CN-121984132-A - Micro-grid dispatching method and system

CN121984132ACN 121984132 ACN121984132 ACN 121984132ACN-121984132-A

Abstract

The application provides a micro-grid dispatching method and a micro-grid dispatching system, which relate to the technical field of micro-grid energy management, wherein the method collects running data of each unit of a micro-grid through edge computing equipment and generates a photovoltaic power prediction sequence of a future period by utilizing a local lightweight time sequence prediction model; the operation data and the prediction result are uploaded to a cloud platform, the cloud platform performs global optimization by combining the time-of-use electricity price and the dynamic carbon-shift factor, a scheduling strategy is generated and issued, and the edge equipment generates regulation and control instructions for the energy storage and charging piles in advance before power fluctuation occurs according to the strategy and the prediction deviation. According to the cloud edge cooperative real-time decision and advanced control method, cloud edge cooperative real-time decision and advanced control are realized, and the photovoltaic absorption rate and the system economy are effectively improved.

Inventors

ZOU ZIKANG
HE TAO
ZHANG XIANGXIN
WU HONGLE
GUO YIRUI
Bao Yuyi

Assignees

温州职业技术学院

Dates

Publication Date: 20260505
Application Date: 20260407

Claims (10)

1. A method for micro-grid scheduling, wherein the method is performed by an edge computing device deployed locally on a micro-grid in conjunction with a remote cloud platform, comprising: collecting operation data of a photovoltaic power generation unit, an energy storage unit, a load unit and a charging pile in the micro-grid through the edge computing equipment; in the edge computing equipment, the operation data are input into a light-weight time sequence prediction model which is trained and deployed locally for processing, and a photovoltaic power prediction value sequence used for representing the trend of the output power variation of the photovoltaic power generation unit in a future set period is generated; The cloud platform calculates time-of-use electricity price information and a dynamic carbon-emission factor as input parameters based on a preset comprehensive scheduling gain function, generates a scheduling strategy for the energy storage unit and the charging pile, and issues the scheduling strategy to the edge computing equipment; And the edge computing equipment receives the scheduling strategy, and generates an advanced control instruction for pre-adjusting the charge and discharge state of the energy storage unit and/or the running power of the charging pile before the future set period of time arrives based on the expected deviation of the photovoltaic power predicted value sequence and a reference scheduling plan in the scheduling strategy.
2. The micro grid scheduling method of claim 1, wherein the edge computing device generating the lead control instruction comprises: analyzing the scheduling strategy to obtain a reference scheduling plan of the energy storage unit and the charging pile in the future set period; based on the photovoltaic power predicted value sequence, calculating the supply-demand relation between the photovoltaic power generation power and the current load power in the future, and comparing the supply-demand relation with the planned value at the corresponding moment in the reference scheduling plan to obtain the expected deviation; And executing advanced control logic according to the expected deviation and the current state of charge of the energy storage unit and in combination with a preset configurable threshold, wherein the advanced control logic comprises: when the expected deviation indicates that the predicted photovoltaic power generation power is higher than a reference plan expected value and the state of charge of the energy storage unit is lower than a preset charging threshold, generating an advanced control instruction for increasing the charging power of the energy storage unit; and when the expected deviation indicates that the predicted photovoltaic power generation power is lower than the expected value of the reference plan and the state of charge of the energy storage unit is higher than the preset discharge threshold value, generating an advanced control instruction for starting or increasing the discharge power of the energy storage unit and/or reducing the running power of the charging pile.
3. The micro-grid dispatching method according to claim 1, wherein the cloud platform calculates, based on a preset comprehensive dispatching profit function, time-of-use electricity price information and a dynamic carbon-displacement factor as input parameters, and generates a dispatching strategy for the energy storage unit and the charging pile, and the method comprises the following steps: The cloud platform receives the operation data and the photovoltaic power predicted value sequence, and obtains time-of-use electricity price information of a power grid and dynamic carbon emission factors of areas; selecting a corresponding comprehensive scheduling benefit function according to a preset scheduling target, wherein the comprehensive scheduling benefit function is a multi-target weighting function and at least comprises an economic cost item and a carbon emission cost item, and the state of charge constraint and the power limit constraint of an energy storage unit are considered; Based on the current load power in the operation data, the state of charge of an energy storage unit, the photovoltaic power predicted value sequence, the time-of-use electricity price information and the dynamic carbon emission factor, solving the comprehensive dispatching yield function by adopting a preset optimization algorithm to obtain an optimal charge and discharge power sequence of the energy storage unit and an optimal power dispatching sequence of the charging pile in a future set period; And converting the optimal charge and discharge power sequence and the optimal power scheduling sequence into scheduling instructions executed by the edge computing equipment to form the scheduling strategy.
4. The micro grid scheduling method according to claim 1, wherein the edge computing device generates the advanced control instruction, specifically comprising: Calculating a photovoltaic power prediction average value in a future set period based on the photovoltaic power prediction value sequence; executing the following strategy decision logic according to the photovoltaic power prediction average value, the current load power, the current charge state of the energy storage unit and the current electricity price period: when the photovoltaic power prediction average value is larger than a preset first power threshold value and multiplied by the current load power, and the current state of charge of the energy storage unit is smaller than the preset first state of charge threshold value, generating an advanced control instruction for increasing the charging power of the energy storage unit; When the photovoltaic power prediction average value is smaller than a preset second power threshold value and multiplied by the current load power, and the current state of charge of the energy storage unit is larger than the preset second state of charge threshold value and is currently in a peak electricity price period, generating an advanced control instruction for starting or increasing the discharge power of the energy storage unit and/or reducing the running power of the charging pile; otherwise, generating a maintaining instruction for maintaining the current state of the energy storage unit and the power is zero.
