A high-precision early warning method for new energy station

Abstract

The application provides a high-precision early warning method for a new energy station; the method comprises the following steps: constructing an output deviation probability model under different operation conditions; quantifying the safety influence of control errors of AGC, AVC and primary frequency modulation on system frequency, voltage and power balance; constructing a multi-source risk feature vector, and fusing the multi-source risk feature vector by using D-S evidence theory to obtain a comprehensive risk confidence; calculating a meteorological risk correction coefficient for updating the comprehensive risk confidence, constructing a future risk prediction model for obtaining a predicted value of the comprehensive risk confidence at a future prediction time; determining an early warning level based on the predicted value of the comprehensive risk confidence at the future prediction time according to a four-level risk classification strategy; when the early warning level is level 3 or above, generating a station cooperative control instruction under a high-risk scenario based on a model predictive control method and executing the instruction; the early warning accuracy and response speed are effectively improved, and early warning is realized.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and specifically to a high-precision early warning method for new energy power plants. Background Technology

[0002] With the rapid development of high-proportion renewable energy power systems, wind power, photovoltaic (PV) and other renewable energy power plants are exhibiting dynamic characteristics characterized by strong randomness, weak disturbance resistance, and multi-timescale coupling. By the end of 2023, my country's installed capacity of wind power and PV exceeded 1 billion kilowatts, accounting for more than 36% of the country's total installed power generation capacity. Driven by the "dual carbon" target, renewable energy power plants are developing from decentralized access to clustering and base-based development, and their operating conditions are becoming increasingly complex, with output deviations having a more significant impact on the safe operation of the power grid.

[0003] Traditional early warning methods for new energy power plants have the following problems: First, traditional early warning power output deviation modeling is too simple and cannot accurately characterize the nonlinear characteristics of power station power output deviation under different operating conditions. The power output of new energy power stations is significantly affected by meteorological conditions. The power output deviation characteristics under normal operating conditions and extreme weather conditions are very different, and a single model is difficult to adapt. Second, existing methods fail to fully consider the coupling relationship between AGC / AVC control accuracy, primary frequency modulation response characteristics and system safe operation. Studies have shown that when the AGC control accuracy deteriorates to ±10%, the system frequency deviation can reach 0.35Hz, which exceeds the safe operating range. When the AVC control error exceeds 10%, it will cause continuous voltage oscillation. Third, the impact of extreme weather events (such as typhoons, low temperatures, and extreme droughts) on new energy power stations has strong uncertainty and spatiotemporal distribution characteristics. Furthermore, traditional methods mainly rely on electrical measurement data and do not effectively integrate multi-dimensional information such as weather forecasts, equipment health status, and power grid structural strength. This results in incomplete risk assessment and delayed early warning, with significant deficiencies in multi-dimensional information integration and risk quantification, making it difficult to achieve accurate early warning. Fourth, existing early warning systems have long response times, typically on the order of seconds or even minutes, making it difficult to meet the real-time requirements of high-proportion renewable energy power grids. DLT1092-2023 clearly stipulates that the total operating time of stability control devices in renewable energy power plants must be less than or equal to 30 milliseconds. Therefore, there is a lack of dynamic hierarchical strategies based on system stability margins, and collaborative optimization control of multiple power plants cannot be achieved in high-risk scenarios.

[0004] Therefore, existing methods need to be optimized. Summary of the Invention

[0005] This invention addresses the problems existing in the prior art by providing a high-precision early warning method for new energy power plants.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Step 1: Collect historical operating data and meteorological data of each new energy power station. Based on the collected data, divide the operating conditions. For each type of operating condition, establish a piecewise exponential distribution model of the output deviation and calculate the standard deviation of the output deviation. Then, establish a spatial correlation model and a spatial attenuation model of the output deviation of multiple power stations. Step 2: Establish a coupling model between AGC control accuracy and system frequency stability, a coupling relationship model between AVC control accuracy and voltage stability of grid-connected units, a voltage support boundary model of grid-connected units, and a quantitative model of the impact of primary frequency regulation control accuracy. Quantify the impact of AGC, AVC, and primary frequency regulation control errors on the safety of system frequency, voltage, and power balance. Step 3: Construct a multi-source risk feature vector, use DS evidence theory to fuse the multi-source risk feature vector, assign a basic probability allocation function to each feature, calculate the conflict coefficient between each feature, and then obtain the comprehensive risk confidence based on the evidence synthesis rules. Step 4: Based on the comprehensive risk reliability, calculate the meteorological severity index, the power loss coefficient under typhoon weather, and the power prediction error increase factor during typhoons to update the comprehensive risk reliability. Then, perform risk evolution trend analysis to obtain the risk evolution rate. Based on the comprehensive risk reliability and risk evolution rate at the current moment, construct a future risk prediction model and output the comprehensive risk reliability prediction value at the future prediction moment. Step 5: Construct a three-tiered early warning indicator system including system-level, site-level, and equipment-level indicators; set early warning activation criteria; trigger an early warning decision when any early warning activation criterion is met; determine the early warning level based on the comprehensive risk confidence prediction value of the future prediction time output in Step 4, according to the four-level risk classification strategy; and output the early warning level and corresponding response strategy. Step 6: Receive the warning level and corresponding response strategy. When the warning level is level 3 or above, generate and execute the station collaborative control command for high-risk scenarios based on the model predictive control method.

[0007] Optionally, the historical operating data includes the active power, reactive power, voltage amplitude, frequency, unit start-up and shutdown status, AGC / AVC control commands and response values, and primary frequency regulation action records of each new energy power station; The meteorological data includes wind speed, light intensity, ambient temperature, humidity, air pressure, rainfall, and extreme weather indicator information; The operating conditions include normal operating conditions, low wind speed operating conditions, high wind speed operating conditions, and extreme weather operating conditions.

[0008] Optionally, the piecewise exponential distribution model of the output deviation is specifically as follows: , In the formula, ΔP, λ1, e, μ1, λ2, and μ2 are, respectively, the output deviation, the exponential distribution rate parameter of the first interval, the natural constant, the position parameter of the first interval, the exponential distribution rate parameter of the second interval, and the position parameter of the second interval; The output deviation standard deviation is the statistical value of the standard deviation of samples within each type of working condition after K-means clustering to classify the working condition categories; The spatial correlation model of the output deviation of the multiple stations uses the Copula function to describe the joint distribution. The spatial decay model is based on the Kendall rank correlation coefficient.

