CN-121998287-A - Energy management method and device for distributed energy system, system and storage medium

CN121998287ACN 121998287 ACN121998287 ACN 121998287ACN-121998287-A

Abstract

The embodiment of the application provides an energy management method and device of a distributed energy system, a system and a storage medium, and relates to the technical field of the distributed energy system and energy management. The method comprises the steps of extracting a typical load characteristic mode through clustering analysis based on historical operation data, carrying out short-term load prediction through a neural network model based on the typical load characteristic mode and external prediction information, establishing a multi-objective optimization model comprising an economical objective and an environmental objective, wherein decision variables of the optimization model at least comprise output plans of controllable energy units in a distributed energy system, and solving the multi-objective optimization model by adopting an optimization algorithm based on a short-term load prediction result to generate a scheduling scheme of the distributed energy system. The method is beneficial to improving the utilization rate of renewable energy sources, promoting the effective balance between the running cost and environmental emission, enhancing the capability of an algorithm to jump out of local optimum and improving the overall running performance of a distributed energy system.

Inventors

WANG YUEJIAO
ZHAO PU
LIU YIYUAN
YANG SONG
LU ZHIPENG
XING JIAWEI
ZHENG ZHIJIE
YU PI
WANG CHUNYI
SUN SHUMIN
CHENG YAN

Assignees

国网山东省电力公司电力科学研究院
国家电网有限公司

Dates

Publication Date: 20260508
Application Date: 20251210

Claims (16)

1. A method of energy management for a distributed energy system, comprising: extracting a typical load characteristic mode through cluster analysis based on historical operation data; based on the typical load characteristic mode and external prediction information, carrying out short-term load prediction through a neural network model; Establishing a multi-objective optimization model comprising an economical objective and an environmental objective, wherein decision variables of the optimization model at least comprise output plans of all controllable energy units in the distributed energy system; and solving the multi-objective optimization model by adopting an optimization algorithm based on the short-term load prediction result to generate a scheduling scheme of the distributed energy system.
2. The method of claim 1, wherein the step of determining the position of the substrate comprises, The distributed energy system comprises at least one of a photovoltaic power generation unit, an energy storage unit, a diesel generator, a utility grid interface, and a load unit, and/or, The controllable energy unit comprises at least one of an energy storage unit, a diesel generator and a public power grid interface.
3. The method of claim 1, wherein building a multi-objective optimization model comprising an economic objective and an environmental objective comprises: constructing an operation cost objective function, wherein the operation cost objective function comprises at least two of outsourcing electricity cost, diesel generator fuel cost, energy storage unit loss cost and renewable energy waste electricity punishment cost; an emissions objective function is constructed that quantifies diesel generator operation and pollutant emissions generated from utility grid purchases.
4. The method of claim 3, wherein the step of, The renewable energy waste penalty cost is related to the waste power of the photovoltaic power generation unit and/or, The pollutants quantified by the emissions objective function include at least one of carbon dioxide, sulfur dioxide, and nitrogen oxides.
5. The method of claim 3, wherein building a multi-objective optimization model comprising an economic objective and an environmental objective further comprises: Establishing a system power balance constraint for ensuring that at least one of the output of the photovoltaic power generation unit, the energy storage unit, the diesel generator and the utility grid interface is balanced with the load demand and the waste electric power, and/or, Establishing upper and lower limit constraints of the output of each energy unit, and/or, And establishing an energy storage unit operation constraint, wherein the energy storage unit operation constraint comprises at least one of a charge state constraint, a charge and discharge power constraint and a charge state update model.
6. The method according to any one of claims 1 to 5, wherein extracting a typical load feature pattern by cluster analysis based on historical operating data comprises: clustering the historical data sequence by adopting a clustering algorithm, wherein the historical data sequence at least comprises an electricity price sequence, an outdoor temperature sequence and a load sequence; and outputting each class of central sequences obtained after clustering as a typical load characteristic mode.
7. The method of claim 6, wherein the step of providing the first layer comprises, The clustering algorithm is a K-means algorithm, and/or, Using euclidean distance as a similarity measure, and/or, With a change in cluster center less than the preset threshold is used as a convergence condition.
8. The method of claim 6, wherein prior to clustering the historical data sequences using a clustering algorithm, further comprising: and carrying out normalization processing on each dimension data in the historical data sequence, wherein the historical data sequence also comprises a characteristic for identifying the date type or season.
9. The method according to any one of claims 1 to 5, wherein short-term load prediction by a neural network model based on a typical load characteristic pattern and external prediction information, comprises: According to the predicted electricity price of the predicted period, the predicted outdoor temperature, the typical load characteristic mode of the corresponding period and the load data at the same time of history, constructing an input characteristic vector of the neural network; Inputting the input feature vector to a feedforward neural network model; And receiving a predicted load value output by the feedforward neural network model.
10. The method of claim 9, wherein the step of determining the position of the substrate comprises, The load data of the same time historic includes load data of the same time of the previous day of the predicted day and load data of the same time of the same day of the previous week of the predicted day, and/or, The hidden layer activation function of the feedforward neural network adopts a Sigmoid function or a ReLU function.
11. The method of any one of claims 1 to 5, wherein solving the multi-objective optimization model using an optimization algorithm based on the short-term load prediction results generates a scheduling scheme for the distributed energy system, comprising: Taking the short-term load prediction result as an input parameter of a load demand in a multi-objective optimization model; initializing to form an initial solution population by taking the output plan of each controllable energy unit in each future period as a decision variable; carrying out iterative solution based on a one-to-one optimization algorithm, and updating the population through a one-to-one comparison and repositioning mechanism between solution vectors according to the objective function value and constraint satisfaction condition in each iteration; and when the iteration termination condition is met, outputting the optimal solution vector in the current population as a scheduling scheme.
12. The method of claim 11, wherein the step of determining the position of the probe is performed, When the one-to-one optimization algorithm is used for population updating, the basis for evaluating the solution vector comprises an objective function value and the violation degree of the system operation constraint.
13. The method according to any one of claims 1 to 5, wherein after the short-term load prediction by the neural network model based on the typical load characteristic pattern and the external prediction information, further comprising: And evaluating the prediction result by using a prediction precision evaluation index to quantify the prediction performance, wherein the prediction precision evaluation index comprises at least two of an average absolute error, a root mean square error, an average absolute percentage error and a decision coefficient.
14. An energy management device of a distributed energy system comprising a processor and a memory storing program instructions, wherein the processor is configured to perform the energy management method of the distributed energy system of any of claims 1 to 13 when the program instructions are executed.
15. A distributed energy system, comprising: And distributed energy system body The energy management device of a distributed energy system of claim 14, mounted to the distributed energy system body.
16. A computer readable storage medium storing program instructions which, when executed, are configured to cause a computer to perform the energy management method of a distributed energy system according to any one of claims 1 to 13.

