Search

CN-122026314-A - Operation optimization method considering virtual power plant resource confidence interval

CN122026314ACN 122026314 ACN122026314 ACN 122026314ACN-122026314-A

Abstract

The invention relates to an operation optimization method considering a virtual power plant resource confidence interval, which comprises the following steps of 1, constructing a distributed new energy output probability model, quantifying output fluctuation ranges of resources under different confidence levels, obtaining comprehensive output probability interval estimation, 2, selecting a proper calculation method from minkowski and theory and related optimization strategies by combining physical constraints of internal resources of a virtual power plant, determining a probability feasible domain of the virtual power plant, 3, comprehensively considering power grid scheduling requirements, resource adjustment cost and operation safety constraints, establishing a multi-objective operation optimization model, solving the model through an improved genetic algorithm, and outputting real-time collaborative operation strategies of all resources of the virtual power plant, thereby realizing high-precision operation optimization in an uncertainty environment. The method and the device can improve the accuracy, dynamic adaptability and engineering practicability of the operation optimization of the virtual power plant.

Inventors

  • WANG XIAOXUE
  • LIU JINTAO
  • LIU YIXIN
  • CHENG JIUDONG
  • Sun Miaohe

Assignees

  • 河北工业大学
  • 天津大学
  • 天津天电瑞联能源科技有限公司

Dates

Publication Date
20260512
Application Date
20251231

Claims (10)

