Early warning events are presented in a list format, with each record including the warning time, level (general / critical), and associated abnormal indicators. Clicking an event displays historical information such as the response plan and recovery time for similar faults. The terminal also supports automated periodic report generation: daily reports automatically summarize the daily comprehensive score fluctuation range, the number of indicator violations, and early warning event statistics, embedded in the indicator trend chart to form a daily report. Weekly reports are generated, adding analysis dimensions such as the weekly score change rate and the operating condition proportion (distribution of stable / fluctuating operating conditions). Reports can be exported as PDFs and automatically sent to the operation and maintenance team via email. All data in the report can be traced back to the original sampling records of the data acquisition and preprocessing unit 100 through the terminal interface, meeting the needs of data auditing and fault reproduction.

[0052] As a further explanation of this embodiment, in the human-computer interaction level of this embodiment, the terminal provides an interactive operation interface: that is, the warning threshold is modified through the slider control (the adjustment instruction is synchronized to the performance evaluation model unit 400 in real time, and the threshold takes effect immediately after the update); the network-related indicators of concern (static, dynamic or harmonic indicators) are selected through the drop-down menu, and the indicator trend area is refreshed with the corresponding curve synchronously; the evaluation data and indicator performance of any time period are queried through the time selector, and the comparative analysis of historical results and real-time data is supported; at the same time, three levels of authority are set for operation and maintenance, engineers, and administrators, and differentiated open data viewing, threshold modification, report export and other functions are ensured to ensure data security.

[0053] For example, take the photovoltaic station inspection scenario as an example: the operation and maintenance personnel learned from the terminal ring chart that the comprehensive score of the station is 88 points (good), and the voltage deviation peak mark in the indicator trend area triggered the warning details. Combined with the historical records of similar events (including handling measures and recovery time), it helps to determine whether the current fluctuation is abnormal; when the weekly report shows that the "power factor qualification rate" has dropped, the harmonic indicator abnormality is located through the indicator screening function, triggering the filter maintenance process.

[0054] It can be understood that the visual interactive terminal unit 500 in this embodiment not only connects to the real-time evaluation results of the performance evaluation model unit 400 to achieve visual presentation of scores and warnings; it also associates the original data of the data acquisition and preprocessing unit 100 to support report traceability and historical comparison, thereby providing an intuitive tool for the dynamic management of the network-related performance of new energy stations.

[0055] It should be added that the visual interactive terminal unit 500 is also used to trigger the model iteration process to achieve dynamic optimization of the evaluation system. Specifically, the system supports dual mechanisms of periodic triggering and event triggering: the periodic trigger is set to automatically start the full iteration in the early morning of the first non-working day of each week; the event trigger includes the fluctuation of the comprehensive score exceeding the preset threshold for many consecutive days, the detection of new equipment types in the new energy site (such as the connection of the energy storage system), or the operation and maintenance personnel manually initiating the iteration command through the terminal. During the iteration process, the system automatically retrieves the historical operating data of the data acquisition and preprocessing unit 100 (including the original sampling values ​​of electrical parameters and environmental parameters), the historical scoring records of the performance evaluation model unit 400, and the abnormal events and disposal feedback marked by the operation and maintenance personnel in the visual interactive terminal unit 500. At the same time, it synchronizes the latest standards of the power grid industry and the public fault case library through the external data interface to ensure the comprehensiveness of the iterative data; At the same time, in terms of core optimization strategies, based on the distribution of historical operating condition data, the subjective and objective weights of each evaluation indicator are recalculated through an improved fusion algorithm of the hierarchical analysis method and the entropy weight method. For example, when voltage sag events occur frequently during a certain period, the system automatically increases the weight of the "dynamic stability index" to enhance the evaluation accuracy of the corresponding dimension. Secondly, the input feature dimensions of the evaluation model are automatically expanded for newly added equipment types. For example, when a photovoltaic station adds an SVG, the model simultaneously incorporates reactive power compensation parameters as a new evaluation indicator and updates the decision tree structure of the random forest model through an incremental learning algorithm. In addition, an online learning algorithm is used to process real-time data streams, gradually optimizing model parameters without interrupting real-time evaluation. Incremental learning results are merged with historical data during the weekly full iteration to complete the overall model update. In addition, the iterated model is first run in a test environment for 48 hours. Time series cross-validation is used to compare the scoring consistency and warning accuracy of the new and old models. Once verified, the model is switched to the production environment using a blue-green deployment strategy, while retaining the first three valid versions for rollback. The terminal displays model iteration logs and performance comparison reports in real time. Operations and maintenance personnel can review iteration results and submit optimization suggestions through a permissioned interface, forming a closed-loop management system of "data collection-evaluation-iteration." Example 2