Description

Energy management method and device for distributed energy system, system and storage medium Technical Field The present application relates to the field of distributed energy systems and energy management technologies, and in particular, to an energy management method and apparatus for a distributed energy system, and a storage medium. Background At present, as the permeability of renewable energy sources in distributed energy systems is continuously increased, the operation of the systems exhibits significant randomness, intermittence and multi-source coupling characteristics. Traditional centralized energy management mode is difficult to adapt to fluctuation and dispersion of distributed energy, so that system operation efficiency is low, power supply reliability is insufficient, and local renewable energy is difficult to fully utilize. Especially in the scene of high proportion photovoltaic access, the rapid change of illumination condition often causes power supply and demand unbalance, brings serious challenges to the safe and economic operation of the power grid. In the prior art, a local control strategy based on rules or empirical thresholds is adopted, or an optimized scheduling model with a single economic goal is introduced. For example, basic power balance and running cost control are achieved by setting photovoltaic output priority, energy storage charge and discharge fixed thresholds, or diesel generator start-stop conditions. Some methods also attempt to optimize a certain operation objective of the system by using an optimization algorithm, such as linear programming or dynamic programming, so as to improve economy. In the process of implementing the embodiment of the application, the related art is found to have at least the following problems: The adoption of the related technology enhances the autonomous operation capability of the system to a certain extent and reduces the dependence on an external power grid. However, in practical applications, the related art focuses on optimization of a single objective, or relies on a simplified model and empirical rules, and cannot fully consider the coupling relationship among load prediction accuracy, multi-objective coordination and global optimization capability, so that the scheduling scheme is often difficult to achieve effective balance between economy, environmental protection and renewable energy utilization, and is easy to fall into local optimization, which limits further improvement of the overall performance of the distributed energy system. Disclosure of Invention The embodiment of the application provides an energy management method and device of a distributed energy system, the distributed energy system and a storage medium. In a first aspect of an embodiment of the present application, there is provided an energy management method of a distributed energy system, including: extracting a typical load characteristic mode through cluster analysis based on historical operation data; based on the typical load characteristic mode and external prediction information, carrying out short-term load prediction through a neural network model; Establishing a multi-objective optimization model comprising an economical objective and an environmental objective, wherein decision variables of the optimization model at least comprise output plans of all controllable energy units in the distributed energy system; and solving the multi-objective optimization model by adopting an optimization algorithm based on the short-term load prediction result to generate a scheduling scheme of the distributed energy system. In an alternative embodiment of the application, the distributed energy system comprises at least one of a photovoltaic power generation unit, an energy storage unit, a diesel generator, a utility grid interface, and a load unit, and/or the controllable energy unit comprises at least one of an energy storage unit, a diesel generator, and a utility grid interface. In an alternative embodiment of the application, building a multi-objective optimization model comprising an economic objective and an environmental objective comprises: constructing an operation cost objective function, wherein the operation cost objective function comprises at least two of outsourcing electricity cost, diesel generator fuel cost, energy storage unit loss cost and renewable energy waste electricity punishment cost; an emissions objective function is constructed that quantifies diesel generator operation and pollutant emissions generated from utility grid purchases. In an alternative embodiment of the application, the renewable energy waste penalty cost is related to the waste light power of the photovoltaic power generation unit and/or the pollutant quantified by the emission objective function comprises at least one of carbon dioxide, sulfur dioxide and nitrogen oxides. In an alternative embodiment of the present application, establishing a multi-objective optimizatio