Search

CN-121983969-A - Hydrogen energy micro-grid optimal scheduling method and system for short-term prediction uncertainty

CN121983969ACN 121983969 ACN121983969 ACN 121983969ACN-121983969-A

Abstract

The invention relates to a method and a system for optimizing and dispatching a hydrogen energy micro-grid with short-term prediction uncertainty, wherein the method comprises the following steps of S1, collecting time sequence meteorological data and power states of units of a micro-grid site in real time, carrying out cleaning, denoising and normalization preprocessing on the collected data, S2, obtaining a power point prediction sequence in a future prediction time domain based on a trained long-short-term memory network LSTM model, collecting historical residual samples, and establishing a non-parameter probability density model by utilizing a kernel density estimation KDE method, S3, optimizing and deciding stage S31, constructing a random optimizing model which aims at the minimum total running cost of the system and comprises equipment fading cost and power balance opportunity constraint, S32, converting the power balance opportunity constraint into deterministic linear inequality constraint, and S4, executing and feeding back stage. Compared with the prior art, the method and the device have the advantages that the accuracy of uncertainty quantification is improved, economical operation of the system is ensured, and the calculation complexity of random optimization is remarkably reduced.

Inventors

  • HOU ZHIPENG
  • CHEN FENGXIANG
  • GUO YAFENG

Assignees

  • 同济大学

Dates

Publication Date
20260505
Application Date
20260403

Claims (10)

