A millisecond-level assessment system, method, equipment, and storage medium for new energy grid-connected operation oriented towards power grid dispatch.

Abstract

This invention relates to the field of new energy grid-connected operation assessment technology, and in particular to a millisecond-level assessment system, method, equipment, and storage medium for new energy grid-connected operation oriented towards grid dispatch. The system includes an assessment instrument deployed at the grid connection point and an assessment system deployed at the dispatch center. The assessment instrument internally incorporates a recursive variance adaptive event triggering discrimination module. It employs a recursive algorithm with a forgetting factor to update the local benchmark mean and noise floor variance of frequency and voltage online. Based on the noise floor variance, it dynamically generates an adaptive envelope threshold and uses bidirectional accumulation and statistics to accumulate time-domain deviations to determine event triggering. Simultaneously, it determines the theoretical compensation energy using the squared difference of frequency endpoints and the actual compensation energy using the time-domain integral of power deviation, outputting the assessment score through the energy ratio. This invention solves the problems of high false triggering and missed detection rates in fixed threshold triggering mechanisms under strong noise environments and the numerical instability of the instantaneous differential assessment model.

Description

Technical Field

[0001] This invention relates to the field of new energy grid-connected operation assessment technology, and in particular to a millisecond-level assessment system, method, equipment and storage medium for new energy grid-connected operation oriented towards grid dispatch. Background Technology

[0002] As the penetration rate of renewable energy sources such as wind and solar power in the power system continues to rise, a large number of new energy power generation devices connected to the grid via power electronic inverters are gradually replacing traditional synchronous generator units. This has led to a significant decrease in the overall equivalent inertia level of the power system. Consequently, the system faces increased frequency variation rate and a deeper dip in the frequency minimum when encountering active power disturbances, posing a severe challenge to the safety and stability of the power grid operation. To address this, power grid dispatching agencies are gradually incorporating the frequency support capability and voltage regulation capability of new energy power plants into their grid-connected operation assessment system. This requires power plants to be equipped with functions such as virtual inertia control, fast primary frequency regulation, and fast voltage regulation, and to accurately quantify their actual response performance in millisecond-level transient processes.

[0003] Existing technologies have proposed deploying dedicated millisecond-level hardware testing instruments at the grid connection points of new energy power plants to achieve on-site acquisition and processing of high-frequency data. This addresses the issues of insufficient time resolution and data source objectivity in traditional second-level remote control systems at the dispatch master station in terms of physical architecture. However, the core software algorithms within these hardware testing instruments still suffer from two interrelated technical defects. First, in the event triggering judgment stage, existing solutions generally employ triggering logic based on fixed thresholds, comparing the rate of change of frequency or voltage with a preset static constant threshold. However, new energy power plant grid connection points are located in a strong noise environment superimposed with high-frequency switching harmonics from inverters and complex electromagnetic interference. After numerical differentiation, the noise amplitude of the high-frequency sampled data is drastically amplified. A fixed low threshold will generate a massive number of invalid triggers due to frequent background noise exceeding the threshold, while a fixed high threshold will cause the system to lose its sensitivity to real, small, slowly changing disturbances, resulting in the missed detection of effective auxiliary service events. Secondly, in the quantification of assessment indicators, existing solutions rely on extracting the instantaneous extreme frequency change rate and substituting it into differential equations to calculate the theoretical support power requirement. Since the frequency derivative approaches zero or undergoes a zero-crossing abrupt change near the extreme point during the transient process, using it as a denominator or multiplier in calculations easily leads to numerical oscillations or even singularity divergence. Furthermore, measuring station support performance solely based on instantaneous power extremes ignores the cumulative effect of energy injection over time, making the assessment results susceptible to measurement noise interference and lacking objectivity at the physical energy level. The high false alarm and false positive rate of the aforementioned event triggering mechanism, coupled with the numerical instability of the assessment indicator calculations, creates a negative synergistic effect, severely restricting the reliability of millisecond-level assessment systems in practical engineering applications. Summary of the Invention

[0004] To address the technical problems of existing millisecond-level assessment systems for new energy grid-connected operation, which generate numerous false triggers or missed judgments in strong noise and broadband harmonic environments due to the use of fixed threshold logic in the event trigger judgment stage, and the numerical oscillations and singularity divergence caused by the reliance on the instantaneous frequency change rate differential model in the assessment index quantification stage, thus failing to objectively reflect the energy support contribution of power stations, this invention provides a millisecond-level assessment system, method, equipment, and storage medium for new energy grid-connected operation oriented towards power grid dispatch.

[0005] A millisecond-level assessment system for renewable energy grid-connected operation oriented towards power grid dispatch includes a renewable energy grid-connected operation millisecond-level assessment instrument deployed at the renewable energy power plant grid connection point and a dispatch-side renewable energy assessment system deployed in the power grid dispatch data center; the renewable energy grid-connected operation millisecond-level assessment instrument includes:

[0006] The wideband power quality synchronous sampling and preprocessing module is used to synchronously sample the three-phase AC voltage and current at the grid connection point at a sampling frequency of not less than 10kHz, and to calculate and output the fundamental voltage amplitude in real time through a sliding data window. Instantaneous active power Instantaneous reactive power and instantaneous frequency estimate ;

[0007] A high-precision time synchronization and tag alignment module is used to add microsecond-level absolute timestamps to each discrete parameter output by the wideband power quality synchronization sampling and preprocessing module;

[0008] The recursive variance adaptive event-triggered discrimination module is used to employ a method incorporating a forgetting factor. The recursive algorithm for frequency sequences and voltage amplitude sequence The local baseline mean and local noise floor variance are updated online respectively. An adaptive envelope threshold boundary and tolerance drift amount are dynamically generated based on the local noise floor variance. A bidirectional cumulative sum test statistic is constructed to accumulate persistent deviations exceeding the tolerance drift amount in the time domain. When the cumulative sum test statistic exceeds the preset limit trigger threshold, the event trigger is determined to be established and the event time domain slice window is locked. This allows the trigger determination line to dynamically scale with the real-time background noise level of the grid connection point, effectively shielding invalid triggers caused by high-frequency harmonic ripple and measurement glitch. At the same time, it captures small, slowly changing disturbances through the cumulative effect.