[0009] Optionally, establishing the coupling model between AGC control accuracy and system frequency stability includes: Calculate the difference between the system load disturbance and the sum of the AGC control errors of all AGC-participating stations; The reciprocal of the frequency response transfer function, which is composed of the equivalent inertia of the calculation system, the load damping coefficient, and the gain and time constant of the AGC control loop of each station; Multiplying the difference by the reciprocal yields the system frequency deviation; The establishment of the coupling relationship model between AVC control accuracy and voltage stability of grid-connected generator units includes: calculating the ratio of system short-circuit reactance to rated voltage, and multiplying the ratio by reactive power control error to approximate the system voltage deviation; The establishment of the voltage support boundary model for grid-type generating units includes: Calculate the sum of the rated internal potential and the internal potential control deviation; The active-reactive power correction term is calculated based on the equivalent reactance of the grid-type unit, the active power output of the unit, the rated internal potential of the unit, and the reactive power output of the unit. Subtracting the active-reactive power output correction term from the sum yields the critical support voltage of the grid-type unit. The quantitative model for the impact of primary frequency modulation control accuracy includes: Calculate the product of the rated adjustment rate and the actual adjustment rate; Calculate the ratio of the negative descent rate control error to the product; The primary frequency regulation power response deviation is obtained by multiplying the ratio by the frequency deviation and the rated capacity of the unit.

[0010] Optionally, the multi-source risk feature vector includes electrical risk features, meteorological risk features, equipment risk features, and geographical risk features; the electrical risk features include system frequency deviation, system voltage deviation, and power fluctuation rate; the meteorological risk features include wind speed change rate, temperature extreme values, and disaster warning level. The generation of the equipment risk characteristics includes: For each piece of equipment in the station, the normalized deviation index of each state quantity is weighted and summed with the corresponding weight to obtain the risk characteristic value of a single piece of equipment; The weighting coefficients are determined based on the sensitivity of each state variable to the health status of the device. The state variable with higher sensitivity has a larger weight, and the sum of the weights of each state variable within the same device is 1. The maximum value of the risk characteristic of a single piece of equipment among all the equipment in the site is taken as the equipment risk characteristic of the site.

[0011] Optionally, the generation of the geographic risk features includes: For the target new energy power station area, based on the elevation standard deviation and slope change rate of the station area, the influence of topographic undulation on airflow / solar radiation distribution is quantified to obtain the topographic complexity factor; based on the land cover type of the station area, the influence of underlying surface roughness on wind speed attenuation or solar radiation reflection is quantified to obtain the underlying surface type factor; based on the statistical frequency of extreme meteorological events such as typhoons, thunderstorms, and freezing in the station area within a preset historical period, the historical risk level of meteorological disasters in the area is quantified to obtain the meteorological disaster historical frequency factor; based on indicators such as the number of transmission lines at the station grid connection point, line length, proportion of transmission lines erected on the same tower, and N-1 throughput rate, the structural strength and redundancy of the power transmission channel of the station are quantified to obtain the power grid structural vulnerability factor. Each geographical feature factor is normalized to the [0,1] interval, and weight coefficients are assigned according to the significance of each factor's impact on the station's operational risk. The sum of the weights equals 1. The geographical risk characteristic index of the station is obtained by weighting and summing the normalized geographical characteristic factors with their corresponding weights.

[0012] Optionally, the generation of the power volatility includes: After K-means clustering to classify the working conditions, the standard deviation of the samples within each working condition is used as the basic power fluctuation intensity of the target station under the corresponding working condition. Based on the joint distribution of multiple power stations described by the Copula function and the Kendall spatial attenuation model, the spatial correlation coefficient of the output deviation between the target power station and all surrounding power stations is calculated. Based on the correlation coefficient, the base power fluctuation intensity of the target power station is spatially superimposed and corrected to obtain the corrected fluctuation intensity taking into account the spatial correlation of multiple power stations. Within a preset evaluation time window, the spatially corrected instantaneous fluctuation intensity is time-weighted and averaged to obtain the power fluctuation rate within that time window.

[0013] Optionally, the modified comprehensive risk confidence includes: Real-time meteorological data is collected, and the measured values ​​of each meteorological parameter are converted into dimensionless values ​​in the interval [0,1] according to the preset severity mapping function. Weights are assigned according to the significance of the impact of each meteorological parameter on the output of the new energy power station, and the meteorological severity index is obtained by weighted summation. For typhoon weather, the rated power curve of the station units is compared with the actual operating power curve. The cumulative energy loss during the typhoon is calculated by integration, and the turbulence intensity factor is introduced to correct the loss value to obtain the output loss coefficient under typhoon weather. Using the measured maximum wind speed as the core input, the power prediction error is increased by a factor of 100% by the fitting coefficients of historical typhoon events. The meteorological risk correction coefficient is obtained by weighting and aggregating the meteorological severity index, the power loss coefficient, and the power prediction error amplification factor. The meteorological risk characteristics at the current moment output in step three are corrected using the meteorological risk correction coefficient, and the meteorological-corrected comprehensive risk confidence at the current moment is obtained after DS evidence fusion.

[0014] Optionally, the risk evolution trend analysis includes: Subtract the comprehensive risk confidence score of the previous time from the comprehensive risk confidence score of the current time after weather correction, and then divide by the sampling time interval to obtain the risk evolution rate of the current time. When the risk evolution rate is greater than 0, the risk is judged to be increasing; when the risk evolution rate is equal to 0, the risk is judged to be in a stable state; when the risk evolution rate is less than 0, the risk is judged to be decreasing. The method of constructing a future risk prediction model based on the comprehensive risk confidence and risk evolution rate at the current moment, and outputting the comprehensive risk confidence prediction value at the future moment, includes: adding the risk value at the current moment to the multiplication value of the prediction time window and the risk evolution rate at the current moment, to obtain the comprehensive risk confidence prediction value at the future prediction moment.