[0056] See also Figure 2 This embodiment further provides a multi-dimensional network-related performance index evaluation method based on measured data of new energy stations. The multi-dimensional network-related performance index evaluation system based on measured data of new energy stations according to embodiment 1 includes the following steps: S100, Multi-source data synchronous acquisition and preprocessing: Collect electrical parameters and environmental data through distributed arrays and meteorological sensor networks, and store and transmit them after spatiotemporal synchronization, noise suppression, and normalization processing; S200, three-dimensional indicator system construction: Based on steady-state and transient operation data, an evaluation system including static electrical indicators, dynamic stability indicators and harmonic pollution indicators is constructed; S300, dynamic weight fusion calculation: Use the Delphi method to build an expert experience database, and combine the operating conditions to generate dynamic indicator weights by integrating the improved hierarchical analysis method and the entropy weight method; S400, Comprehensive Performance Evaluation and Early Warning: Use a model that integrates the random forest algorithm to score indicator data and trigger risk warnings based on thresholds; S500, Visual Interaction and Model Iteration: Display evaluation results through the terminal and provide an interactive interface, and regularly iterate and optimize the evaluation model based on new data.

[0057] Those skilled in the art will appreciate that the process of implementing all or part of the steps of the above embodiments may be accomplished by hardware, or by instructing related hardware through a program.

[0058] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-dimensional network-related performance index evaluation system based on measured data from new energy stations, characterized by: include: A data acquisition preprocessing unit (100), the data acquisition preprocessing unit (100) is used to synchronously acquire electrical operation data and environmental parameters of a new energy station, and perform denoising and normalization preprocessing; A multi-dimensional index construction unit (200), the multi-dimensional index construction unit (200) being used to construct a three-dimensional evaluation system including static electrical indicators, dynamic stability indicators and harmonic pollution indicators; A weight adjustment unit (300), wherein the weight adjustment unit (300) dynamically adjusts the indicator weight according to the operating conditions based on the expert scoring judgment matrix and consistency test results of the expert experience database and a fusion algorithm of the improved hierarchical analysis method and the entropy weight method; A performance evaluation model unit (400), wherein the performance evaluation model unit (400) uses a comprehensive evaluation model integrated with a random forest algorithm to generate a multi-dimensional performance score and risk warning; A visual interaction terminal unit (500) is used to display evaluation results in real time, generate trend analysis reports, and provide a human-computer interaction interface.

2. The multi-dimensional network-related performance index evaluation system based on measured data of new energy stations according to claim 1 is characterized in that: The data acquisition and preprocessing unit (100) comprises a multi-source heterogeneous data acquisition module (110) and a spatiotemporal synchronization preprocessing module (120), wherein: The multi-source heterogeneous data acquisition module (110) uses a distributed array composed of 16 synchronized phasor acquisition devices, combined with a meteorological sensor network, to achieve sub-millisecond synchronous acquisition of electrical parameters and environmental parameters; The spatiotemporal synchronization preprocessing module (120) realizes multi-source data timestamp alignment based on the GPS timing system, constructs a three-level cache architecture, and transmits the preprocessed data to the multi-dimensional indicator construction unit (200) via the OPCUA protocol, realizing breakpoint resume and CRC-32 check.

3. The multi-dimensional network-related performance index evaluation system based on measured data of new energy stations according to claim 2 is characterized in that: The spatiotemporal synchronization preprocessing module (120) includes an adaptive noise suppression submodule (121), a dynamic dimension normalization submodule (122) and a data quality assessment submodule (123), wherein: The adaptive noise suppression submodule (121) adopts a three-stage filtering architecture, sequentially passing through a hardware low-pass filter, a digital median filter, and an adaptive threshold filter, combined with a spectrum analysis automatic matching filtering strategy to suppress various types of noise interference; The dynamic dimension normalization submodule (122) performs dimension normalization and dynamic range scaling on the collected data, and identifies and repairs outliers through sliding window statistical characteristics; The data quality assessment submodule (123) performs multi-dimensional quantitative scoring on the pre-processed data based on preset integrity, consistency, and timeliness assessment standards, divides the data into different quality levels, and automatically isolates data that does not meet the quality standards.

