CN-121791131-B - Short-term prediction and optimization control method for optical storage system based on model prediction control

CN121791131BCN 121791131 BCN121791131 BCN 121791131BCN-121791131-B

Abstract

The invention discloses a short-term prediction and optimization control method of an optical storage system based on model prediction control, and relates to the technical field of optical storage system control. The method comprises the steps of responding to a photovoltaic power sequence of an optical storage system, carrying out signal analysis on the photovoltaic power sequence, determining a power drop event in the photovoltaic power sequence, analyzing the evolution trend of a cloud layer system based on the drop event sequence obtained by constructing each power drop event according to time sequence, extracting cloud evolution characteristics of the cloud layer system, predicting future shielding events caused by the cloud layer system based on the cloud evolution characteristics of the cloud layer system, and predicting energy storage energy gaps of the optical storage system based on the future shielding events caused by the cloud layer system so as to optimally control the optical storage system based on the energy storage energy gaps. The invention can improve the reliability of the optimal control of the optical storage system.

Inventors

HUANG WEI
KONG LIANG
JIA JIANBO
YANG HUI

Assignees

上海联元智能科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260309

Claims (10)

1. The short-term prediction and optimization control method for the optical storage system based on model prediction control is characterized by comprising the following steps of: The method comprises the steps of responding to a photovoltaic power sequence of a photovoltaic storage system, carrying out signal analysis on the photovoltaic power sequence, and determining a power drop event in the photovoltaic power sequence, wherein the photovoltaic power sequence comprises historical photovoltaic power generation powers in historical preset time periods nearest to the current moment, and the power drop event is used for representing a time period when the historical photovoltaic power generation powers show a downward trend; analyzing the evolution trend of a cloud layer system based on a drop event sequence obtained by constructing each power drop event according to a time sequence, and extracting cloud cluster evolution characteristics of the cloud layer system; predicting future shielding events caused by the cloud layer system based on cloud cluster evolution characteristics of the cloud layer system; And predicting an energy storage energy gap of the optical storage system based on a future shielding event caused by the cloud layer system, so that the optical storage system is optimally controlled based on the energy storage energy gap.
2. The method for short-term predictive and optimal control of a light storage system based on model predictive control of claim 1, wherein said performing signal analysis on said photovoltaic power sequence to determine a power sag event in said photovoltaic power sequence comprises: Performing discrete Fourier transform on the photovoltaic power sequence to obtain a frequency domain spectral density curve of the photovoltaic power sequence; determining dominant frequency of the photovoltaic power sequence based on the frequency domain spectral density curve, wherein the dominant frequency is used for separating a cloud cluster slow-change process from a cloud cluster fast-change process; Extracting a fluctuation component sequence with the frequency greater than the dominant frequency from the frequency domain spectral density distribution; a power sag event in the photovoltaic power sequence is determined based on the sequence of fluctuating components.
3. The method for short-term predictive and optimal control of a light storage system based on model predictive control of claim 2, wherein said determining a dominant frequency of the photovoltaic power sequence based on the frequency domain spectral density curve comprises: Performing second order derivative operation on the frequency domain spectrum density curve, and determining the frequency value of each inflection point in the frequency domain spectrum density curve; And determining the inflection point frequency value with the maximum frequency domain spectral density as the dominant frequency of the photovoltaic power sequence.
4. The method for short-term predictive and optimal control of a light storage system based on model predictive control of claim 2, wherein said determining a power drop event in said photovoltaic power sequence based on said sequence of wave components comprises: Performing Hilbert transformation on the fluctuation component sequence to obtain an analysis signal corresponding to the fluctuation component sequence; The instantaneous amplitude envelope curve is used for representing the change degree of the power fluctuation intensity along with time; And carrying out first-order differential operation on the instantaneous amplitude envelope curve to determine the power drop event formed between the adjacent envelope peak time and the envelope valley time.
5. The method for short-term predictive and optimal control of an optical storage system based on model predictive control according to claim 1, wherein the step of analyzing the evolution trend of a cloud system based on a drop event sequence obtained by constructing each power drop event in time sequence and extracting the cloud evolution characteristics of the cloud system comprises the following steps: Acquiring normalized drop depth and normalized drop rate of each power drop event in the drop event sequence; Based on the time sequence among the power falling events in the falling event sequence, taking the normalized falling depth of the power falling events in the falling event sequence as a first dimensional coordinate, and taking the normalized falling rate of the power falling events in the falling event sequence as a second dimensional coordinate, constructing a two-dimensional characteristic phase space, wherein event points in the two-dimensional characteristic phase space are used for representing the physical characteristic states of one-time historical shielding events; and extracting cloud cluster evolution characteristics of the cloud layer system based on the two-dimensional characteristic phase space.