[0015] Optionally, generating station collaborative control commands for high-risk scenarios based on model predictive control methods includes: With the control objectives of minimizing overall risk and minimizing the power adjustment costs of each station, an objective function is constructed, with power balance constraints, voltage constraints, and control quantity constraints as constraints. At the current moment, based on the future risk prediction sequence provided in step four, the comprehensive risk confidence provided in step three, the output deviation characteristics provided in step one, and the control accuracy constraints provided in step two, the above-mentioned constrained optimization problem is solved online to obtain the optimal adjustment sequence of active power and reactive power of each new energy power station in the prediction time domain. The optimal control command at the current moment is taken as the actual execution command, and the actual execution command is sent to each new energy power station for execution.

[0016] Compared with the prior art, the present invention has the following beneficial effects: First, the operating conditions are divided into normal operating conditions, low wind speed operating conditions, high wind speed operating conditions, and extreme weather operating conditions according to wind speed, light intensity, and temperature. A piecewise exponential distribution model is independently established for each type of operating condition, and K-means clustering is used to calculate the output deviation standard deviation of each type of operating condition. This overcomes the problem of poor adaptability of a single model, achieves accurate characterization under different operating conditions, and improves the pertinence and effectiveness of early warning. Second, we establish AGC coupled with frequency stability, grid-type AGC coupled with voltage stability, grid-type voltage support boundary model, and primary frequency modulation influence quantification model, respectively, to quantify the safety impact boundary of control error on frequency (0.35Hz) and voltage (continuous oscillation), thus making up for the shortcomings of existing methods in control accuracy coupling analysis. Third, a four-dimensional risk feature vector of electrical, meteorological, equipment, and geographical dimensions is constructed. DS evidence theory is used to fuse multi-source heterogeneous information, assigning a basic probability allocation function to each feature and calculating the conflict coefficient between features. Then, a comprehensive risk reliability is obtained based on evidence synthesis rules. Based on the comprehensive risk reliability, the meteorological severity index, the power loss coefficient under typhoon weather, and the power prediction error increase factor during typhoons are calculated to update the comprehensive risk reliability. Finally, risk evolution trend analysis is performed to obtain the risk evolution rate. Based on the comprehensive risk reliability and risk evolution rate at the current moment, a future risk prediction model is constructed, outputting the comprehensive risk reliability prediction value at the future prediction moment. This model can provide an early warning of extreme weather risks 45 minutes in advance, allowing dispatchers sufficient response time. The early warning accuracy rate has increased from 75% to over 92%, solving the problems of incomplete assessment and delayed early warning. Fourth, a three-tiered early warning indicator system is constructed, including system-level, station-level, and equipment-level indicators. Early warning activation criteria are set, and an early warning decision is triggered when any one of the criteria is met. Based on the comprehensive risk confidence prediction value at the future prediction time, the early warning level is determined according to a four-level risk classification strategy, and the early warning level and corresponding response strategy are output. When the early warning level is level 3 or above, station collaborative control instructions for high-risk scenarios are generated and executed based on model predictive control methods. Compared with independent control of a single station, this can optimize the overall operating status of the new energy power grid, improve the safe and stable operation level of a high-proportion new energy power grid, and the early warning response time is less than 80ms, meeting the real-time requirements of online early warning. This is conducive to realizing multi-station collaborative optimization control in high-risk scenarios. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the steps of a high-precision early warning method for new energy power plants. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention. It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application.

[0019] Example 1 like Figure 1 As shown, this embodiment provides a high-precision early warning method for new energy power plants, including the following steps: Step 1: Collect historical operating data and meteorological data of each new energy power station. Based on the collected data, divide the operating conditions. For each type of operating condition, establish a piecewise exponential distribution model of the output deviation and calculate the standard deviation of the output deviation. Then, establish a spatial correlation model and a spatial attenuation model of the output deviation of multiple power stations. Step 2: Establish a coupling model between AGC control accuracy and system frequency stability, a coupling relationship model between AVC control accuracy and voltage stability of grid-connected units, a voltage support boundary model of grid-connected units, and a quantitative model of the impact of primary frequency regulation control accuracy. Quantify the impact of AGC, AVC, and primary frequency regulation control errors on the safety of system frequency, voltage, and power balance. Step 3: Construct a multi-source risk feature vector, use DS evidence theory to fuse the multi-source risk feature vector, assign a basic probability allocation function to each feature, calculate the conflict coefficient between each feature, and then obtain the comprehensive risk confidence based on the evidence synthesis rules. Step 4: Based on the comprehensive risk reliability, calculate the meteorological severity index, the power loss coefficient under typhoon weather, and the power prediction error increase factor during typhoons to update the comprehensive risk reliability. Then, perform risk evolution trend analysis to obtain the risk evolution rate. Based on the comprehensive risk reliability and risk evolution rate at the current moment, construct a future risk prediction model and output the comprehensive risk reliability prediction value at the future prediction moment. Step 5: Construct a three-tiered early warning indicator system including system-level, site-level, and equipment-level indicators; set early warning activation criteria; trigger an early warning decision when any early warning activation criterion is met; determine the early warning level based on the comprehensive risk confidence prediction value of the future prediction time output in Step 4, according to the four-level risk classification strategy; and output the early warning level and corresponding response strategy. Step 6: Receive the warning level and corresponding response strategy. When the warning level is level 3 or above, generate and execute the station collaborative control command for high-risk scenarios based on the model predictive control method.

[0020] Example 2 Based on Example 1, the historical operating data includes the active power, reactive power, voltage amplitude, frequency, unit start-up and shutdown status, AGC / AVC control commands and response values, and primary frequency regulation action records of each new energy power station; The meteorological data includes wind speed, light intensity, ambient temperature, humidity, air pressure, rainfall, and extreme weather indicator information; The operating conditions include normal operating conditions, low wind speed operating conditions, high wind speed operating conditions, and extreme weather operating conditions.

[0021] It should be noted that the extreme weather indicators include, but are not limited to, typhoon warning levels and low temperature warning levels.