4. The multi-dimensional network-related performance index evaluation system based on measured data of new energy stations according to claim 1 is characterized in that: The multi-dimensional index construction unit (200) comprises a static electrical index module (210) and a dynamic stability index module (220), wherein: The static electrical index module (210) constructs a voltage quality comprehensive evaluation system based on steady-state operation data, and generates static evaluation indicators reflecting the basic power quality of the station by quantifying voltage deviation, frequency stability, power factor, voltage amplitude fluctuation and harmonic distortion rate parameters; wherein the voltage quality evaluation integrates the voltage amplitude fluctuation and harmonic distortion rate parameters to form a comprehensive quantitative value to characterize the compatibility of the power grid; The dynamic stability index module (220) constructs a dynamic evaluation system including voltage sag recovery capability, frequency regulation rate and active power ramp rate according to the transient response characteristics of the new energy station; and quantifies the stability performance of the new energy station under load fluctuation or fault conditions by capturing the parameter change rate and recovery time during the disturbance process.

5. The multi-dimensional network-related performance index evaluation system based on measured data of new energy stations according to claim 4 is characterized in that: The multi-dimensional index construction unit (200) further includes a harmonic pollution index module (230), wherein the harmonic pollution index module (230) includes a harmonic separation calculation submodule (231) and a source end positioning analysis submodule (232); wherein: The harmonic separation calculation submodule (231) uses fast Fourier transform technology to separate 2-50 harmonic components based on the electrical parameters collected synchronously by multiple channels, calculates the content rate of each harmonic and the total distortion rate of harmonic voltage, generates a harmonic content distribution matrix, and outputs the harmonic component parameters to the source end positioning analysis submodule (232); The source end positioning analysis submodule (232) constructs a phase difference matrix of each harmonic , by setting the phase difference threshold, the main harmonic source equipment is located, a harmonic source location map is generated, and the location result is output to the performance evaluation model unit (400) as a multi-dimensional scoring parameter.

6. The multi-dimensional network-related performance index evaluation system based on measured data of new energy stations according to claim 1 is characterized in that: The weight adjustment unit (300) includes an expert experience library construction module (310) for constructing an expert experience library. The construction of the expert experience library includes the following steps: S310.

1. Select experts and select at least five experts in the fields of new energy grid connection and power system analysis; S310.

2. Use the Delphi method to conduct multiple rounds of indicator importance ratings: First round: Issue scale questionnaires to experts, using a 1-9 scale system, and collect the original judgment matrix set ,in, , indicating the Experts on indicators and The importance scoring matrix, is the total number of experts, ,in, is the number of indicators; Second round: Summarize all expert scores from the first round and calculate the mean of each matrix element and standard deviation , feedback to experts and invite revision of scores; define coefficient of variation , when all elements When the score converges, the iteration is stopped; S310.

3. Perform consistency check on each expert matrix: Calculate the maximum eigenvalue , through the formula Get consistency index ; Check the standard random consistency index table to obtain the corresponding standard random consistency index ; Calculating the consistency ratio ,filter <0.1 matrix, constitutes the candidate set of the experience library ; S310.

4. Repeat S310.1, S310.2, and S310.3 every 4-8 months in combination with new expert opinions or new energy technology iteration data to iteratively optimize the expert experience database.

7. The multi-dimensional network-related performance index evaluation system based on measured data of new energy stations according to claim 6 is characterized in that: The weight adjustment unit (300) further includes a weight calculation module (320), wherein the weight calculation module (320) calculates the dynamic weight based on the expert experience database, including the following steps: S320.

1. Call the candidate matrix set that has passed the consistency test from the expert experience database , according to the experts' years of experience Calculating weights : Perform weighted averaging on the candidate matrices to generate the initial judgment matrix : S320.

2. Collect the operating data of new energy stations in real time and calculate the active power fluctuation rate and voltage deviation rate, including: Active power fluctuation rate ;in, for Active power at all times, is the active power at the previous moment, is the rated active power of the station; Voltage deviation rate ;in, for The effective value of voltage at the moment, is the rated voltage; S320.3, through the improved Gaussian mixture model Correct the initial matrix: ； in, is the correction coefficient, which controls the correction amplitude; S320.