6. The model predictive control-based short-term predictive and optimal control method for an optical storage system of claim 5, wherein the cloud evolution characteristics include a depth evolution trend and a period delay of the power drop event; The extracting cloud cluster evolution features of the cloud layer system based on the two-dimensional feature phase space comprises the following steps: Extracting a target analysis period of the cloud layer system in a steady state evolution mode based on the two-dimensional characteristic phase space; Performing linear fitting on event points located in the target analysis period in the two-dimensional characteristic phase space, and determining the depth evolution trend of the power drop event; and carrying out autocorrelation analysis on the interval duration sequence of the power drop event to determine the period delay quantity of the power drop event, wherein the interval duration sequence comprises a plurality of event interval durations which are used for representing interval durations between adjacent power drop events.
7. The method for short-term predictive and optimal control of a light storage system based on model predictive control as set forth in claim 6, wherein said extracting a target analysis period of the cloud layer system in a steady state evolution mode based on the two-dimensional characteristic phase space includes: Performing variance calculation on Euclidean distances between adjacent event points in the two-dimensional characteristic phase space to obtain a global variance of the two-dimensional characteristic phase space; performing variance calculation on Euclidean distances between the adjacent event points in the two-dimensional characteristic phase space, which are positioned in a preset sliding window, to obtain local variances of the sliding window; And determining a period corresponding to the sliding window as the target analysis period in response to the local variance being smaller than a preset proportion of the global variance.
8. The model predictive control-based short-term predictive and optimal control method for an optical storage system according to claim 1, wherein the cloud evolution characteristics comprise a depth evolution trend and a period delay amount of the power drop event; The predicting future shielding events caused by the cloud layer system based on cloud cluster evolution characteristics of the cloud layer system comprises the following steps: Predicting the estimated occurrence time of the future shielding event based on the absolute value of the autocorrelation coefficient of the period delay and the waiting time, wherein the waiting time is the difference between the current time and the recovery completion time, and the recovery completion time is from the valley time of the power falling event nearest to the current time, and the envelope value of the instantaneous amplitude envelope reaches a preset fluctuation interval for the first time; predicting the estimated shielding depth of the future shielding event based on the depth evolution trend of the power drop event; And determining an average value of the duration sequence of the power drop event as the estimated shielding duration of the future shielding event, wherein the duration sequence comprises a plurality of event durations used for representing the duration of the power drop event.
9. The method for short-term predictive and optimal control of a light storage system based on model predictive control of claim 8, wherein predicting the energy storage gap of the light storage system based on future occlusion events caused by the cloud cover system comprises: inputting a solar irradiance predicted value in a future preset time length into a power conversion model of the light storage system, and determining a reference power predicted track; Acquiring a reference predicted power of the reference power predicted track at the estimated occurrence time in response to the estimated occurrence time of the future shielding event being within the future preset time; Multiplying the reference predicted power by the estimated occlusion depth of the future occlusion event to determine a peak power loss of the future occlusion event; and determining an energy storage energy gap of the light storage system based on the peak power loss of the future shielding event and the estimated shielding duration.
10. The method for short-term predictive and optimal control of a light storage system based on model predictive control of claim 1, wherein the predicting the energy storage gap of the light storage system based on the future occlusion event caused by the cloud cover system further comprises: Comparing the energy storage energy gap of the light storage system with available discharge energy of the light storage system to obtain an energy comparison result; determining that the light storage system is in an energy-sufficient scene in response to the energy comparison result indicating that the stored energy gap is less than or equal to the available discharge energy; and determining that the light storage system is in an energy shortage scene in response to the energy comparison result indicating that the stored energy gap is larger than the available discharge energy.