[0022] Specifically, it should be noted that the operating conditions are divided based on the collected data, including: Extreme weather conditions are assessed when: wind speeds consistently exceed the cut-out wind speed threshold for the generating units (typically 25 m / s) or a typhoon warning is in effect; ambient temperatures consistently fall below the unit's lower operating temperature threshold (typically -30°C) or an extreme low temperature warning is in effect; and solar irradiance consistently falls below the effective power generation threshold (typically below 50 W / m²). 2 Furthermore, if any of the above four conditions are present, the work will be classified as an extreme weather condition. If the extreme weather conditions are not met, the remaining operating conditions will be divided based on wind speed. The wind speed is within the normal operating range between the unit's rated wind speed and the cut-out wind speed (usually 12m / s~25m / s, depending on the unit model), and the solar irradiance is within the effective operating range of the photovoltaic modules (usually 200W / m²). 2 ~1000W / m 2 If the ambient temperature is within the normal operating temperature range of the unit (usually -10℃ to 40℃), it is considered to be in normal operating condition. The wind speed is below the unit's cut-in wind speed threshold (usually 3 m / s) but above zero, or the wind speed is between the cut-in wind speed and the rated wind speed but remains consistently low (usually 3 m / s to 12 m / s), and the light intensity is low (usually 50 W / m²). 2 ~200W / m 2 If the wind speed is low, it is considered a low wind speed operating condition. If the wind speed is higher than the upper limit of the normal operating range but does not reach the cut-out wind speed threshold (usually below 25m / s but significantly higher, such as 18m / s~25m / s), it is judged as a high wind speed condition. When there is a conflict in the judgment results of the three dimensions of wind speed, light intensity, and temperature (e.g., the wind speed is high but the light intensity is normal), the wind speed dimension is the main criterion, and the light intensity and temperature are auxiliary correction factors. The threshold boundary points of each working condition adopt the left-closed and right-open principle (i.e., the lower limit is included, but the upper limit is not included) to ensure that each sample uniquely belongs to one working condition category, with no overlap or omission. After the division is completed, all samples are checked for consistency to ensure that the union of the four working conditions covers the entire time period and the intersection is empty.

[0023] Example 3 Based on Example 2, the piecewise exponential distribution model of the output deviation is specifically as follows: , In the formula, ΔP, λ1, e, μ1, λ2, and μ2 are, respectively, the output deviation, the exponential distribution rate parameter of the first interval, the natural constant, the position parameter of the first interval, the exponential distribution rate parameter of the second interval, and the position parameter of the second interval; The output deviation standard deviation is the statistical value of the standard deviation of samples within each type of working condition after K-means clustering to classify the working condition categories; The spatial correlation model of the output deviation of the multiple stations uses the Copula function to describe the joint distribution. The spatial decay model is based on the Kendall rank correlation coefficient.

[0024] Specifically, it should be noted that the formula for calculating the standard deviation of the output deviation is as follows: , In the formula, σ k N represents the overall standard deviation of output deviation under the k-th operating condition; k is the operating condition category number, k=1, 2, 3, 4 correspond to normal operating condition, low wind speed condition, high wind speed condition, and extreme weather condition, respectively; k μ is the total number of samples for the k-th type of working condition; ΔP，k ΔP is the arithmetic mean of the output deviations of all samples under the k-th working condition; k，j This represents the output deviation value of the j-th sample point under the k-th working condition.

[0025] Specifically, it should be noted that the joint distribution of the Copula function is as follows: , In the formula, C(u) i ,u j ) is the Copula joint distribution function, which connects the marginal distributions of two power stations into a joint distribution, used to describe the complete dependent structure of the output deviation between the two power stations; u i u j The marginal cumulative probability values ​​for stations i and j are, respectively, u. i The range is [0,1], u j The range is [0,1]; ΔP i ΔP j The output deviation random variables for stations i and j are, respectively; F i -1 (u) i ), F j -1 (u)j The cumulative distribution function F at the edge of station i is shown in the following order. i The inverse function (quantile function), the cumulative distribution function F at the edge of station j j The inverse function of the cumulative probability u i u j The mapping is to the corresponding output deviation value; P(·) is a probability operator, representing the joint probability of the events within the brackets occurring simultaneously.

[0026] Specifically, it should be noted that the spatial attenuation model is as follows: , In the formula, τ(d ij The Kendall rank correlation coefficient (KRC) is the output deviation between power stations i and j, and is a measure of their spatial correlation. Its value ranges from [...]. [1,1], positive values ​​indicate positive correlation, and negative values ​​indicate negative correlation; d ij τi represents the geographical distance between station i and station j, usually in km, and is the core spatial factor affecting the correlation; τ0 is the reference correlation coefficient, i.e., when d ij The theoretical maximum correlation coefficient, approaching 0, reflects the upper limit of correlation within the station itself or at very close distances, typically τ0 ≤ 1; α is the spatial attenuation coefficient, usually measured in km. 1 The rate at which the correlation coefficient decreases with increasing distance is controlled; the larger the α, the faster the spatial correlation decreases with distance.

[0027] Example 4 Based on Example 1, the establishment of the coupling model between AGC control accuracy and system frequency stability includes: Calculate the difference between the system load disturbance and the sum of the AGC control errors of all AGC-participating stations; The reciprocal of the frequency response transfer function, which is composed of the equivalent inertia of the calculation system, the load damping coefficient, and the gain and time constant of the AGC control loop of each station; Multiplying the difference by the reciprocal yields the system frequency deviation; The establishment of the coupling relationship model between AVC control accuracy and voltage stability of grid-connected generator units includes: calculating the ratio of system short-circuit reactance to rated voltage, and multiplying the ratio by reactive power control error to approximate the system voltage deviation; The establishment of the voltage support boundary model for grid-type generating units includes: Calculate the sum of the rated internal potential and the internal potential control deviation; The active-reactive power correction term is calculated based on the equivalent reactance of the grid-type unit, the active power output of the unit, the rated internal potential of the unit, and the reactive power output of the unit. Subtracting the active-reactive power output correction term from the sum yields the critical support voltage of the grid-type unit. The quantitative model for the impact of primary frequency modulation control accuracy includes: Calculate the product of the rated adjustment rate and the actual adjustment rate; Calculate the ratio of the negative descent rate control error to the product; The primary frequency regulation power response deviation is obtained by multiplying the ratio by the frequency deviation and the rated capacity of the unit.

[0028] Specifically, it should be noted that the coupling model expression between AGC control accuracy and system frequency stability is as follows: , In the formula, Δf(s) is the Laplace transform of the system frequency deviation, reflecting the frequency shift caused by AGC control error; s is the Laplace operator, used for frequency domain analysis; H s D is the system's equivalent inertia constant, reflecting the power system's inertial resistance to frequency changes; L The load damping coefficient reflects the load's regulation characteristics with frequency variations; its unit is K. i T represents the gain coefficient of the AGC control loop of the i-th new energy power station, which is dimensionless; i ΔP represents the time constant of the AGC control loop of the i-th renewable energy power station, reflecting the AGC response speed; NAGC represents the total number of renewable energy power stations participating in AGC regulation; ΔP L (s) represents the Laplace transform of the system load disturbance, reflecting the power fluctuations on the load side; E AGC，i (s) represents the AGC control error of the i-th station, which is the difference between the actual AGC response value and the ideal command value.