[0009] The integral model time-domain quantization assessment module is used to determine the theoretical expected response compensation energy based on the squared difference of frequency endpoints for the data within the time-domain slice window of the event. The actual output compensation energy is determined by the time-domain integral of the active power deviation. The ratio of the two outputs the comprehensive score of the inertial response effectiveness, and the integral deviation of the reactive power deviation in the time domain and the voltage recovery trajectory from the standard corridor outputs the fast voltage regulation assessment results. The assessment dimension is transformed from instantaneous power space to energy space to eliminate the numerical singularity problem of differential equations.

[0010] The multi-dimensional situational data and assessment result compression communication module is used to package and transmit the assessment indicator scoring dictionary and the downsampled event window key feature sequence to the scheduling-side new energy assessment system.

[0011] As a preferred embodiment of the present invention, the recursive variance adaptive event-triggered discrimination module updates the local benchmark mean online according to the following formula. and local noise variance :

[0012]

[0013]

[0014] in The forgetting factor is used; an adaptive envelope threshold upper limit is generated based on the local noise floor variance. Lower limit and tolerance drift ,in Based on the sensitivity multiplier, This represents the drift margin. Through this recursive estimation, the system can obtain statistical parameters characterizing the current electromagnetic environment purity at the grid connection point in real time during each sampling period without storing a large amount of historical data, thus achieving automatic adaptation to different noise levels.

[0015] As a preferred embodiment of the present invention, the recursive formula for the bidirectional cumulative sum test statistic is as follows:

[0016]

[0017]

[0018] The recursive variance adaptive event-triggered discrimination module also features a benchmark locking protection mechanism. When the cumulative sum statistic in any direction first exceeds a preset locking threshold, the recursive updates of the local benchmark mean and local noise floor variance are frozen until the event trigger decision is completed and the event recording window is closed. This locking mechanism prevents outliers from continuously flooding the recursive benchmark during disturbances, ensuring that the cumulative sum test always uses the steady-state benchmark as a reference throughout the entire disturbance evolution.

[0019] As a preferred embodiment of the present invention, the calculation formula for the theoretical expected response compensation energy in the integral model time-domain quantization assessment module is as follows:

[0020]

[0021] in As the reference virtual inertia time constant, The rated apparent power capacity of the station, The nominal rated frequency of the power grid. The starting point of the event. The moment of frequency extremum; the formula for calculating the actual output compensation energy is:

[0022]

[0023] in As the reference for steady-state active power before the disturbance, The sampling interval is given. This integral formula uses the squared difference of frequency values ​​instead of the instantaneous rate of change of frequency as the theoretical basis for calculation, thus completely avoiding the risk of numerical divergence when the differential component approaches zero.

[0024] As a preferred embodiment of the present invention, the integral model time-domain quantization assessment module outputs the final inertial response effectiveness comprehensive score after response delay correction according to the following formula:

[0025]

[0026] in , The response start-up delay experienced when the actual active power increment curve crosses the preset percentage threshold dead zone of the rated capacity. To the maximum allowable start-up delay, This is the penalty attenuation factor. The station is not penalized when the response delay does not exceed the allowable limit; otherwise, the score decreases exponentially with the amount of delay.

[0027] A millisecond-level assessment method for the grid-connected operation of new energy sources oriented towards power grid dispatch includes the following steps:

[0028] The three-phase AC voltage and current at the grid connection point of the new energy power station are synchronously sampled at a sampling frequency of not less than 10kHz. The instantaneous frequency estimate, active power and reactive power are calculated in real time through a sliding data window, and a microsecond-level absolute timestamp is added to each sampled data.

[0029] A recursive algorithm with a forgetting factor is used to update the local baseline mean and local noise floor variance of the frequency sequence online. Based on the local noise floor variance, an adaptive envelope threshold boundary and tolerance drift amount are dynamically generated. A bidirectional cumulative sum test statistic is constructed to accumulate persistent deviations exceeding the tolerance drift amount in the time domain. When the cumulative sum test statistic exceeds the preset limit trigger threshold, the event trigger is determined to be established and the event time domain slice window is locked.

[0030] For the data within the time-domain slice window of the event, the theoretical expected response compensation energy is determined by the squared difference between the event start time and the frequency extreme time, and the actual output compensation energy is determined by the time-domain integral of the active power deviation relative to the steady-state reference before the disturbance. The comprehensive score of the inertial response effectiveness is output by the ratio of the two.

[0031] For concurrently triggered voltage transient events, the actual reactive power regulation energy is quantified by the time-domain integral area of ​​the reactive power deviation, and the conformity of the voltage recovery trajectory is quantified by the integral deviation between the actual voltage trajectory and the preset standard recovery corridor.

[0032] As a preferred technical solution of the present invention, in the step of constructing the bidirectional cumulative sum test statistic, when the cumulative sum statistic in either direction exceeds the preset lock-in threshold for the first time, the recursive update of the local benchmark mean and local noise floor variance is frozen until the event recording window is closed, so as to avoid the continuous contamination of the recursive benchmark by the disturbance value.

[0033] As a preferred technical solution of the present invention, the output inertia response effectiveness comprehensive score further includes a response delay penalty correction step, which retrieves the response start delay when the actual active power increment curve crosses the preset percentage threshold dead zone of the rated capacity. Output the corrected score using the following formula:

[0034]

[0035] The penalty mechanism applies to the base score in a multiplicative manner, ensuring that the corrected score always falls within a reasonable percentage range.

[0036] A computer device includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0037] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] This invention introduces a two-layer event triggering discrimination mechanism within the edge assessment instrument, combining online estimation of recursive variance with a forgetting factor and dynamic adaptive cumulative sum. This mechanism dynamically scales the event triggering threshold according to the real-time background noise level of the grid connection point. Furthermore, it uses cumulative sum statistics to accumulate deviations in the time domain rather than instantaneously comparing them. While maintaining high sensitivity, this mechanism effectively shields invalid triggers caused by random noise pulses and high-frequency harmonic ripples, thereby reducing the false triggering rate and missed detection rate in high-noise environments.