  1. 1. An operation optimization method considering a virtual power plant resource confidence interval is characterized by comprising the following steps: The method comprises the steps of 1, constructing a distributed new energy output probability model by adopting a mode of combining Bootstrap resampling, ensemble learning, quantile regression and conformal quantile regression, quantifying output fluctuation ranges of resources under different confidence levels, obtaining comprehensive output probability interval estimation, and providing a precise uncertainty boundary for subsequent operation optimization; Step 2, selecting a proper calculation method from Minkowski and theory and related optimization strategies by combining physical constraint of internal resources of the virtual power plant on the basis of the comprehensive output probability interval estimation of the distributed new energy obtained in the step 1, and determining a probability feasible region of the virtual power plant; And 3, on the basis of the probability feasible region of the virtual power plant obtained in the step 2, comprehensively considering the power grid scheduling requirement, the resource adjusting cost and the operation safety constraint, establishing a multi-objective operation optimization model, solving the model through an improved genetic algorithm, outputting a real-time cooperative operation strategy of each resource of the virtual power plant, and realizing high-precision operation optimization in an uncertainty environment.
  2. 2. The method for optimizing operations in consideration of virtual power plant resource confidence intervals as set forth in claim 1, wherein the specific steps of step 1 include: (1) Performing data preprocessing, collecting historical output data of internal resources of the virtual power plant, removing abnormal values and missing data, and performing normalization processing on the data; (2) Based on the preprocessed data obtained in the step (1), constructing a probability model of distributed resource output by using an integrated learning and quantile regression method, and calculating a result of a distributed resource output probability interval; (3) Based on the result of the distributed resource output probability interval calculated in the step (2), the result of all training data sets is aggregated through an integrated learning method to obtain a resource output probability interval of a total sample, and the upper and lower boundaries of the probability interval are corrected through introducing a conformal quantile regression method, wherein the deviation rectifying amplitude of the upper and lower boundaries can be determined according to the mode shown in a formula (4): ; Wherein, the Indicating that at a quantile level of 1-alpha/2, the pair Phi (#) is a corresponding fractional number calculation function; Representing the upper boundary order Column or lower boundary sequences ; The modified probability interval can be represented by formulas (5) - (6), so that the output probability interval estimation of the distributed new energy integration is obtained: ; ; wherein the upper and lower boundaries of the corrected section are respectively recorded as And And, correspondingly, the number of the parts of the device, And (3) with Representation is directed to And (3) with Is described.
  3. 3. The method for optimizing operation of claim 2 wherein step 1, step (2), comprises the steps of: ① Sampling the historical data by a Bootstrap method in a put-back way, and dividing the historical data into a plurality of training data sets; ② For each training data set, calculating a probability interval of resource output by using a base learner and a quantile regression method; In the model training process, if the input variable comes from the distribution X, the quantile regression model is shown in the formula (1): ; In this setting, y is the output value of the model, α represents the selected quantile, for the distributed new energy output, the probability that its actual power lies within the corresponding confidence interval can be represented as (1- α), the symbol F Y (y) is used to represent the cumulative distribution function of the random variable y, and the corresponding probability interval can be determined according to the upper and lower quantiles corresponding to the given quantile α, specifically represented by formula (2): ; Here, note that And (3) with Respectively representing a quantile level alpha/2 and a 1-alpha/2, which are estimated values obtained by a quantile regression method based on historical sample data of distributed new energy output; to obtain this result, an optimization problem in the form of equation (3) needs to be solved: 。
  4. 4. the method for optimizing operations taking into account the confidence interval of virtual power plant resources according to claim 1, wherein said step 2 comprises the following steps: (1) Modeling various distributed resources in the virtual power plant on the basis of the comprehensive output probability interval estimation of the distributed new energy obtained in the step 1, and describing the output characteristics, physical constraint and adjustment capacity of the distributed resources; (2) Based on various distributed resource models in the virtual power plant constructed in the step (1), physical constraints of internal resources of the virtual power plant are obtained, the output probability interval estimation of the distributed new energy synthesis obtained in the step (1) is combined, and a proper optimization algorithm is selected from an optimization method to solve the probability feasible region of the virtual power plant.
  5. 5. The method for optimizing operations in consideration of confidence intervals of resources of a virtual power plant as set forth in claim 4, wherein said step 2 (1) models various distributed resources in the virtual power plant, and said modeling comprises: ① A distributed photovoltaic model, as follows: ; wherein: -the radiation intensity of solar energy (W/m 2 ); 、 Obtaining the expected value of the illumination radiation intensity based on the historical data of the illumination radiation intensity Sum of variances As shown in formulas (8), (9): ; ; Intensity of solar radiation Is shown by the formula (10): ; wherein: max -upper limit value of solar radiation intensity (W/m 2 ); min -lower limit of solar radiation intensity (W/m 2 ), the output function of photovoltaic power generation is represented by formula (11): ; wherein: -photovoltaic power generation conversion efficiency; -photovoltaic panel active area (m 2 ); -solar radiation intensity at moment (W/m 2 ); the constraint conditions to be satisfied by photovoltaic power generation are shown in the following formulas (12) and (13): the output constraint of the photovoltaic generator set is shown in a formula (12): ; wherein: the actual output (MW) of the photovoltaic unit i during period t, -Maximum output (MW) of the photovoltaic unit i during period t; the photovoltaic generator set light rejection constraint is shown in formula (13): ; wherein: -maximum light rejection rate; ② A distributed fan model, as follows: the probability density function is as follows as in equation (14): ; Wherein f (v) -wind speed probability distribution density, v-actual wind speed (m/s), k-shape parameters, c-scale parameters; When determining the expected value and variance of wind speed, the values of k and c can be derived by a moment estimation method as shown in equations (15), (16): ; ; wherein: -expected value of wind speed, i.e. wind speed mean (m/s); -standard deviation of wind speed (m/s); -Gamma function; The output power formula is shown as (17): ; wherein: -rated power (MW) of the wind turbine; -cut-in wind speed (m/s); -cut-out wind speed (m/s); -rated wind speed (m/s); The constraint conditions to be satisfied by wind power generation are shown in the formula (18) and the formula (19): The output constraint of the wind generating set is shown in a formula (18): ; wherein: actual output (MW) of wind turbine k during period t, -Maximum output (MW) of wind turbine k during period t; the wind curtailment constraint of the wind generating set is shown in a formula (19): ; wherein: -maximum wind reject rate; ③ The gas turbine unit model is as follows: The actual force G t is shown in formula (20): ; wherein: -micro gas turbine power generation conversion efficiency; -natural gas heating value of (34.12 MJ/m 3); -using the volume of natural gas (m 3) at time t, -unit conversion constant of the D-time period length; the constraints that should be satisfied by the gas turbine operation are shown in formulas (21) to (22): the gas turbine output constraint is shown in equation (21): ; wherein: -lower limit of the gas turbine output (MW); -upper limit of gas turbine output (MW); -the start-stop state variable of the gas