[0029] Specifically, it should be noted that there are three typical scenarios for the coupled model of AGC control accuracy and system frequency stability. The first scenario is that the load disturbance is exactly offset by the sum of all AGC errors at all stations, meaning the numerator of the model expression is 0. In this case, the system frequency deviation is 0, which is the ideal state, i.e., no frequency deviation. The second scenario is that the load disturbance is greater than the sum of AGC errors, meaning the numerator of the model expression is greater than 0. In this case, the system frequency deviation is greater than 0, the system power is excessive, the frequency is too high, and power generation needs to be reduced. The third scenario is that the load disturbance is less than the sum of AGC errors, meaning the numerator of the model expression is less than 0. In this case, the system frequency deviation is less than 0, the system power is insufficient, the frequency is too low, and power generation needs to be increased.

[0030] Specifically, it should be noted that AGC control error refers to the deviation between the actual AGC response value of the station and the dispatch command value, which can be positive or negative; system load disturbance refers to the amount by which the actual load deviates from the planned load, which can be positive or negative; when the sum of the AGC control errors of all stations equals the load disturbance, it means that the total deviation of the actual output of each station exactly compensates for the load deviation, the system reaches a new power balance point, and the frequency deviation is zero.

[0031] Specifically, it should be noted that the coupling relationship model expression between the AVC control accuracy and voltage stability of grid-connected generator units is as follows: , In the formula, ΔU is the system voltage deviation, i.e., the difference between the actual voltage and the rated voltage; x a U0 is the system short-circuit reactance, reflecting the short-circuit capacity characteristics of the power grid under faults or disturbances; U0 is the system rated voltage, which is the reference voltage value; ΔQ is the reactive power control error, which is the difference between the actual reactive power output and the AVC command value.

[0032] Specifically, it should be noted that the expression for the voltage support boundary model of grid-connected units is as follows: , In the formula, U crit E0 is the critical support voltage for grid-connected generating units, i.e., the lowest voltage boundary value at which the unit can maintain stable operation; E0 is the rated internal electromotive force, the rated value of the internal electromotive force of the grid-connected generating unit, reflecting the voltage source characteristic reference of the unit; ΔE is the internal electromotive force control deviation, i.e., the difference between the actual internal electromotive force and the rated internal electromotive force, reflecting the error of AVC control accuracy in regulating the internal electromotive force; x b P represents the equivalent reactance of the grid-connected generating unit (including the step-up transformer and lines); Q represents the active power output of the generating unit; and Q represents the reactive power output of the generating unit.

[0033] Specifically, it should be noted that the quantitative model expression for the impact of primary frequency modulation control accuracy is as follows: , In the formula, ΔP freq This refers to the primary frequency modulation power response deviation, i.e., the deviation between the actual primary frequency modulation output and the ideal primary frequency modulation output; δ R The droop control error, i.e., the deviation between the actual droop rate and the rated droop rate, reflects the degree of degradation in the accuracy of the primary frequency modulation control and is dimensionless; R i The rated droop rate is a rated parameter of the primary frequency regulation characteristic of a generator set. It reflects the proportion of power adjustment per unit change in frequency and is dimensionless. N The rated capacity of the unit is denoted as ; the negative sign in the expression indicates that the primary frequency regulation has a negative feedback regulation characteristic, that is, the power generation increases when the frequency decreases and decreases when the frequency increases.

[0034] Example 5 Based on Example 4, the multi-source risk feature vector includes electrical risk features, meteorological risk features, equipment risk features, and geographical risk features; the electrical risk features include system frequency deviation, system voltage deviation, and power fluctuation rate; the meteorological risk features include wind speed change rate, temperature extreme values, and disaster warning level. The generation of the equipment risk characteristics includes: For each piece of equipment in the station, the normalized deviation index of each state quantity is weighted and summed with the corresponding weight to obtain the risk characteristic value of a single piece of equipment; The weighting coefficients are determined based on the sensitivity of each state variable to the health status of the device. The state variable with higher sensitivity has a larger weight, and the sum of the weights of each state variable within the same device is 1. The maximum value of the risk characteristic of a single piece of equipment among all the equipment in the site is taken as the equipment risk characteristic of the site.

[0035] Example 6 Based on Example 5, the generation of the geographic risk features includes: For the target new energy power station area, based on the elevation standard deviation and slope change rate of the station area, the influence of topographic undulation on airflow / solar radiation distribution is quantified to obtain the topographic complexity factor; based on the land cover type of the station area, the influence of underlying surface roughness on wind speed attenuation or solar radiation reflection is quantified to obtain the underlying surface type factor; based on the statistical frequency of extreme meteorological events such as typhoons, thunderstorms, and freezing in the station area within a preset historical period, the historical risk level of meteorological disasters in the area is quantified to obtain the meteorological disaster historical frequency factor; based on indicators such as the number of transmission lines at the station grid connection point, line length, proportion of transmission lines erected on the same tower, and N-1 throughput rate, the structural strength and redundancy of the power transmission channel of the station are quantified to obtain the power grid structural vulnerability factor. Each geographical feature factor is normalized to the [0,1] interval, and weight coefficients are assigned according to the significance of each factor's impact on the station's operational risk. The sum of the weights equals 1. The geographical risk characteristic index of the station is obtained by weighting and summing the normalized geographical characteristic factors with their corresponding weights.