[0040] This invention reconstructs the mathematical basis of the assessment indicators from a differential model based on the instantaneous frequency change rate to an energy integral model based on the squared difference of the frequency endpoints and the area of ​​the power time-domain integral. The theoretical assessment benchmark is directly determined by the frequency squared difference without the introduction of noisy differential variables. The actual support contribution is solved by the time-domain integral of the power deviation. The integral operation has a natural low-pass smoothing characteristic for high-frequency noise, eliminating the numerical singularity problem of traditional differential equations when the frequency change rate approaches zero. It can also reflect the cumulative effective work done by the station throughout the entire transient process, improving the numerical stability and physical objectivity of the assessment results. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the system architecture according to an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of the assessment method flow according to an embodiment of the present invention.

[0043] The attached diagrams are labeled as follows: 100, Millisecond-level assessment instrument for new energy grid-connected operation; 101, Wideband power quality synchronous sampling and preprocessing module; 102, High-precision time synchronization and tag alignment module; 103, Recursive variance adaptive event triggering discrimination module; 104, Integral model time-domain quantization assessment module; 104-A, Inertial response effectiveness integral index calculation submodule; 104-B, Fast voltage regulation auxiliary service integral index calculation submodule; 105, Multi-dimensional situational data and assessment result compression communication module; 200, Dispatch-side new energy assessment system; 201, Time consistency alignment and multi-source massive data aggregation verification module; 202, Global assessment index penetration recalculation and dispatch settlement application output module. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the embodiments of this invention will be described in detail below with reference to the accompanying drawings. It should be noted that the following embodiments are for illustrative purposes only and are not intended to limit the invention. Other implementation methods obtained by those skilled in the art from these embodiments without creative effort are all within the protection scope of this invention.

[0045] like Figure 1As shown, this invention provides a millisecond-level assessment system for renewable energy grid-connected operation oriented towards power grid dispatch. The system adopts an edge-cloud collaborative architecture in its overall physical topology, including a millisecond-level assessment instrument 100 deployed at the grid connection points of various renewable energy power plants, and a dispatch-side renewable energy assessment system 200 centrally deployed in the power grid dispatch data center. The millisecond-level assessment instrument 100, acting as an edge computing device, is directly installed at the grid connection point between the renewable energy power plant and the power grid. It is responsible for high-frequency synchronous sampling of electrical physical quantities at the grid connection point and performing event identification and quantification assessment calculations locally. The dispatch-side renewable energy assessment system 200 is deployed in the provincial or municipal power grid dispatch control center, serving as a global data aggregation and decision output platform. It undertakes functions such as spatiotemporal alignment verification, cross-validation, and auxiliary service settlement and clearing of assessment results from multiple power plants.

[0046] The hardware computing platform of the millisecond-level assessment instrument 100 for grid-connected new energy can adopt a heterogeneous architecture combining a field-programmable gate array (FPGA) and a multi-core digital signal processor (DSP), or it can use a high-performance embedded processor platform running a multi-tasking real-time operating system to meet the real-time scheduling requirements at the millisecond or even microsecond level. The internal software system of the millisecond-level assessment instrument 100 is functionally divided into a wideband power quality synchronous sampling and preprocessing module 101, a high-precision time synchronization and tag alignment module 102, a recursive variance adaptive event triggering discrimination module 103, an integral model time-domain quantization assessment module 104, and a multi-dimensional situational data and assessment result compression communication module 105. Real-time data flow and interaction between these modules are achieved through a shared high-speed circular memory queue, which will be described in detail below.

[0047] The broadband power quality synchronous sampling and preprocessing module 101 directly connects to the underlying hardware acquisition interface of the millisecond-level assessment instrument 100 for new energy grid-connected operation. It continuously drives and acquires the instantaneous values ​​of the three-phase AC voltage of the voltage transformer and current transformer at the grid connection point through a high-precision analog-to-digital converter. Instantaneous value of three-phase alternating current Considering that the transient changes at the grid connection points of new energy power plants contain rich broadband characteristic information, the system will use the basic sampling frequency. The sampling interval is set to no less than 10kHz. No more than 0.1ms. In actual engineering deployments, this sampling frequency can be appropriately increased according to the accuracy requirements and hardware conditions. For example, for applications that need to capture higher frequency harmonic components, the sampling frequency can be increased to 20kHz or even higher. For discrete digital sequences obtained after analog-to-digital conversion... and The broadband power quality synchronous sampling and preprocessing module 101 maintains a sliding data window with a width of one standard power frequency cycle. Taking a power system with a nominal frequency of 50Hz as an example, this window length corresponds to 20ms, containing 200 sampling points at a sampling rate of 10kHz. The broadband power quality synchronous sampling and preprocessing module 101 uses a sliding discrete Fourier transform algorithm or a phase-locked loop algorithm to perform recursive calculations when each new sampling point arrives, and calculates and outputs the fundamental voltage amplitude of the grid connection point at the current moment in real time. fundamental current amplitude Instantaneous active power Instantaneous reactive power and instantaneous frequency estimate The pre-cleaned feature scalar data is continuously written to a circular memory queue, which is implemented using a first-in-first-out circular buffer structure, allowing downstream advanced algorithm modules to access and retrieve data as needed. In an optional implementation, the broadband power quality synchronous sampling and preprocessing module 101 can also calculate auxiliary power quality indicators such as total harmonic distortion rate in parallel, in order to provide a more complete signal feature profile for subsequent event analysis.

