CN118798584B - A method, system, medium and device for robust optimization scheduling of building microgrid - Google Patents
A method, system, medium and device for robust optimization scheduling of building microgrid Download PDFInfo
- Publication number
- CN118798584B CN118798584B CN202411266152.6A CN202411266152A CN118798584B CN 118798584 B CN118798584 B CN 118798584B CN 202411266152 A CN202411266152 A CN 202411266152A CN 118798584 B CN118798584 B CN 118798584B
- Authority
- CN
- China
- Prior art keywords
- load
- stage
- source
- day
- source load
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000005457 optimization Methods 0.000 title claims abstract description 50
- 230000002787 reinforcement Effects 0.000 claims abstract description 17
- 238000004146 energy storage Methods 0.000 claims description 29
- 230000004044 response Effects 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 14
- 230000007704 transition Effects 0.000 claims description 13
- 230000006399 behavior Effects 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 8
- 230000009471 action Effects 0.000 claims description 7
- 238000011217 control strategy Methods 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 238000003860 storage Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 230000003993 interaction Effects 0.000 claims description 4
- 238000011084 recovery Methods 0.000 claims 6
- 238000010248 power generation Methods 0.000 description 28
- 239000010410 layer Substances 0.000 description 22
- 238000010438 heat treatment Methods 0.000 description 17
- 238000009423 ventilation Methods 0.000 description 17
- 239000013598 vector Substances 0.000 description 13
- 238000009826 distribution Methods 0.000 description 11
- 238000004378 air conditioning Methods 0.000 description 10
- 230000006870 function Effects 0.000 description 8
- 230000001186 cumulative effect Effects 0.000 description 3
- 230000005611 electricity Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 238000005315 distribution function Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004904 shortening Methods 0.000 description 2
- 101100001674 Emericella variicolor andI gene Proteins 0.000 description 1
- 240000002853 Nelumbo nucifera Species 0.000 description 1
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 1
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000007363 regulatory process Effects 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 238000010977 unit operation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06311—Scheduling, planning or task assignment for a person or group
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/092—Reinforcement learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06315—Needs-based resource requirements planning or analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Health & Medical Sciences (AREA)
- Operations Research (AREA)
- General Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Software Systems (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Primary Health Care (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明提出了一种建筑微电网鲁棒优化调度方法、系统、介质及装置，属于建筑微电网优化调度技术领域，包括：基于日前源荷预测值生成若干源荷样本，表示源荷在次日的可能实现值，从而构造出源荷不确定性集；建立日前调度层，包括：将源荷不确定集划分为较大部分源荷不确定集、较小部分源荷不确定集及建立日前三阶段鲁棒优化调度模型；在日前调度层的基础上，再基于模型强化学习建立日内实时优化调度层，输出的日内实时最优调度计划。
The present invention proposes a method, system, medium and device for robust optimization scheduling of a building microgrid, which belongs to the technical field of building microgrid optimization scheduling, including: generating a number of source load samples based on the day-ahead source load prediction value, representing the possible realization value of the source load on the next day, thereby constructing a source load uncertainty set; establishing a day-ahead scheduling layer, including: dividing the source load uncertainty set into a larger part of the source load uncertainty set, a smaller part of the source load uncertainty set and establishing a three-stage day-ahead robust optimization scheduling model; on the basis of the day-ahead scheduling layer, establishing an intraday real-time optimization scheduling layer based on model reinforcement learning, and outputting an intraday real-time optimal scheduling plan.
Description
Technical Field
The invention belongs to the technical field of building micro-grid optimal scheduling, and particularly relates to a building micro-grid robust optimal scheduling method and system.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
At present, the building energy structure is relatively single and mainly takes fossil energy, and along with the low-carbon development of the building, the building energy structure tends to be diversified, and renewable energy sources gradually take the dominant role. In the process of continuously and synchronously expanding the renewable energy power generation and the building power consumption scale, a building micro-grid is generated.
To ensure efficient and safe operation of the building microgrid, the key is an efficient scheduling strategy and sufficient adjustable resources. However, when new energy power generation such as photovoltaic power generation and wind power discharge occupies a large proportion in the system, scheduling work of the micro-grid can encounter some difficulties. Firstly, the power demand presents randomness due to the uncertainty of the power consumption habit of a user, and secondly, the yield of new energy power generation such as photovoltaic power generation, wind power discharge and the like is also influenced by natural conditions such as sunlight intensity, wind power size and the like, and the power generation system has instability.