5. The micro grid scheduling method of claim 1, further comprising a dynamic carbon footprint tracking step: The cloud platform or the edge computing equipment periodically acquires a dynamic power grid carbon emission factor of an area where the micro power grid is located; Calculating real-time electricity purchasing power of the micro-grid from a main grid based on the operation data, and calculating the range two-carbon emission of the micro-grid in real time by combining the dynamic grid carbon emission factor and time integral; And adding a digital signature to the carbon bank record containing the timestamp, the carbon emission and the associated operation data, uploading the signed record to a blockchain network for certification, and acquiring a corresponding blockchain transaction identifier as a certification.
6. The micro grid scheduling method according to claim 1, further comprising: When the communication network between the edge computing equipment and the cloud platform is interrupted, the edge computing equipment continuously executes control on the energy storage unit and/or the charging pile based on the locally stored photovoltaic power predicted value sequence and a preset local control rule, wherein the local control rule at least comprises a reference scheduling plan which is obtained by analyzing a latest received scheduling strategy; The edge computing equipment records the operation data collected during interruption and the executed control instruction in a local storage medium according to the time stamp sequence; When the communication network is restored, the edge computing device automatically supplements the data with the time stamp recorded during the interruption to the cloud platform, and marks that the data are synchronized.
7. The micro grid scheduling method of claim 1, wherein the training and updating of the timing prediction model is performed by the cloud platform, comprising: The cloud platform acquires a training data set containing historical photovoltaic power, historical illumination intensity and historical environment temperature from historical operation data; Performing supervised training on an original deep learning model adopting an encoder-decoder structure by using the training data set so as to minimize errors between a predicted value and an actual value of the photovoltaic power in a future set period predicted by the model; After training and verification are completed, converting the original deep learning model into a lightweight inference format suitable for edge calculation, and carrying out quantization compression to generate the lightweight time sequence prediction model; and issuing and deploying the generated lightweight time sequence prediction model on the edge computing equipment.
8. The micro grid scheduling method according to claim 5, wherein the dynamic carbon footprint tracking step specifically comprises: Acquiring total load power P load , photovoltaic power generation power P pv and energy storage charge-discharge power P storage of the micro-grid; According to the power balance principle, calculating real-time electricity purchasing power P grid of the micro-grid from the main grid according to a formula P grid =P load -P pv -P storage ; acquiring a dynamic carbon emission factor f of an area where the micro-grid is located; calculating and accumulating the carbon emission increment delta E in the current time period according to the formula delta E=P grid multiplied by delta t multiplied by f to obtain the two-carbon emission in the range; Wherein P storage is positive in charging and negative in discharging.
9. A micro grid scheduling method according to claim 3, wherein the comprehensive scheduling benefit function is a total cost J in a minimum scheduling period, and the specific expression is: ; Wherein J is the total cost in the dispatching cycle, and comprises three parts of electricity purchasing cost, carbon emission cost and energy storage loss cost, t is the time step in the dispatching cycle; Time-sharing electricity price is the t time step; Purchasing electric power from a main power grid for the t time step; is a preset carbon emission cost coefficient; a dynamic carbon number factor which is the t time step; the energy storage loss cost coefficient is preset; Charging power for the energy storage unit of the t time step; And discharging power for the energy storage unit in the t time step.
10. A microgrid dispatching system, the system comprising: The data acquisition module is used for acquiring operation data of the photovoltaic power generation unit, the energy storage unit, the load unit and the charging pile in the micro-grid; The edge computing device is in communication connection with the data acquisition module and comprises: The time sequence prediction model unit is used for inputting the operation data into a light time sequence prediction model which is trained and deployed locally for processing, and generating a photovoltaic power prediction value sequence used for representing the trend of the output power variation of the photovoltaic power generation unit in a future set period; The data uploading unit is used for uploading the operation data and the photovoltaic power predicted value sequence to a cloud platform; the instruction generation unit is used for receiving a scheduling strategy issued by the cloud platform, and generating an advanced control instruction for pre-adjusting the charge and discharge state of the energy storage unit and/or the running power of the charging pile before the future set period of time comes based on the expected deviation of the photovoltaic power predicted value sequence and a reference scheduling plan in the scheduling strategy; a local storage unit for caching operation data and control instructions during interruption of the communication network; The communication recovery unit is used for automatically supplementing the buffered data after the network is recovered; the cloud platform, in communication connection with the edge computing device, comprises: The strategy optimization unit is used for calculating based on the received operation data and the photovoltaic power predicted value sequence and on a preset comprehensive scheduling income function by taking time-of-use electricity price information and a dynamic carbon-emission factor as input parameters to generate a scheduling strategy for the energy storage unit and the charging pile; and the strategy issuing unit is used for issuing the scheduling strategy to the edge computing equipment.