[0048] The core task of the high-precision time synchronization and tag alignment module 102 is to eliminate data time differences caused by geographical distance and communication links between widely distributed new energy power stations. The high-precision time synchronization and tag alignment module 102 has a built-in timing receiver chip for the BeiDou Navigation Satellite System or Global Positioning System, and maintains continuous synchronization calibration with the master clock source within the substation through a precise time protocol. Every time the broadband power quality synchronization sampling and preprocessing module 101 generates a discrete parameter, the high-precision time synchronization and tag alignment module 102 immediately appends an absolute timestamp based on International Atomic Time to it. The timestamp has a resolution at the microsecond level. Therefore, the data corresponding to each sampling moment is organized into a format containing... A complete time-series data dictionary with time tags. This time tagging mechanism ensures that when the dispatch center aggregates data from multiple sites, it can accurately align time-domain slice data from different geographical locations to a unified absolute time coordinate system. In the extreme case of temporary loss of satellite signal, the high-precision time synchronization and tag alignment module 102 can switch to a local high-stability temperature-controlled crystal oscillator for timekeeping and automatically complete the smooth correction of time deviation after the satellite signal is restored.

[0049] The recursive variance adaptive event triggering discrimination module 103 is one of the core innovative modules that distinguishes this invention from existing technologies. Traditional grid connection point assessment devices generally rely on a hard comparison of the frequency change rate or voltage change rate with a pre-set fixed threshold in the event identification stage. This static threshold determination mechanism faces an irreconcilable contradiction in the real grid edge environment: if the threshold setting is too low, it is easily affected by frequent breakthroughs in background noise and inverter high-frequency switching ripple, resulting in a large number of invalid triggers; if the threshold setting is too high, it loses the sensitivity to real small, slowly changing disturbances. The recursive variance adaptive event triggering discrimination module 103 fundamentally abandons the above-mentioned static threshold determination paradigm and instead adopts a statistical trend detection algorithm based on dynamic environment learning, opening parallel inspection threads for grid connection point frequency events and voltage events respectively. Taking the frequency anomaly event triggering thread as an example, its algorithm flow includes the following three interconnected processing stages.

[0050] In the online recursive estimation stage of the local environmental baseline and noise floor variance, the discrete sampling index at the current time is defined as... The system continuously receives frequency time series containing small random noise components. To quantify the current normal noise floor level and avoid the consumption of limited memory resources in the embedded system by storing large amounts of historical data, the system employs a forgetting factor. A recursive statistical algorithm is used to update local background signal properties online. Forgetting factor. The value is typically set between 0.95 and 0.99. This parameter controls the degree to which the system retains historical observations. The closer the value is to 1, the greater the inertia in maintaining historical mean and the slower the response to newly arriving data. Conversely, a lower value assigns higher weight to new sampled values, allowing the baseline to track signal changes more quickly. Local baseline mean The recursive update is performed according to the following formula:

[0051]

[0052] Local noise floor variance The recursive update is performed according to the following formula:

[0053]

[0054] by For example, the current sample value It accounts for only 1% of the weight in each update, while the historical cumulative average By retaining 99% of the weights, the local baseline is ensured to have good smoothness and resistance to transient pulse interference. Through the two extremely lightweight recursive formulas mentioned above, the recursive variance adaptive event-triggered discrimination module 103 can obtain the standard deviation characterizing the electromagnetic background purity of the current grid connection point environment in real time within each sampling period. When the high-frequency switching harmonics of the inverter's pulse width modulation increase or when it is subjected to irregular local load disturbances, This will automatically expand, accurately reflecting the objective physical fact that the background noise level at the grid connection point is rising. During the system initialization phase, and The system can use the power grid's nominal frequency and the preset engineering experience initial variance as starting values, respectively. After several power frequency cycles of data input, it can converge to a stable online tracking state.

[0055] During the adaptive dynamic envelope threshold generation stage, the recursive variance adaptive event-triggered discrimination module 103 utilizes real-time acquired local noise characteristic indicators. It automatically generates event judgment threshold boundaries that continuously shift over time. Two fundamental parameters, freed from the constraints of absolute physical dimensions, are pre-configured: the fundamental sensitivity multiplier. and drift margin .in The typical value range is 3 to 5. The typical value range is 0.5 to 1.5. Both parameters can be calibrated and configured by the dispatching agency according to the specific operating characteristics and accuracy requirements of the power grid under its jurisdiction. Adaptive envelope threshold upper limit. and lower limit It evolves dynamically according to the following formula:

[0056]

[0057]

[0058] Meanwhile, the core tolerance drift amount used for subsequent cumulative sum algorithms is calculated:

[0059]

[0060] After this processing stage, the trigger determination line is no longer a pre-defined constant straight line, but evolves into an adaptive envelope band that dynamically changes in sync with the real-time noise level at the grid connection point. When the harmonic level of the inverter inside the station rises or is subjected to regular external electromagnetic interference, the envelope band automatically widens, effectively shielding the interference of high-frequency electromagnetic disturbances and measurement glitches on the trigger determination; when the operating environment at the grid connection point becomes calm and the background noise decreases, the envelope band narrows accordingly, and the sensitivity naturally increases, thus enabling the system to always maintain a high level of alertness to real disturbances.

[0061] In the dynamic adaptive accumulation and statistical accumulation and state discrimination stages, considering that the ordinary instantaneous envelope crossing judgment may only be caused by a single extremely strong impulse noise, the recursive variance adaptive event-triggered discrimination module 103 further constructs upper and lower bidirectional cumulative sum and test statistics based on the aforementioned adaptive envelope judgment. and These two statistics are used to accumulate persistent positive and negative biases exceeding the tolerance drift, respectively, and their recursive formulas are as follows:

[0062]

[0063]

[0064] The nonlinear truncation operation in the above formula This endows accumulation and statistics with a remarkably sophisticated automatic discrimination capability. For random white noise with a mean of zero and alternating positive and negative values, the positive and negative deviations exceeding the tolerance drift roughly cancel each other out on the time axis, and the statistics will rapidly decay to zero under the truncation operation, failing to form an effective unidirectional accumulation. However, real power grid frequency slippage or rise events, even if the initial deviation is extremely small, will continuously accumulate due to their monotonically continuous nature over time. or Injecting positive contributions in the same direction causes the statistic to exhibit a monotonically increasing snowball effect. This enables a dual cumulative effect test across both the time duration and deviation magnitude dimensions, effectively avoiding spurious triggers caused by isolated noise impulses.