[0036] Specifically, it should be noted that the generation of the comprehensive risk confidence score includes: A risk identification set for the operation of new energy power stations is defined, namely Z = {high risk H, medium risk M, low risk L}. This framework covers all possible risk level assumptions, and the assumptions are mutually exclusive and complete. For four information sources—electrical risk characteristics, meteorological risk characteristics, equipment risk characteristics, and geographical risk characteristics—the risk characteristic indicators with continuously taking values ​​are mapped to the basic probability allocation function m on the identification framework. j(A) For each information source j, its risk characteristic value is transformed into a basic confidence assignment to each hypothesis (high risk H, medium risk M, low risk L) in the identification set according to a preset membership function, satisfying: , where m j (Z) represents the uncertainty confidence level assigned to the entire identification set, reflecting the incompleteness of information from the information source. The higher the uncertainty, the higher the confidence level. j (Z) The larger the value, the better; Calculate the conflict coefficient K between each pair of basic probability assignment functions of each information source: , In the formula, A and B are arbitrary subsets on the identification set Z, and K reflects the degree of evidence conflict between different information sources. The larger the value of K, the more serious the disagreement between information sources. The basic probability allocation functions of the four information sources are synthesized in pairs using an iterative process. The formula is a publicly available formula and its specific expression is not given here. When K < 1, the synthesis result is valid. When K ≥ 1, it indicates that the evidence is completely conflicting and the conflict coefficient needs to be corrected (by using the weighted average method or the discount coefficient method to reduce the weight of highly conflicting evidence) before synthesis. After synthesis, the comprehensive basic reliability allocation m(H), m(M), m(L) of each risk level hypothesis and the overall uncertainty reliability m(Z) are obtained. The weighted sum of the weight coefficients of each risk level assumption and the corresponding basic reliability assignment is used to obtain the comprehensive risk reliability. Uncertain reliability m(Z) that is not assigned to a specific risk assumption after synthesis is not included in the calculation of comprehensive risk reliability.

[0037] Example 7 Based on Example 5, the generation of the power fluctuation rate includes: After K-means clustering to classify the working conditions, the standard deviation of the samples within each working condition is used as the basic power fluctuation intensity of the target station under the corresponding working condition. Based on the joint distribution of multiple power stations described by the Copula function and the Kendall spatial attenuation model, the spatial correlation coefficient of the output deviation between the target power station and all surrounding power stations is calculated. Based on the correlation coefficient, the base power fluctuation intensity of the target power station is spatially superimposed and corrected to obtain the corrected fluctuation intensity taking into account the spatial correlation of multiple power stations. Within a preset evaluation time window, the spatially corrected instantaneous fluctuation intensity is time-weighted and averaged to obtain the power fluctuation rate within that time window.

[0038] Example 8 Based on Example 5, the modified comprehensive risk confidence includes: Real-time meteorological data is collected, and the measured values ​​of each meteorological parameter are converted into dimensionless values ​​in the interval [0,1] according to the preset severity mapping function. Weights are assigned according to the significance of the impact of each meteorological parameter on the output of the new energy power station, and the meteorological severity index is obtained by weighted summation. For typhoon weather, the rated power curve of the station units is compared with the actual operating power curve. The cumulative energy loss during the typhoon is calculated by integration, and the turbulence intensity factor is introduced to correct the loss value to obtain the output loss coefficient under typhoon weather. Using the measured maximum wind speed as the core input, the power prediction error is increased by a factor of 100% by the fitting coefficients of historical typhoon events. The meteorological risk correction coefficient is obtained by weighting and aggregating the meteorological severity index, the power loss coefficient, and the power prediction error amplification factor. The meteorological risk characteristics at the current moment output in step three are corrected using the meteorological risk correction coefficient, and the meteorological-corrected comprehensive risk confidence at the current moment is obtained after DS evidence fusion.

[0039] Example 9 Based on Example 8, the risk evolution trend analysis includes: Subtract the comprehensive risk confidence score of the previous time from the comprehensive risk confidence score of the current time after weather correction, and then divide by the sampling time interval to obtain the risk evolution rate of the current time. When the risk evolution rate is greater than 0, the risk is judged to be increasing; when the risk evolution rate is equal to 0, the risk is judged to be in a stable state; when the risk evolution rate is less than 0, the risk is judged to be decreasing. The method of constructing a future risk prediction model based on the comprehensive risk confidence and risk evolution rate at the current moment, and outputting the comprehensive risk confidence prediction value at the future moment, includes: adding the risk value at the current moment to the multiplication value of the prediction time window and the risk evolution rate at the current moment, to obtain the comprehensive risk confidence prediction value at the future prediction moment.

[0040] Specifically, a three-tiered early warning indicator system was constructed, comprising system-level, site-level, and equipment-level early warning indicators. The system-level early warning indicators include global system frequency deviation and global system voltage deviation. The site-level early warning indicators include single-site output deviation, power fluctuation rate, and control response deviation. The equipment-level early warning indicators include key equipment status parameters such as transformer oil temperature, circuit breaker contact resistance, wind turbine gearbox vibration amplitude, and photovoltaic inverter power module junction temperature.

[0041] Specifically, it should be noted that the warning activation criteria include voltage over-limit warning criteria, power deviation warning criteria, and frequency stability warning criteria; When the system's global voltage deviation or the voltage deviation of any critical node exceeds the preset normal operating voltage threshold range (usually ±5% of the rated voltage), the voltage over-limit warning criterion is triggered. When the absolute value of the output deviation of any new energy power station exceeds the preset percentage threshold (usually ±10%) of the station's rated capacity, the power deviation warning criterion is triggered. When the global frequency deviation of the system exceeds the preset normal operating frequency threshold range (usually ±0.2Hz of the rated frequency), the frequency stability warning criterion is triggered.

[0042] Specifically, it should be noted that the four-level risk classification strategy includes Level 1 warning, Level 2 warning, Level 3 warning, and Level 4 warning; When the comprehensive risk confidence prediction value at the future prediction time is greater than or equal to 0.2 and less than 0.4, it is judged as a Level 1 warning, which belongs to the prompt level. The enhanced monitoring and response strategy is implemented, including encrypting the data sampling frequency to the second level, increasing the frequency of status quantity inspection, and generating operation prompt information to push to the scheduling terminal. When the comprehensive risk confidence prediction value at the future prediction time is greater than or equal to 0.4 and less than 0.6, it is judged as a Level 2 warning, which belongs to the attention level. The prevention and control response strategy is implemented, including initiating internal prevention and control measures of the station (such as adjusting the AVC voltage setting value and pre-adjusting the reactive power output of SVG) and notifying the dispatch terminal to pay attention to the operating status of the station. When the comprehensive risk confidence prediction value at the future prediction time is greater than or equal to 0.6 and less than 0.8, it is judged as a Level 3 warning, which belongs to the alarm level. An emergency control response strategy is executed, including initiating the Model Predictive Control (MPC) collaborative optimization control process in step six, adjusting the active and reactive power output plans of the station, sending an emergency alarm signal to the dispatching terminal, and suggesting the implementation of system-level control measures. When the comprehensive risk confidence prediction value at the future prediction time is greater than or equal to 0.8, it is judged as a Level 4 warning, which is a dangerous level. The unit tripping protection response strategy is executed, including immediately cutting off some or all high-risk units, activating the station emergency shutdown protection logic, sending the unit tripping action signal to the dispatch terminal and recording the action event.