Description

Short-term prediction and optimization control method for optical storage system based on model prediction control Technical Field The invention relates to the technical field of optical storage system control, in particular to an optical storage system short-term prediction and optimization control method based on model prediction control. Background At the moment of the vigorous development of renewable energy sources, the light storage system is used as a key device for organically combining photovoltaic power generation and energy storage technology, and has great significance on improving the energy utilization efficiency and guaranteeing the power supply stability. The model predictive control (ModelPredictiveControl, MPC) is widely applied to the field of optical storage system control by virtue of the advantages of rolling optimization and look-ahead decision, and realizes the efficient operation of the optical storage system. At present, a short-term prediction and optimization control method of an optical storage system based on MPC (MPC) solves the problem of finite time domain optimization by relying on macroscopic parameters such as solar irradiance, cloud cover and the like provided by a weather forecast system in each control period, comprehensively considers conditions such as charge and discharge power limitation, power grid interaction constraint and the like, outputs a control instruction at the current moment, and re-optimizes according to new measurement values and updated predictions in the next period. However, the macroscopic parameters provided by weather forecast make it difficult to capture short term power drop events caused by cloud shielding on the photovoltaic array. The lack of the prediction granularity causes blindness in energy storage scheduling, such as unexpected power sudden drop when the sun is in a cloudy state, the energy storage is forced to be powered from the power grid in an emergency, and the reliability of optimal control of the optical storage system is poor. Disclosure of Invention The embodiment of the invention provides a short-term prediction and optimization control method for an optical storage system based on model prediction control, which can improve the reliability of the optimal control of the optical storage system. In a first aspect of the embodiment of the present invention, a method for short-term predictive and optimization control of an optical storage system based on model predictive control is provided, including: responding to a photovoltaic power sequence obtained from a photovoltaic storage system, carrying out signal analysis on the photovoltaic power sequence, and determining a power drop event in the photovoltaic power sequence, wherein the photovoltaic power sequence comprises historical photovoltaic power generation powers in a historical preset time period nearest to the current moment, and the power drop event is used for representing a time period in which the historical photovoltaic power generation power is in a downward trend; analyzing the evolution trend of the cloud layer system based on a drop event sequence obtained by constructing each power drop event according to a time sequence, and extracting cloud cluster evolution characteristics of the cloud layer system; based on cloud evolution characteristics of the cloud layer system, predicting future shielding events caused by the cloud layer system; And predicting an energy storage energy gap of the optical storage system based on a future shielding event caused by the cloud layer system, so that the optical storage system is optimally controlled based on the energy storage energy gap. Further, the present invention also provides for signal analysis of the photovoltaic power sequence to determine a power sag event in the photovoltaic power sequence, including: performing discrete Fourier transform on the photovoltaic power sequence to obtain a frequency domain spectral density curve of the photovoltaic power sequence; determining dominant frequency of the photovoltaic power sequence based on the frequency domain spectral density curve, wherein the dominant frequency is used for separating a cloud cluster slow-change process from a cloud cluster fast-change process; extracting a fluctuation component sequence with the frequency larger than the dominant frequency from the frequency domain spectral density distribution; A power sag event in the photovoltaic power sequence is determined based on the sequence of fluctuating components. Further, the present invention also proposes that determining a dominant frequency of a photovoltaic power sequence based on a frequency domain spectral density curve, comprising: performing second order derivative operation on the frequency domain spectrum density curve to determine the frequency value of each inflection point in the frequency domain spectrum density curve; and determining the inflection point frequency value with the maximum fre