[0065] It should be noted that the deviation benchmark in the above cumulative sum formula uses the average value of the previous time step. Instead of the current time average This is because the current sampling point Already participated Recursive updates, if directly used This will result in a systematic reduction of the deviation. Simultaneously, to prevent the continuous influx of abnormal frequency values ​​after a disturbance from gradually contaminating the recursive mean baseline and causing the deviation term to... Over time, the cumulative sum and statistics may be improperly compressed, their growth rate may slow down, or they may even prematurely reach zero. The recursive variance adaptive event-triggered discrimination module 103 incorporates a benchmark locking protection mechanism. Specifically, when the cumulative sum and statistics in any direction are improperly compressed, their growth rate may slow down, or they may prematurely reach zero. or The first time the preset lock-up threshold is exceeded When (the threshold is lower than the final trigger threshold) Typical value (30% to 50%), the system immediately freezes the local benchmark mean. and local variance The recursive update process locks both values ​​to the last steady-state estimate before entering a suspected disturbance state. This locked state remains until the event trigger decision is completed and the event recording window is closed, at which point the recursive estimation resumes normal operation and re-tracks the new power grid operating background. Through this locking protection mechanism, the accumulation and verification always use an uncontaminated steady-state benchmark as a reference for deviation accumulation throughout the entire disturbance evolution, ensuring that the verification sensitivity does not decrease due to benchmark drift.

[0066] The system pre-sets cumulative and limit trigger thresholds. This threshold is pre-calibrated based on the minimum disturbance tolerance energy level required by the power grid dispatching. In each sampling period... Upon arrival, the recursive variance adaptive event-triggered discrimination module 103 executes the following core event-triggered decision logic: If Greater than or Greater than If the triggering condition is deemed irreversible, the event trigger flag is set to valid, and the baseline time of the actual physical event is recorded. The system then traces forward a preset safe backward buffer number of points in the memory history circular buffer. Extract the undisturbed steady-state baseline value. and the corresponding power reference Based on this, initiate the event time-domain slice waveform recording action, and set the time range of the waveform recording window to [specify value]. ,in This refers to the pre-recorded duration before the event. Both the duration of continuous waveform recording after the event and the duration of recording can be flexibly configured according to assessment requirements. In a typical embodiment of the present invention, Set to 200ms to 500ms. The time is set from 30s to 120s to fully cover the entire process from the steady-state baseline before the disturbance to the recovery of frequency or voltage to a quasi-steady state. After triggering, and All conditions are reset to zero, the baseline lock protection is released, and the system enters the recovery monitoring state to prepare for possible subsequent events. If none of the above triggering conditions are met, the event trigger flag remains invalid, and the recursive variance adaptive event triggering discrimination module 103 continues to perform the regular recursive estimation and accumulation update operations.

[0067] For grid connection point voltage events designed to assess rapid voltage regulation ancillary service capabilities, the recursive variance adaptive event triggering discrimination module 103 targets voltage amplitude sequences. The instantiated parallel execution of the recursive variance and dynamic adaptive cumulative sum test algorithm structure is completely symmetrical with the aforementioned frequency event triggering thread, and is also configured with an independent benchmark locking protection mechanism. The two threads share the same algorithm framework but each maintains an independent set of parameters, including their respective local mean, local variance, envelope threshold boundary, and cumulative sum statistic. Regarding parameter configuration, the sensitivity multiplier and drift margin of the voltage event triggering thread can be independently set according to the physical characteristics of the voltage disturbance and scheduling evaluation criteria. For example, for the detection of voltage sag events, the sensitivity multiplier can be appropriately reduced to improve the detection speed of rapid voltage drops.

[0068] like Figure 2 As shown, the integral model time-domain quantization assessment module 104 is another core innovative module of this invention. After the recursive variance adaptive event triggering discrimination module 103 locks and extracts a segment of abnormal event window sequence data, the system delivers this data packet to the integral model time-domain quantization assessment module 104 for in-depth quantization evaluation calculations. The integral model time-domain quantization assessment module 104, starting from the underlying physical mechanism, resolutely rejects any form of noisy differential variable. Instead of division, an energy integral mapping algorithm based on the macroscopic physical dimension is used to calculate the actual work released by the power station to the grid during abnormal periods and the regulation effectiveness. The integral model time-domain quantization assessment module 104 includes two functional sub-units: the inertial response effectiveness integral index calculation sub-module 104-A and the rapid voltage regulation auxiliary service integral index calculation sub-module 104-B.

[0069] In the inertial response effectiveness integral index calculation submodule 104-A, the mathematical basis for the assessment index is derived from the synchronous generator rotor motion equation. However, instead of directly using the differential equation for instantaneous mapping calculation, the equation is simultaneously mapped along the time axis from the starting point of the event. Integrate until the frequency reaches its deepest extreme point After the integral transformation, the differential term is eliminated, and the change in equivalent inertial kinetic energy corresponding to the system frequency deviation process can be completely represented by a function of the frequency endpoint values. Specifically, the reference virtual inertial time constant allocated by the power grid to this new energy power station is known to be... The rated apparent power capacity of the station is The nominal rated frequency of the power grid is The theoretically expected response compensation energy required for the new energy power station during this transient process is calculated using the following formula:

[0070]

[0071] The physical meaning of this formula is clear and precise: the theoretical benchmark is entirely based on the frequency squared difference at the two endpoints, the start of the event and the frequency extreme value, without needing to introduce the instantaneous frequency change rate containing noise components as an intermediate calculation variable at any stage. Since the frequency value itself is a state quantity smoothly output by the sliding window algorithm of the broadband power quality synchronous sampling and preprocessing module 101, the numerical stability of its squared difference is much higher than that of the differential component after differentiating the original sampling sequence, fundamentally eliminating the mathematical divergence and numerical singularity problems caused by the frequency change rate approaching or crossing zero in traditional schemes.