Description

Micro-grid dispatching method and system Technical Field The application relates to the technical field of micro-grid energy management, in particular to a micro-grid dispatching method and system. Background In the prior art, a centralized cloud computing architecture is generally adopted in an energy management system of a micro-grid, namely, telemetry data (such as photovoltaic power, load power and the like) acquired on site are all uploaded to a cloud server, and unified predictive analysis and scheduling decision is carried out by a cloud. However, the architecture has obvious technical defects that firstly, because data needs to be transmitted through a network and processed in a cloud end in a centralized way, the end-to-end response delay usually reaches the second level or even the minute level, the requirement on millisecond-level real-time control under the photovoltaic power rapid fluctuation scene is difficult to meet, secondly, the system is highly dependent on network connection, once the network is interrupted, the cloud end decision function is invalid, the system can only be degenerated into simple local logic control and intelligent advanced scheduling cannot be realized, and in addition, huge pressure is caused to network bandwidth due to continuous uploading of massive high-frequency data, and data privacy and security risks exist. Although the edge computing technology is introduced to perform data processing locally, the existing scheme still faces technical bottlenecks that a large AI model is difficult to deploy, prediction accuracy and reasoning speed are difficult to consider, and a prediction result cannot be fused with a deep closed loop of an underlying control strategy under the condition of limited computing resources of the edge equipment. Disclosure of Invention The application provides a micro-grid dispatching method and a micro-grid dispatching system, which are used for solving the problems of local prediction, global optimization control disconnection and control instruction lag caused by the limitation of a computing architecture in the existing micro-grid dispatching scheme. In a first aspect, the present application provides a micro-grid dispatching method, which is cooperatively executed by an edge computing device deployed locally on a micro-grid and a remote cloud platform, and includes: collecting operation data of a photovoltaic power generation unit, an energy storage unit, a load unit and a charging pile in the micro-grid through the edge computing equipment; in the edge computing equipment, the operation data are input into a light-weight time sequence prediction model which is trained and deployed locally for processing, and a photovoltaic power prediction value sequence used for representing the trend of the output power variation of the photovoltaic power generation unit in a future set period is generated; The cloud platform calculates time-of-use electricity price information and a dynamic carbon-emission factor as input parameters based on a preset comprehensive scheduling gain function, generates a scheduling strategy for the energy storage unit and the charging pile, and issues the scheduling strategy to the edge computing equipment; And the edge computing equipment receives the scheduling strategy, and generates an advanced control instruction for pre-adjusting the charge and discharge state of the energy storage unit and/or the running power of the charging pile before the future set period of time arrives based on the expected deviation of the photovoltaic power predicted value sequence and a reference scheduling plan in the scheduling strategy. In a second aspect, the present application also provides a micro grid dispatching system, the system comprising: The data acquisition module is used for acquiring operation data of the photovoltaic power generation unit, the energy storage unit, the load unit and the charging pile in the micro-grid; The edge computing device is in communication connection with the data acquisition module and comprises: The time sequence prediction model unit is used for inputting the operation data into a light time sequence prediction model which is trained and deployed locally for processing, and generating a photovoltaic power prediction value sequence used for representing the trend of the output power variation of the photovoltaic power generation unit in a future set period; The data uploading unit is used for uploading the operation data and the photovoltaic power predicted value sequence to a cloud platform; the instruction generation unit is used for receiving a scheduling strategy issued by the cloud platform, and generating an advanced control instruction for pre-adjusting the charge and discharge state of the energy storage unit and/or the running power of the charging pile before the future set period of time comes based on the expected deviation of the photovoltaic power predicted value sequence and a refere