  1. 1. The hydrogen energy micro-grid optimal scheduling method for short-term prediction uncertainty is characterized by comprising the following steps of: S1, data acquisition and preprocessing, namely acquiring time sequence meteorological data of a micro-grid site in real time and power states of all units; S2, a prediction and uncertainty quantization stage, namely acquiring a power point prediction sequence in a future prediction time domain based on a trained long-short-term memory network LSTM model, acquiring a historical residual error sample, and establishing a non-parameter probability density model by using a kernel density estimation KDE method to acquire error probability distribution of the power point prediction; s3, optimizing a decision stage: S31, constructing a random optimization model which aims at the minimum total running cost of the system and contains equipment fading cost and power balance opportunity constraint, wherein when the random optimization model is constructed, a total objective function is constructed Overall modeling includes the degradation costs caused by start-up and shut-down of the electrolyzer and fuel cell, and power fluctuations And cost of electricity Cost of electricity purchase and sale And a safety cost ; S32, opportunity constraint conversion, namely converting the power balance opportunity constraint into deterministic linear inequality constraint through a risk self-adaptive mechanism and quantile mapping, wherein the allowable violation probability which dynamically increases along with the prediction step length t is calculated Converting the power balance probability constraint into a deterministic linear inequality constraint defined by quantiles by utilizing the inverse cumulative probability distribution function mapping relation of the KDE; And S4, performing a rolling solution on the converted deterministic optimization model, issuing a first optimal control instruction obtained by the solution to each unit controller of the hydrogen energy micro-grid for execution, and returning to the step S2 for closed-loop rolling optimization according to the real-time running state feedback of the system.
  2. 2. The hydrogen energy micro-grid optimal scheduling method for short-term prediction uncertainty according to claim 1, wherein in S1, the preprocessing comprises the steps of eliminating abnormal values through a threshold value judging method, complementing missing data through a linear interpolation method, and eliminating magnitude differences among different dimension data through a normalization algorithm.
  3. 3. The method for optimizing and scheduling a hydrogen energy micro-grid for short-term prediction uncertainty as set forth in claim 1, wherein in S1, the time-series meteorological data comprises irradiance and ambient temperature, and the unit power state comprises solar panel output power, load power, direct current bus voltage, battery energy storage and hydrogen energy system power.
  4. 4. The hydrogen energy micro-grid optimal scheduling method of short-term prediction uncertainty according to claim 1, wherein in S2, a non-parametric probability density model is established by using a kernel density estimation KDE method, and specifically, a probability density function of a prediction error is calculated by the following formula: point prediction sequence for obtaining future prediction step length based on trained long-short-term memory network LSTM Meanwhile, a historical residual error sample is collected, and a non-parameter probability density model is established by using a Kernel Density Estimation (KDE) method: Wherein, the Representing prediction error Probability density estimation functions of (2); a total number of samples representing the historical prediction error; a parameter representing bandwidth for controlling the degree of smoothness of the probability density curve; represented as gaussian kernel functions; represent the first Observing sample values by historical prediction errors; And acquiring non-stationary probability distribution characteristics reflecting the fluctuation difference of the prediction error in different time periods by carrying out subsection statistics on the 24-hour error samples.
  5. 5. The hydrogen energy micro-grid optimal scheduling method for short-term prediction uncertainty as claimed in claim 1, wherein in S31, the total objective function is modeled in an overall way The expression is as follows: ; The decay cost Voltage loss caused by modeling operation time length, start-stop switching frequency and power fluctuation amplitude is comprehensively modeled for an electrolytic tank and a fuel cell, loss quantification is carried out for a storage battery and a hydrogen storage system based on charge-discharge and charge-discharge cycle times, and electricity cost is reduced The renewable energy source power generation cost consumed by all paths meeting the load demand, wherein the paths comprise direct supply, battery energy storage and hydrogen energy storage conversion, and the electricity purchasing and selling cost The method refers to calculating direct economic loss or income generated by electricity purchasing and selling behaviors between a micro-grid and a main grid according to time-sharing electricity price, and safety cost Refers to the security cost of a system state violating a constraint.
  6. 6. The method for optimizing and scheduling a hydrogen energy micro-grid with short-term prediction uncertainty as set forth in claim 1, wherein in S32, the risk adaptive mechanism dynamically adjusts the allowable violation probability by the standard deviation of the prediction error distribution The formula is as follows: ; wherein, alpha min and alpha max are the preset minimum and maximum violation probabilities, sigma max is the maximum value of standard deviation, The standard deviation of the prediction error distribution is obtained by calculating the prediction error distribution obtained by KDE.
  7. 7. The hydrogen energy micro grid optimizing and scheduling method of short term predictive uncertainty as set forth in claim 1, further comprising utilizing variables in S32 The confidence level is allocated to a probability interval q L ,q U corresponding to the high score and the low score, and the calculation formula is as follows: ; Wherein, the The value of (2) depends on the bias of the prediction error distribution.
  8. 8. The method for optimizing and scheduling hydrogen energy micro-grid with short-term predictive uncertainty as set forth in claim 7, wherein in S32, the probability constraint of power balance is converted into a high-low quantile by using the inverse cumulative probability distribution function mapping relation of KDE And The defined deterministic linear inequality constraint, thereby translating the opportunistic constraint into a deterministic constraint.
  9. 9. The hydrogen energy micro-grid optimizing scheduling method for short-term predictive uncertainty as set forth in claim 8, wherein in S32, the specific operation of converting the opportunity constraint into the deterministic constraint is to pass through an inverse cumulative probability distribution function of a KDE probability density model The probability constraint of power balancing is converted into a deterministic linear inequality constraint: 。
  10. 10. A system for implementing the short-term predictive uncertainty hydrogen energy microgrid optimization scheduling method according to any one of claims 1-9, comprising: The data acquisition preprocessing module is used for executing data cleaning, complement and normalization; The prediction and uncertainty quantitative analysis module is used for generating a power point prediction sequence by utilizing an LSTM model and constructing a prediction error probability density model based on a KDE method; The constraint reconstruction and conversion module is used for dynamically determining the permissible violation probability according to the prediction step length and converting the power balance opportunity constraint into a deterministic equivalent constraint by utilizing the quantile of the probability density model; the comprehensive target scheduling solving module is used for constructing and solving a total target function comprising the decay cost, the electricity consumption cost, the electricity purchase and sale cost and the safety cost and outputting an optimal control instruction sequence; And the instruction execution and feedback module is used for executing rolling optimization scheduling and receiving system state feedback.