turbine at time t is a 0-1 variable, 1 representing operation, 0 representing shutdown; the hill climbing constraint of the gas turbine is shown in formula (22): ; wherein: the output of the gas turbine during the period t, The output of the gas turbine during the period t-1, The upper limit of the climbing (MW) of the gas turbine, -Lower limit of the climb (MW) of the gas turbine; the start-stop constraints of the gas turbine are shown in formula (23) and formula (24): ; ; wherein: The gas turbine operating time at time t, The gas turbine shutdown time at time t, The minimum start-up time of the gas turbine, -Gas turbine minimum down time; ④ The electric automobile model is as follows: The electric vehicle charge-discharge cost formula of the period t is shown by (25): ; wherein: -the number of dispatchable electric vehicles (vehicles) of the virtual power plant; v—index number of electric car; Discharge cost (yuan/MWh) of the electric vehicle in period t; Charging cost (yuan/MWh) of the electric vehicle during period t; -a period tv discharge power (MW) of the electric vehicle; -charging power (MW) of the v-th electric vehicle during time period t; -electric vehicle discharge conversion efficiency; -electric car charge conversion efficiency; the electric vehicle operation constraints are shown in formulas (26) - (28): ; ; ; wherein: Lower limit value of SOC of v-th electric automobile, Upper limit value of SOC of v-th electric automobile, Maximum allowable charging power (MW) of the electric vehicle, -Maximum allowable discharge power (MW) of the electric vehicle; ⑤ The energy storage model is as follows: ; ; ; The energy storage system output is shown by formulas (32), (33): ; ; wherein: — the state of charge of the energy storage battery at the moment; -the rated capacity (MWh) of the energy storage battery; 、 -charge-discharge energy conversion efficiency of the energy storage battery; 、 Energy storage battery Charge-discharge power (MW) at time; The constraints on the operation of the energy storage facility should satisfy formulas (34) - (38): ; ; ; ; ; wherein: A variable 0-1 for indicating the charge and discharge state of the energy storage cell, The energy storage system is in a charged state when =1, The energy storage system is in a discharge state when=0.
  6. 6. The method for optimizing operation of said virtual power plant resource confidence interval of claim 4, wherein said step 2 (2) comprises the steps of: Firstly, respectively calculating the virtual power plant probability feasible outer boundary corresponding to the distributed new energy output upper boundary by adopting a Minkowski method and an optimization method, verifying the consistency of two groups of calculation results, if the two groups of calculation results are consistent, completing the solution of the inner boundary by adopting the Minkowski method and the optimization method, and if the results are different, adopting the optimization method to carry out the calculation work of the virtual power plant probability feasible inner boundary.
  7. 7. The method of optimizing operations of claim 4 wherein step (2) is followed by the steps of: (3) Based on the confidence coefficient of each distributed new energy output probability interval, the confidence coefficient of a single resource is aggregated into the comprehensive confidence level of the whole feasible region through a probability propagation algorithm, and the confidence level of the probability feasible region of the virtual power plant is quantitatively evaluated, as shown in a formula (39): ; The parameters in the formula are defined as follows, alpha i represents the quantile of the output probability interval of the ith distributed new energy source, 1-alpha i represents the corresponding confidence coefficient of the interval, and m represents the total number of the distributed new energy sources aggregated by the virtual power plant; And finally, taking beta as a dependent variable, measuring the comprehensive confidence level of the probability feasible region of the whole virtual power plant, and providing a quantitative basis for the reliability evaluation of the multi-resource collaborative operation strategy.
  8. 8. The method for optimizing operations taking into account the confidence interval of virtual power plant resources according to claim 1, wherein said step 3 comprises the following steps: (1) Firstly, analyzing the power grid demand and the operation target; (2) Establishing a multi-objective optimization model in an operation optimization stage based on the power grid demand and the operation objective analysis result in the step (1); (3) And (3) carrying out model solving and collaborative operation strategy outputting based on the multi-objective optimization model constructed in the step (2).
  9. 9. The method for optimizing operations in consideration of virtual power plant resource confidence intervals as set forth in claim 8, wherein the step 3 (2) comprises the steps of: ① Establishing an optimization objective function as shown in formula (40): ; Where T is the total number of time periods in the optimization cycle, The actual total output of the virtual power plant during period t, -Grid reference power for period t, The weighting coefficient of the period t, reflecting the power tracking requirements of the grid for this period, The resources of period t regulate the physical loss, Run risk metric (risk of deviation based on probability feasible domain), The risk aversion coefficient is set by the virtual power plant according to the stability requirement of the power grid, and the running risk measure is calculated specifically as shown in a formula (41): ; wherein: -characterizing a desired level of resource contribution for a central value of a probability feasible region, the specific calculation being as shown in formula (42): ; ② Constraint condition setting, namely setting resource constraint and technical rule constraint, wherein the resource constraint is detailed in each distributed resource modeling part of the virtual power plant in the step 2, and the technical rule constraint is shown in formulas (43) - (45): minimum/maximum force constraint: ; Force change rate constraint (climbing rate constraint): ; Probability feasible region constraint: ; wherein: -a minimum total output (MW) allowed by the virtual power plant; -maximum total output (MW) allowed by the virtual power plant; -maximum ramp rate (MW/h) of the power output of the virtual power plant during adjacent periods.
  10. 10. The method for optimizing operations in consideration of virtual power plant resource confidence intervals as recited in claim 8, wherein the step 3, step (3), comprises the steps of: ① The improved Genetic Algorithm (GA) is adopted for solving, and the specific steps are as follows: Initializing, namely generating an initial population, wherein each individual represents an operation strategy, and the population size is N=100; Fitness calculation, namely calculating the fitness value of each individual according to an objective function, wherein the fitness function is shown as a formula (46): ; selecting, namely selecting individuals with high fitness to enter the next generation by adopting a roulette selection method; the crossover operation, which is to adopt a single-point crossover method, wherein the crossover probability p_c=0.8, and the crossover point selection mode is to randomly select a period of time as the crossover point; the variation operation, namely, the individual is subjected to variation according to variation probability p_m=0.1, and the variation mode is that a resource output plan is randomly adjusted in a probability feasible domain; termination condition that maximum number of iterations is reached Or the fitness value meets a preset threshold; ② Optimizing the multi-resource collaborative operation strategy, namely optimizing the collaborative operation strategy of each resource in the virtual power plant by an improved genetic algorithm and combining a probability feasible region and resource physical constraint; ③ In order to adapt to the change of the power grid demand, the dynamic adjustment strategy comprises the following specific contents: Updating real-time data every other time Updating the resource confidence interval and the power grid reference power information, wherein the updating frequency is 1 hour; local optimization, namely, based on the current time period information, carrying out local optimization on the operation strategy of the subsequent time period, wherein the optimization range is k time periods k=6 in the future; Policy updating, namely updating the optimized operation policy to the virtual power plant energy management system, wherein updating logic is shown as a formula (47): ; Wherein the method comprises the steps of To adjust the threshold, a policy smooth transition is ensured.