[0043] Example 10 Based on Example 9, the generation of station collaborative control commands for high-risk scenarios using the model predictive control method includes: With the control objectives of minimizing overall risk and minimizing the power adjustment costs of each station, an objective function is constructed, with power balance constraints, voltage constraints, and control quantity constraints as constraints. At the current moment, based on the future risk prediction sequence provided in step four, the comprehensive risk confidence provided in step three, the output deviation characteristics provided in step one, and the control accuracy constraints provided in step two, the above-mentioned constrained optimization problem is solved online to obtain the optimal adjustment sequence of active power and reactive power of each new energy power station in the prediction time domain. The optimal control command at the current moment is taken as the actual execution command, and the actual execution command is sent to each new energy power station for execution.

[0044] Specifically, it should be noted that the objective function expression is: , In the formula, N y To predict the number of time-domain steps, R pre (t) represents the predicted comprehensive risk confidence value at each time point within the prediction time domain, ΔP i (t) represents the active power adjustment of the i-th power station, w R Here, w represents the risk item weighting coefficient, n is the total number of new energy power stations, and w is the weighting coefficient. i The weighting coefficients for the control costs of each station are denoted by . The objective function aims to minimize the cumulative risk level in the prediction time domain, while minimizing the power adjustment costs of each station.

[0045] Specifically, it should be noted that the power balance constraint means that the sum of the active power adjustment of all stations equals the total power deficit of the system, maintaining the power generation and consumption balance of the system; the voltage constraint means that the voltage amplitude of each key node must be maintained within the normal operating range of the rated value; and the control quantity constraint means that the adjustment range of active and reactive power of each station must not exceed the upper limit of the adjustable capacity of the station.

[0046] It should be noted that the preset values, thresholds, weighting coefficients, and constant coefficients used without specific calculation processes are all selected based on actual needs, and no specific value restrictions are imposed here.

[0047] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, and is not intended to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of the present invention do not depart from the essence and scope of the technical solution of the present invention.

Claims

1. A high-precision early warning method for new energy power plants, characterized in that, Includes the following steps: Step 1: Collect historical operating data and meteorological data of each new energy power station. Based on the collected data, classify the operating conditions. For each type of operating condition, establish a piecewise exponential distribution model of the output deviation and calculate the standard deviation of the output deviation. Then, establish a spatial correlation model and a spatial attenuation model of the output deviation of multiple power stations. Step 2: Establish a coupling model between AGC control accuracy and system frequency stability, a coupling relationship model between AVC control accuracy and voltage stability of grid-connected units, a voltage support boundary model of grid-connected units, and a quantitative model of the impact of primary frequency regulation control accuracy. Quantify the impact of AGC, AVC, and primary frequency regulation control errors on the safety of system frequency, voltage, and power balance. Step 3: Construct a multi-source risk feature vector, use DS evidence theory to fuse the multi-source risk feature vector, assign a basic probability allocation function to each feature, calculate the conflict coefficient between each feature, and then obtain the comprehensive risk confidence based on the evidence synthesis rules. Step 4: Based on the comprehensive risk reliability, calculate the meteorological severity index, the power loss coefficient under typhoon weather, and the power prediction error increase factor during typhoons to update the comprehensive risk reliability. Then, perform risk evolution trend analysis to obtain the risk evolution rate. Based on the comprehensive risk reliability and risk evolution rate at the current moment, construct a future risk prediction model and output the comprehensive risk reliability prediction value at the future prediction moment. Step 5: Construct a three-tiered early warning indicator system including system-level, station-level, and equipment-level indicators; set early warning activation criteria; trigger an early warning decision when any early warning activation criterion is met; determine the early warning level based on the comprehensive risk confidence prediction value of the future prediction time output in Step 4, according to the four-level risk classification strategy; and output the early warning level and corresponding response strategy. Step 6: Receive the warning level and corresponding response strategy. When the warning level is level 3 or above, generate and execute the station collaborative control command for high-risk scenarios based on the model predictive control method.

2. The high-precision early warning method for new energy power plants according to claim 1, characterized in that, The historical operating data includes the active power, reactive power, voltage amplitude, frequency, unit start-up and shutdown status, AGC / AVC control commands and response values, and primary frequency regulation action records of each new energy power station; The meteorological data includes wind speed, light intensity, ambient temperature, humidity, air pressure, rainfall, and extreme weather indicator information; The operating conditions include normal operating conditions, low wind speed operating conditions, high wind speed operating conditions, and extreme weather operating conditions.

3. The high-precision early warning method for new energy power plants according to claim 2, characterized in that, The piecewise exponential distribution model of the output deviation is as follows: ， In the formula, ΔP, λ1, e, μ1, λ2, and μ2 are, respectively, the output deviation, the exponential distribution rate parameter of the first interval, the natural constant, the position parameter of the first interval, the exponential distribution rate parameter of the second interval, and the position parameter of the second interval; The output deviation standard deviation is the statistical value of the standard deviation of samples within each type of working condition after K-means clustering to classify the working condition categories; The spatial correlation model of the output deviation of the multiple stations uses the Copula function to describe the joint distribution. The spatial decay model is based on the Kendall rank correlation coefficient.