[0072] When calculating the effective compensation electromagnetic energy actually output by a new energy power station, the inertial response effectiveness integral index calculation submodule 104-A targets the high-frequency actual active power sequence within the locked event window. Subtract the steady-state active power benchmark extracted by the recursive variance adaptive event-triggered discrimination module 103 before the event occurs. Perform discrete integration in the time domain:

[0073]

[0074] The integral operation is mathematically equivalent to calculating the area enclosed by the active power increment curve and the time axis, which represents the total net compensation electromagnetic energy actually injected into the grid by the power station during frequency drops. The integral operation inherently possesses low-pass smoothing suppression characteristics for high-frequency noise, significantly attenuating the influence of measurement glitches and inverter switching ripples on the final integral result. As a result, the energy calculation value obtained has high noise resistance stability.

[0075] In obtaining the theoretically expected compensation energy Compensation energy with actual output Next, the system calculates the basic score index of inertial response:

[0076]

[0077] Meanwhile, the inertial response effectiveness integral index calculation submodule 104-A performs parallel retrieval of the actual active power increment curve, specifically the time delay experienced by the curve from the event trigger moment when it crosses the threshold dead zone of 5% of the station's rated capacity, and records this as the actual response start-up delay. According to the power grid dispatch management rules regarding the maximum permissible start-up delay... The regulations (e.g.) The system outputs the final inertial response effectiveness score after response delay correction according to the following formula:

[0078]

[0079] In the formula The penalty attenuation coefficient is uniformly set by the dispatching agency based on the severity of the assessment. hour, When the value is zero, the exponent term equals 1, and the final score is the base score; the venue is not penalized for delay. Exceed When the excess is greater, the exponential decay factor approaches zero, and the final score is significantly compressed. Through this multiplicative penalty mechanism, slow-responding power plants will face score deductions that decrease exponentially with the delay, thereby incentivizing new energy power plants to continuously optimize the transient response speed of their control systems at the assessment level. At the same time, it ensures that the corrected score always remains within a reasonable percentage range and does not result in negative values ​​or dimensional mismatches.

[0080] The rapid voltage regulation auxiliary service integral index calculation submodule 104-B performs quantitative evaluation on voltage transient over-limit events captured by the recursive variance adaptive event-triggered discrimination module 103, and also uses the time-domain integral area comparison method to replace the traditional instantaneous extreme value discrimination scheme. This submodule calculates the actual reactive power response injection total area as the equivalent regulation energy:

[0081]

[0082] in The steady-state reactive power reference extracted before the event occurred. and These represent the start and end times of the event window. The integral area quantifies the total reactive power regulation energy actually invested by the power station throughout the entire voltage transient process.

[0083] Regarding the evaluation of voltage recovery trajectory compliance, the dispatching system pre-sets target recovery curve corridors for voltage dips of different severity. This corridor defines the ideal recovery trajectory that voltage should follow during transient processes and its permissible deviation range. The rapid voltage regulation auxiliary service integral index calculation submodule 104-B calculates the actual measured voltage trajectory. By integrating and comparing with the standard corridor boundary, the absolute area deviation exceeding the corridor's tolerance range can be calculated:

[0084]

[0085] In the formula Rated voltage, The step penalty function is defined as follows: when the actual voltage trajectory is within the reasonable recovery range of the standard corridor... When the actual voltage trajectory deviates from the corridor's tolerance range By introducing this penalty function, only voltage offsets that truly violate the recovery corridor constraints are included in the cumulative calculation of area deviation; normal voltage fluctuations within the corridor will not interfere with the evaluation results. The rapid voltage regulation auxiliary service integral index calculation submodule 104-B integrates reactive power support area. The compliance status and trajectory penalty area The degree of near-zeroing is used to generate an accuracy and pass rate assessment report for a single rapid voltage regulation action, and the report is stored in the local event assessment record database of the millisecond-level tester 100 for new energy grid-connected operation.

[0086] It should be noted that the aforementioned inertia response assessment and rapid voltage regulation assessment share a high degree of mathematical unity. Both follow the core technical approach of replacing instantaneous extrema with the area of ​​time-domain integration. The only difference lies in the integration object changing from active power deviation to reactive power deviation, and from the frequency squared difference reference benchmark to the voltage recovery corridor reference benchmark. This algorithmic unity allows the millisecond-level assessment instrument 100 for new energy grid-connected operation to reuse the same integration calculation core in its software implementation. It can adapt to different types of auxiliary service assessment needs simply by switching parameters and changing the input data source, reducing the complexity of software maintenance and improving the system's scalability.

[0087] The multi-dimensional situational data and assessment result compression communication module 105 serves as the communication hub for the millisecond-level assessment instrument 100 for new energy grid-connected operation, responsible for compressing and securely transmitting the data generated by the aforementioned modules. Given the extremely limited wide-area communication bandwidth resources of the dedicated network for the power grid production control area, the multi-dimensional situational data and assessment result compression communication module 105 implements a hierarchical and differentiated data reporting strategy. For assessment scoring results, this module reports the local assessment indicator scoring dictionary as the first priority data in its entirety. The scoring dictionary includes globally unique event flow identifiers, adaptive trigger type identifiers, integral energy scale benchmark values, actual integral energy area values, the final response score after delay penalty correction, and response delay data. For the waveform data required for global spatiotemporal alignment verification and cross-validation of the new energy assessment system 200 on the dispatch side, the multi-dimensional situational data and assessment result compression communication module 105 does not report the complete 10,000 Hz-level raw sample stream, but instead performs downsampling extraction processing on the key feature parameter sequence within the locked event window. Specifically, a steady-state reference interval is selected before and after the event is triggered, and the characteristic trend sequences of frequency and voltage are preserved at a downsampling frequency of no less than one-tenth of the original sampling rate (i.e., no less than 1kHz). For the transient core interval where the frequency and voltage changes most drastically within the event window, the preservation density is appropriately increased to no less than one-quarter of the original sampling rate to ensure that the key inflection points and extreme features of the waveform are not erased by the downsampling process. In addition, the actual active power and reactive power sequences are preserved synchronously according to the same downsampling specifications as the frequency sequence. The downsampled characteristic sequences and the scoring dictionary are packaged and encapsulated together at the protocol level. Using the IoT message queue telemetry transmission protocol or the power industry standard manufacturing message specification protocol, an encrypted tunnel based on the transport layer security protocol is established to asynchronously push the message to the provincial or municipal dispatch center server. During the stable operation period without event triggering, the massive high-frequency original sampling data is discarded on-site without occupying wide-area communication resources. Through this hierarchical strategy, the multidimensional situational data and assessment results compression communication module 105 compresses the amount of communication data to a few percentages of the original sampling stream, while retaining enough key waveform information for the dispatch-side new energy assessment system 200 to support global cross-correlation alignment and randomized indicator verification.