Description

Operation optimization method considering virtual power plant resource confidence interval Technical Field The invention belongs to the technical field of virtual power plant operation optimization, and particularly relates to an operation optimization method considering a virtual power plant resource confidence interval. Background Virtual power plants (Virtual Power Plant, VPP) are virtual energy systems that aggregate and coordinate optimization of distributed resources (e.g., renewable energy, energy storage systems, demand response resources, etc.) through advanced information communication technologies and intelligent control technologies. The virtual power plant does not actually own a physical power plant, but integrates the scattered and different types of resources into a whole in a virtualized mode, so that the virtual power plant can participate in power grid dispatching and auxiliary service provision like a traditional power plant. The virtual power plant plays an increasingly important role in the electric power market, plays an important role in promoting clean energy consumption, improving the flexibility of an electric power system, optimizing electric power market transactions, promoting energy digital transformation and the like, and is an important tool for realizing energy transformation and sustainable development in the future. In the resource aggregation and operation optimization process of Virtual Power Plants (VPPs), uncertainty and volatility of resources are core challenges. The existing resource aggregation method is mainly divided into two types, namely deterministic aggregation and uncertainty aggregation, wherein the deterministic aggregation method regards distributed resources in a virtual power plant as determined energy sources, and the output of the resources is assumed to be fixed or can be completely predicted. Although the method is simple in model, the fluctuation of the output of renewable energy sources (such as photovoltaic and wind power) and an energy storage system cannot be considered, so that a large deviation exists between a polymerization result and the actual output. The virtual power plant operation strategy is disjointed from the actual regulation capability, and the technical requirements of the power grid on the reliability and the flexibility of the resources are difficult to meet. In order to cope with the uncertainty of resources, scholars propose uncertainty aggregation methods such as scene methods, robust optimization and the like. Scenario methods describe uncertainty in resource output by generating typical scenarios (e.g., high wind speeds, low light, etc.). However, the scenario approach relies on the representativeness of historical data and suffers from insufficient coverage when dealing with multi-dimensional uncertainties (e.g., coupling of multiple factors such as weather, load, electricity prices, etc.), resulting in inaccurate run boundary delineation. The robust optimization method characterizes the fluctuation range of the resource output by constructing an uncertainty set. While robust optimization can provide some protection, it is computationally complex and difficult to apply quickly in the real-time market. Furthermore, robust optimization methods often target conservation and may limit the full exploitation of resource regulatory potential. However, the existing virtual power plant operation optimization method has the following technical defects when processing resource uncertainty: (1) The uncertainty quantization is insufficient, and the output fluctuation range of resources such as renewable energy sources cannot be precisely quantized through technical means such as confidence intervals, so that the reliability of an operation strategy is difficult to guarantee; (2) The method has poor dynamic adaptability, is difficult to dynamically adjust an optimization strategy according to the real-time resource state and the power grid demand, and cannot adapt to the rapid fluctuation characteristic of the high-proportion renewable energy power grid; (3) The defect of insufficient capacity in the aspect of multi-objective collaborative optimization is overcome, and when the multi-technical targets such as resource adjustment precision, power grid stability and operation efficiency are balanced, a systematic optimization framework is lacking, so that the comprehensive performance of an operation strategy is poor; (4) And the calculation efficiency and the precision are difficult to balance, and when the existing algorithm is used for solving the problem of high-dimensional uncertainty, the calculation speed and the optimization precision are difficult to be considered, so that the application in actual engineering is limited. Aiming at the technical problems, the invention provides an operation optimization method considering a virtual power plant resource confidence interval. No prior art publication is found, either the same or