Description

Hydrogen energy micro-grid optimal scheduling method and system for short-term prediction uncertainty Technical Field The invention relates to the technical field of micro-grid operation control and energy management, in particular to a hydrogen energy micro-grid optimal scheduling method and system for short-term prediction uncertainty. Background With the continuous improvement of the permeability of renewable energy sources such as wind power, photovoltaic and the like, the inherent intermittence and randomness of the renewable energy sources form a serious challenge for the frequency stability and peak shaving capacity of a main power grid. Integrating a distributed power supply, an energy storage system and a local load and accessing a main network in a micro-grid mode has become an important direction for improving energy consumption capability. CN202510928944.3 discloses an extremely high Wen Changjing multi-energy complementary optimization scheduling method considering uncertainty, which mainly comprises the steps of 1, fusing a two-way time convolution, a two-way long-short-term memory network, a two-way short-term memory network, a concentration mechanism and a split regression forest, realizing high-precision prediction and uncertainty modeling of wind speed, solar irradiation and load, constructing a typical day scene set, 2, constructing a dual-stage scheduling model fusing epsilon-constraint multi-objective optimization and opportunistic constraint mixed integer programming, optimizing adjustment margin in the day and rolling a scheduling path in the day, 3, providing three types of physical correction mechanisms of wind power air density correction, photovoltaic temperature response and water-electricity evaporation-water level coupling aiming at extremely high Wen Raodong, 4, integrating and constructing a prediction-optimization-feedback-correction closed loop flow, and improving stability and response toughness of the system in extreme weather. However, the prediction precision is limited, the physical response model is absent, the scheduling cooperativity is poor, and the capability of the optimization method for processing complex uncertainty is insufficient. In a micro-grid Energy Management System (EMS), a scheduling strategy is formulated by integrating grid electricity price, load demand, renewable energy source output and energy storage cost. However, the prior art suffers from the following disadvantages: 1 uncertainty modeling is inaccurate, namely the existing scheduling method excessively depends on a point prediction result, and it is difficult to accurately describe prediction errors of Non-Gaussian distribution. 2. The equipment loss consideration is lost, namely the fine operation cost of key hydrogen energy components such as an electrolytic tank, a fuel cell and the like under frequent start-stop and power fluctuation is ignored, so that the scheduling scheme is difficult to consider the long-term operation reliability and the economy. Disclosure of Invention The invention aims to overcome the defects of the prior art and provide the hydrogen energy micro-grid optimal scheduling method and system for short-term prediction uncertainty, which improve the accuracy of uncertainty quantification, ensure the economic operation of the system and remarkably reduce the computational complexity of random optimization. The aim of the invention can be achieved by the following technical scheme: the invention provides a hydrogen energy micro-grid optimal scheduling method for short-term prediction uncertainty, which comprises the following steps of: S1, data acquisition and preprocessing, namely acquiring time sequence meteorological data of a micro-grid site in real time and power states of all units; S2, a prediction and uncertainty quantization stage, namely acquiring a power point prediction sequence in a future prediction time domain based on a trained long-short-term memory network LSTM model, acquiring a historical residual error sample, and establishing a non-parameter probability density model by using a kernel density estimation KDE method to acquire error probability distribution of the power point prediction; s3, optimizing a decision stage: S31, constructing a random optimization model which aims at the minimum total running cost of the system and contains equipment fading cost and power balance opportunity constraint, wherein when the random optimization model is constructed, a total objective function is constructed Overall modeling includes the degradation costs caused by start-up and shut-down of the electrolyzer and fuel cell, and power fluctuationsAnd cost of electricityCost of electricity purchase and saleAnd a safety cost; S32, opportunity constraint conversion, namely converting the power balance opportunity constraint into deterministic linear inequality constraint through a risk self-adaptive mechanism and quantile mapping, wherein the allowable violation proba