4. The high-precision early warning method for new energy power plants according to claim 1, characterized in that, The establishment of the coupling model between AGC control accuracy and system frequency stability includes: Calculate the difference between the system load disturbance and the sum of the AGC control errors of all AGC-participating stations; The reciprocal of the frequency response transfer function, which is composed of the equivalent inertia of the calculation system, the load damping coefficient, and the gain and time constant of the AGC control loop of each station; Multiplying the difference by the reciprocal yields the system frequency deviation; The establishment of the coupling relationship model between AVC control accuracy and voltage stability of grid-connected generator units includes: calculating the ratio of system short-circuit reactance to rated voltage, and multiplying the ratio by reactive power control error to approximate the system voltage deviation; The establishment of the voltage support boundary model for grid-type generating units includes: Calculate the sum of the rated internal potential and the internal potential control deviation; The active-reactive power correction term is calculated based on the equivalent reactance of the grid-type unit, the active power output of the unit, the rated internal potential of the unit, and the reactive power output of the unit. Subtracting the active-reactive power output correction term from the sum yields the critical support voltage of the grid-type unit. The quantitative model for the impact of primary frequency modulation control accuracy includes: Calculate the product of the rated adjustment rate and the actual adjustment rate; Calculate the ratio of the negative descent rate control error to the product; The primary frequency regulation power response deviation is obtained by multiplying the ratio by the frequency deviation and the rated capacity of the unit.

5. A high-precision early warning method for new energy power plants according to claim 4, characterized in that, The multi-source risk feature vector includes electrical risk features, meteorological risk features, equipment risk features, and geographical risk features; the electrical risk features include system frequency deviation, system voltage deviation, and power fluctuation rate. The meteorological risk characteristics include wind speed change rate, extreme temperature values, and disaster warning level; The generation of the equipment risk characteristics includes: For each piece of equipment in the station, the normalized deviation index of each state quantity is weighted and summed with the corresponding weight to obtain the risk characteristic value of a single piece of equipment; The weighting coefficients are determined based on the sensitivity of each state variable to the health status of the device. The state variable with higher sensitivity has a larger weight, and the sum of the weights of each state variable within the same device is 1. The maximum value of the risk characteristic of a single piece of equipment among all the equipment in the site is taken as the equipment risk characteristic of the site.

6. A high-precision early warning method for new energy power plants according to claim 5, characterized in that, The generation of the geographic risk features includes: For the target new energy power station area, based on the elevation standard deviation and slope change rate of the station area, the influence of topographic undulation on airflow / solar radiation distribution is quantified to obtain the topographic complexity factor; based on the land cover type of the station area, the influence of underlying surface roughness on wind speed attenuation or solar radiation reflection is quantified to obtain the underlying surface type factor; based on the statistical frequency of extreme meteorological events such as typhoons, thunderstorms, and freezing in the station area within a preset historical period, the historical risk level of meteorological disasters in the area is quantified to obtain the meteorological disaster historical frequency factor; based on indicators such as the number of transmission lines at the station grid connection point, line length, proportion of transmission lines erected on the same tower, and N-1 throughput rate, the structural strength and redundancy of the power transmission channel of the station are quantified to obtain the power grid structural vulnerability factor. Each geographical feature factor is normalized to the [0,1] interval, and weight coefficients are assigned according to the significance of each factor's impact on the station's operational risk. The sum of the weights equals 1. The geographical risk characteristic index of the station is obtained by weighting and summing the normalized geographical characteristic factors with their corresponding weights.

7. A high-precision early warning method for new energy power plants according to claim 5, characterized in that, The generation of the power volatility includes: After K-means clustering to classify the working conditions, the standard deviation of the samples within each working condition is used as the basic power fluctuation intensity of the target station under the corresponding working condition. Based on the joint distribution of multiple power stations described by the Copula function and the Kendall spatial attenuation model, the spatial correlation coefficient of the output deviation between the target power station and all surrounding power stations is calculated. Based on the correlation coefficient, the base power fluctuation intensity of the target power station is spatially superimposed and corrected to obtain the corrected fluctuation intensity taking into account the spatial correlation of multiple power stations. Within a preset evaluation time window, the spatially corrected instantaneous fluctuation intensity is time-weighted and averaged to obtain the power fluctuation rate within that time window.

8. A high-precision early warning method for new energy power plants according to claim 5, characterized in that, The revised comprehensive risk confidence includes: Real-time meteorological data is collected, and the measured values ​​of each meteorological parameter are converted into dimensionless values ​​in the interval [0,1] according to the preset severity mapping function. Weights are assigned according to the significance of the impact of each meteorological parameter on the output of the new energy power station, and the meteorological severity index is obtained by weighted summation. For typhoon weather, the rated power curve of the station units is compared with the actual operating power curve. The cumulative energy loss during the typhoon is calculated by integration, and the turbulence intensity factor is introduced to correct the loss value to obtain the output loss coefficient under typhoon weather. Using the measured maximum wind speed as the core input, the power prediction error is increased by a factor of 100% by the fitting coefficients of historical typhoon events. The meteorological risk correction coefficient is obtained by weighting and aggregating the meteorological severity index, the power loss coefficient, and the power prediction error amplification factor. The meteorological risk characteristics at the current moment output in step three are corrected using the meteorological risk correction coefficient, and the meteorological-corrected comprehensive risk confidence at the current moment is obtained after DS evidence fusion.

9. A high-precision early warning method for new energy power plants according to claim 8, characterized in that, The risk evolution trend analysis includes: Subtract the comprehensive risk confidence score of the previous time from the comprehensive risk confidence score of the current time after weather correction, and then divide by the sampling time interval to obtain the risk evolution rate of the current time. When the risk evolution rate is greater than 0, the risk is judged to be increasing; when the risk evolution rate is equal to 0, the risk is judged to be in a stable state; when the risk evolution rate is less than 0, the risk is judged to be decreasing. The method of constructing a future risk prediction model based on the comprehensive risk confidence and risk evolution rate at the current moment, and outputting the comprehensive risk confidence prediction value at the future moment, includes: adding the risk value at the current moment to the multiplication value of the prediction time window and the risk evolution rate at the current moment, to obtain the comprehensive risk confidence prediction value at the future prediction moment.

10. A high-precision early warning method for new energy power plants according to claim 9, characterized in that, The generation of station collaborative control commands for high-risk scenarios based on model predictive control methods includes: With the control objectives of minimizing overall risk and minimizing the power adjustment costs of each station, an objective function is constructed, with power balance constraints, voltage constraints, and control quantity constraints as constraints. At the current moment, based on the future risk prediction sequence provided in step four, the comprehensive risk confidence provided in step three, the output deviation characteristics provided in step one, and the control accuracy constraints provided in step two, the above-mentioned constrained optimization problem is solved online to obtain the optimal adjustment sequence of active power and reactive power of each new energy power station in the prediction time domain. The optimal control command at the current moment is taken as the actual execution command, and the actual execution command is sent to each new energy power station for execution.

Images