[0088] The dispatch-side new energy assessment system 200 is deployed on the data base of the power grid dispatch control center as a software cluster with a distributed microservice architecture. It includes a time consistency alignment and multi-source massive data aggregation and verification module 201 and a global assessment indicator penetration recalculation and dispatch settlement application output module 202.

[0089] The Time Consistency Alignment and Multi-Source Massive Data Aggregation Verification Module 201 undertakes the crucial task of spatiotemporal alignment of massive asynchronously reported data. Taking a global frequency disturbance event in a large regional power grid as a typical scenario, the millisecond-level assessment instruments 100 of hundreds of renewable energy power plants connected to this grid will be triggered successively within similar time windows, each asynchronously sending grid connection point event data packets to the cloud. Since the physical fluctuations of the same power grid disturbance require tens to hundreds of milliseconds of electrical distance propagation time in the grid, the event initiation times determined by each power plant will not completely overlap. Furthermore, due to the randomness of network transmission delays, the time order in which data packets arrive at the dispatch center may not be consistent with the actual physical occurrence order. The Time Consistency Alignment and Multi-Source Massive Data Aggregation Verification Module 201 constructs a spatiotemporal slice slot based on a ring-shaped time buffer architecture, using the absolute microsecond-level timestamp carried in each edge data packet. The system utilizes the downsampling frequency and voltage characteristic trend sequences reported by each station, and performs global alignment analysis using a multidimensional signal time-series sliding cross-correlation test algorithm to accurately eliminate out-of-order and time-series errors in communication messages caused by network congestion. Based on this, a clustering algorithm is used to merge local event perceptions from different stations into a unique global time axis interval containing a clear geographical topology view of homogeneous disturbance events. This global time axis interval constitutes a unified time benchmark for subsequent cross-validation of assessment results and settlement clearing.

[0090] The global assessment indicator recalculation and scheduling settlement application output module 202, based on the unified global event timeline generated by the time consistency alignment and multi-source massive data aggregation verification module 201, extracts the independent power integral assessment scores uploaded by each station and introduces a higher-level cross-verification mechanism to eliminate possible deviations caused by relying solely on the edge perspective. The recalculation performed by the global assessment indicator recalculation and scheduling settlement application output module 202 is not a complete recalculation of the full 10,000 Hz data at the edge side, but rather a randomized verification of the integral assessment scores output by the edge side based on the downsampled key feature sequences reported by each station. Specifically, the global assessment indicator recalculation and scheduling settlement application output module 202 independently recalculates the energy area using the downsampled power sequence and frequency sequence according to the same integral formula as the integral model time-domain quantization assessment module 104, and compares this verification value with the integral results reported by the edge. If the deviation exceeds the preset allowable range, the assessment result for that station is marked as abnormal and a manual verification process is initiated. Meanwhile, the global performance indicator recalculation and dispatch settlement application output module 202 calls upon historical operation records archived in the power grid energy management system to verify whether the system's automatic generation control module or automatic voltage control module issued mandatory intervention commands for a specific renewable energy power station during a global disturbance. By deeply comparing the intentional coupling between the power station's spontaneous response behavior and dispatch commands, the global performance indicator recalculation and dispatch settlement application output module 202 can confirm whether the power station's inertia or voltage regulation auxiliary response action direction is consistent with the damping interests of the overall power grid security, and identify abnormal behaviors such as reverse deterioration regulation caused by the execution of erroneous commands or deliberate avoidance of performance evaluation by the power station.

[0091] After completing the above logical penetration verification, the global assessment indicator penetration recalculation and scheduling settlement application output module 202 will take the comprehensive score of the effectiveness of the station inertial response generated by a single event and the pass rate of the rapid voltage regulation trajectory integral, and according to the assessment configuration standards and integral accumulation rules agreed in the auxiliary service management implementation rules, perform segmented weighted accumulation according to the time span of weeks, months or quarters, and finally automatically generate a multi-dimensional new energy grid-connected operation auxiliary service execution settlement bill. This bill is pushed unidirectionally to the market-based settlement platform of the power trading center through the dedicated line safety network gate to complete the issuance and payment of assessment rewards or the systematic penalty of deduction for violations, thereby closing the complete chain of the entire grid-connected assessment mechanism from physical edge monitoring and sampling to commercial financial settlement.

[0092] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention. For example, under different engineering deployment conditions, the specific values ​​of parameters such as the forgetting factor in recursive variance estimation, the sensitivity multiplier and drift margin of the adaptive envelope threshold, the cumulative and limit triggering thresholds, the benchmark locking start threshold, and the response delay penalty attenuation coefficient can be adaptively adjusted according to the actual power grid operation characteristics and dispatch management requirements, but the technical principles and algorithm framework they follow all fall within the scope of protection of the present invention. Similarly, specific communication protocols, hardware computing platform selections, downsampling ratio strategies, and software architecture implementation methods can also be equivalently replaced according to actual engineering needs. As long as they achieve the technical effects of the core technical solutions disclosed in the present invention, they are considered equivalent implementations within the scope of protection of the present invention.