4. Calculate the dynamic consistency ratio for the correction matrix: ； ； in, is the dynamic consistency ratio, which is used to measure the consistency of the correction matrix. is the consistency indicator, is the standard random consistency indicator; is the maximum eigenvalue of the matrix, is the order of the correction matrix; when When >0.1, a secondary correction is triggered: ； in, is the correction intensity factor; is the correction matrix at time , No. Rank The elements of the column are the objects of the secondary correction operation; After the second correction, the matrix is , No. Rank New elements of the column; S320.5, indicator data matrix Perform extreme value normalization: Benefit indicators: ； Cost indicators: ； in, All are Samples, The standardized value of each indicator; For the Samples, The original measurement value of each indicator; and Respectively The minimum and maximum values ​​of an indicator in all samples; Constructing the probability matrix , calculate information entropy : and calculate the objective weight : S320.6, calculate the working condition complexity index, through the dynamic fusion factor Fusion of subjective weights and objective weight .

8. The multi-dimensional network-related performance index evaluation system based on measured data of new energy stations according to claim 7 is characterized in that: The dynamic fusion of subjective and objective weights in S320.6 includes the following steps: S330.

1. Calculate the standard deviation for each indicator and coefficient of variation , weighted average to get the working condition complexity: ； in, Where, for Moment The first sample indicator values, is the mean value of this indicator, The number of samples for calculating the standard deviation; , used to reflect the relative fluctuation of indicator data; is the total number of network-related performance indicators; For the current moment The working condition complexity index; S330.

2. Use Sigmoid function to define fusion factor , to achieve a smooth transition between subjective and objective weights: ； in, is the regulating factor; is the complexity benchmark value; S330.

3. Subjective weights of integrated improved analytic hierarchy process Objective weights of the entropy weight method : ； in, time No. The final dynamic weight of each indicator.

9. The multi-dimensional network-related performance index evaluation system based on measured data of new energy stations according to claim 1 is characterized in that: The performance evaluation model unit (400) uses a comprehensive evaluation model integrated with a random forest algorithm to generate multi-dimensional performance scores and risk warnings, including the following steps: S410, extracting the pre-processed and weight-adjusted network-related indicator data and integrating them into a feature set, covering dimensional indicators such as power control, voltage regulation, and frequency response, as well as dynamic weight information; S420: Divide the historical operation data into a training set and a validation set, construct multiple decision trees based on the training set, randomly select samples and feature subsets for each tree, and learn the mapping relationship between indicators and performance; and optimize the number of decision trees, maximum depth, number of features per tree, and minimum number of samples for node splitting using the validation set to minimize prediction error and form a stable random forest model. S430: Input the real-time or to-be-evaluated data into the trained random forest model, and perform predictions on multiple decision trees in parallel; integrate the output results of each decision tree, combine the dynamic weights of the indicators, and calculate a comprehensive score for the grid-related performance of the new energy station; S440. Pre-set performance score warning thresholds based on historical fault data and abnormal operating condition data; Compare the real-time calculated performance score with the warning threshold. If the score is lower than the threshold, a risk warning is triggered. The warning level and major related abnormal indicators are output to support operation and maintenance decisions.

10. A multi-dimensional network-related performance index evaluation method based on measured data of new energy stations, based on a multi-dimensional network-related performance index evaluation system based on measured data of new energy stations according to any one of claims 1 to 9, characterized in that: The steps include: S100, Multi-source data synchronous acquisition and preprocessing: Collect electrical parameters and environmental data through distributed arrays and meteorological sensor networks, and store and transmit them after spatiotemporal synchronization, noise suppression, and normalization processing; S200, three-dimensional indicator system construction: Based on steady-state and transient operation data, an evaluation system including static electrical indicators, dynamic stability indicators and harmonic pollution indicators is constructed; S300, dynamic weight fusion calculation: Use the Delphi method to build an expert experience database, and combine the operating conditions to generate dynamic indicator weights by integrating the improved hierarchical analysis method and the entropy weight method; S400, Comprehensive Performance Evaluation and Early Warning: Use a model that integrates the random forest algorithm to score indicator data and trigger risk warnings based on thresholds; S500, Visual Interaction and Model Iteration: Display evaluation results through the terminal and provide an interactive interface, and regularly iterate and optimize the evaluation model based on new data.

Images