Claims

1. A millisecond-level assessment system for the grid-connected operation of new energy sources oriented towards power grid dispatch, characterized in that, The system includes a millisecond-level assessment instrument (100) for new energy grid-connected operation deployed at the grid connection point of new energy power plants and a dispatch-side new energy assessment system (200) deployed in the power grid dispatch data center; the millisecond-level assessment instrument (100) for new energy grid-connected operation includes: The wideband power quality synchronous sampling and preprocessing module (101) is used to synchronously sample the three-phase AC voltage and current at the grid connection point at a sampling frequency of not less than 10kHz, and to calculate and output the fundamental voltage amplitude in real time through a sliding data window. Instantaneous active power Instantaneous reactive power and instantaneous frequency estimate ; The high-precision time synchronization and tag alignment module (102) is used to add microsecond-level absolute timestamps to each discrete parameter output by the broadband power quality synchronization sampling and preprocessing module (101); The recursive variance adaptive event-triggered discrimination module (103) is used to employ a forgetting factor. The recursive algorithm for frequency sequences and voltage amplitude sequence The local baseline mean and local noise floor variance are updated online respectively. An adaptive envelope threshold boundary and tolerance drift amount are dynamically generated based on the local noise floor variance. A bidirectional cumulative sum test statistic is constructed to accumulate persistent deviations exceeding the tolerance drift amount in the time domain. When the cumulative sum test statistic exceeds the preset limit trigger threshold, the event trigger is determined to be established and the event time domain slice window is locked. The integral model time-domain quantization assessment module (104) is used to determine the theoretical expected response compensation energy by using the squared difference of frequency endpoints for the data within the time-domain slice window of the event. The actual output compensation energy is determined by the time-domain integral of the active power deviation. The comprehensive score of inertial response effectiveness is output by the ratio of the two, and the fast voltage regulation assessment results are output by the time-domain integral area of ​​reactive power deviation and the integral deviation of voltage recovery trajectory from the standard corridor. The multidimensional situation data and assessment result compression communication module (105) is used to package and transmit the assessment indicator scoring dictionary and the downsampled event window key feature sequence to the scheduling side new energy assessment system (200).

2. The millisecond-level assessment system for new energy grid-connected operation oriented towards grid dispatch as described in claim 1, characterized in that: The recursive variance adaptive event-triggered discrimination module (103) updates the local benchmark mean online according to the following formula. and local noise variance : ; in The forgetting factor is used; an adaptive envelope threshold upper limit is generated based on the local noise floor variance. Lower limit and tolerance drift ,in Based on the sensitivity multiplier, This represents the drift margin.

3. The millisecond-level assessment system for new energy grid-connected operation oriented towards grid dispatch as described in claim 2, characterized in that: The recursive formula for the bidirectional cumulative sum test statistic is: ; The recursive variance adaptive event triggering discrimination module (103) is also equipped with a benchmark locking protection mechanism. When the cumulative sum and statistics in any direction exceed the preset locking start threshold for the first time, the recursive update of the local benchmark mean and local noise floor variance is frozen until the event triggering decision is completed and the event recording window is closed.

4. The millisecond-level assessment system for new energy grid-connected operation oriented towards grid dispatch as described in claim 1, characterized in that: The formula for calculating the theoretical expected response compensation energy in the integral model time-domain quantization assessment module (104) is as follows: ; in As the reference virtual inertia time constant, The rated apparent power capacity of the station, The nominal rated frequency of the power grid. The starting point of the event. The moment of frequency extremum; the formula for calculating the actual output compensation energy is: ; in As the reference for steady-state active power before the disturbance, The sampling interval is denoted as .

5. The millisecond-level assessment system for new energy grid-connected operation oriented towards grid dispatch as described in claim 4, characterized in that: The integral model time-domain quantization assessment module (104) outputs the final inertial response effectiveness comprehensive score after response delay correction according to the following formula: ; in , The response start-up delay experienced when the actual active power increment curve crosses the preset percentage threshold dead zone of the rated capacity. To the maximum allowable start-up delay, This is the penalty attenuation coefficient.

6. A millisecond-level assessment method for the grid-connected operation of new energy sources oriented towards power grid dispatch, characterized in that, Includes the following steps: The three-phase AC voltage and current at the grid connection point of the new energy power station are synchronously sampled at a sampling frequency of not less than 10kHz. The instantaneous frequency estimate, active power and reactive power are calculated in real time through a sliding data window, and a microsecond-level absolute timestamp is added to each sampled data. A recursive algorithm with a forgetting factor is used to update the local baseline mean and local noise floor variance of the frequency sequence online. Based on the local noise floor variance, an adaptive envelope threshold boundary and tolerance drift amount are dynamically generated. A bidirectional cumulative sum test statistic is constructed to accumulate persistent deviations exceeding the tolerance drift amount in the time domain. When the cumulative sum test statistic exceeds the preset limit trigger threshold, the event trigger is determined to be established and the event time domain slice window is locked. For the data within the time-domain slice window of the event, the theoretical expected response compensation energy is determined by the squared difference between the event start time and the frequency extreme time, and the actual output compensation energy is determined by the time-domain integral of the active power deviation relative to the steady-state reference before the disturbance. The comprehensive score of the inertial response effectiveness is output by the ratio of the two. For concurrently triggered voltage transient events, the actual reactive power regulation energy is quantified by the time-domain integral area of ​​the reactive power deviation, and the conformity of the voltage recovery trajectory is quantified by the integral deviation between the actual voltage trajectory and the preset standard recovery corridor.

7. The millisecond-level assessment method for new energy grid-connected operation oriented towards grid dispatch as described in claim 6, characterized in that: In the step of constructing the bidirectional cumulative sum test statistic, when the cumulative sum statistic in either direction exceeds the preset lock-in threshold for the first time, the recursive update of the local benchmark mean and local noise floor variance is frozen until the event recording window is closed.

8. The millisecond-level assessment method for new energy grid-connected operation oriented towards grid dispatch as described in claim 6, characterized in that: The output inertia response effectiveness comprehensive score also includes a response delay penalty correction step, which retrieves the response start delay when the actual active power increment curve crosses the preset percentage threshold dead zone of the rated capacity. Output the corrected score using the following formula: .

9. A computer device comprising a processor and a memory, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method according to any one of claims 6 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 6 to 8.

Images