At present, the artificial intelligence technology has limitations in accurately predicting new energy power generation and power requirements such as photovoltaic power generation, wind power discharge and the like, which causes prediction deviation and increases the problem of uncertainty of source load, and the method has direct influence on scheduling planning of a micro-grid. And secondly, with large-scale grid connection of new energy power generation such as photovoltaic power generation and wind power generation, the capacity of the traditional adjustable power generation equipment is relatively reduced, so that adjustable resources for balancing new energy power generation fluctuation and power demand of photovoltaic power generation and wind power generation in a micro-grid become insufficient, and the influence of the source load uncertainty problem is further aggravated.
Through retrieval, related technical schemes of the building micro-grid exist in the prior art, but the technical problem is that the traditional building micro-grid optimization scheduling generally comprises two layers, namely, two-stage robust optimization scheduling before the day and real-time optimization scheduling in the day. In day-ahead scheduling, the construction of the set of source load uncertainties does not take full advantage of the correlation between the source load prediction bias and the predicted value, which results in the set of constructed uncertainties being too conservative. Furthermore, two-phase robust optimized scheduling does not take into account the multi-time scale characteristics of source-load devices and uncertainties in building micro-grids, which makes scheduling plans for those source-load side devices with short time scale characteristics too conservative. In the real-time scheduling layer in the day, the randomness of the source load prediction deviation cannot be fully considered by the traditional method based on mathematical programming, so that the influence on the source load prediction deviation is difficult to effectively influence.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a robust optimal scheduling method for a building micro-grid, which is based on a daily real-time optimal scheduling model for model reinforcement learning, and calculates the daily real-time scheduling result by utilizing the daily scheduling result so as to ensure the stable and efficient operation of the building micro-grid system.
To achieve the above object, one or more embodiments of the present invention provide the following technical solutions:
in a first aspect, a robust optimal scheduling method for a building micro-grid is disclosed, comprising:
Generating a plurality of source load samples based on the day-ahead source load predicted value, wherein the source load samples represent possible realization values of the source load in the next day, so as to construct a source load uncertainty set;
The method comprises the steps of dividing a source load uncertainty set into a larger part of source load uncertainty set and a smaller part of source load uncertainty set, and establishing a three-stage robust optimal scheduling model before the day, wherein the three-stage robust optimal scheduling model before the day comprises the following steps:
determining a scheduling plan under a source load predicted value nominal scene;
The second stage is to calculate the pursuit plan of the heating, ventilation, air conditioning and electric vehicle lumped load of the gas turbine, energy storage battery and load aggregator with different response speed with the aim of maximizing the minimum pursuit cost of the building micro-grid in the worst scene of the uncertainty concentration of the larger part of source load;
The second-stage pursuit plan is overlapped on the basis of the first-stage nominal plan, the third stage aims at maximizing the minimum pursuit cost of the building micro-grid in the condition of uncertainty concentration of small partial source load and worst, and calculates the pursuit plan of the lumped load of the energy storage battery and the electric vehicle of the load aggregator, which has higher response speed;
on the basis of the day-ahead scheduling layer, the day-ahead real-time optimal scheduling layer is built based on model reinforcement learning, and the day-ahead real-time optimal scheduling plan is output.
As a further technical solution, the specific process of constructing the source load uncertainty set is as follows:
acquiring the latest day-ahead predicted value;
acquiring an expected value and a covariance matrix of conditional correlation between a source load predicted value and a true value;
sampling the conditional correlation between the predicted value and the true value of the source load to generate a plurality of samples, and then obtaining the samples of the true value of the source load;
Based on The sample obtained by time sampling and the latest day-ahead predicted value establish a source load uncertainty set.
As a further technical scheme, the source load uncertainty set is decomposed into an uncertainty set with larger fluctuation amplitude and long time scale characteristics and an uncertainty set with smaller fluctuation amplitude and short time scale characteristics.
The method comprises the following steps of determining a scheduling plan under a source load predicted value nominal scene, wherein the scheduling plan comprises a first-stage optimization target, constraint conditions and decision variables;
The first stage constraints include power balance constraints, gas turbine operating constraints, grid constraints, energy storage battery operating constraints, and load aggregator constraints.
As a further technical scheme, the optimization target of the second stage is the second stage additional cost, and specifically comprises the additional cost of the gas turbine, the additional cost of the energy storage battery and the compensation cost of the building micro-grid on the lumped load regulation behavior of the heating ventilation air conditioner of the load aggregator;
The second stage is a new constraint relative to the first stage, which is an upper limit value and a lower limit value of the user thermal comfort temperature of the heating ventilation air conditioner in the building, and the corresponding constraint condition aims to limit the indoor temperature to the thermal comfort range of the user.
As a further technical scheme, the third-stage additional cost comprises additional cost of an energy storage battery and compensation cost of a building micro-grid on lumped load adjustment behavior of the electric automobile of the load aggregator;
The third stage is to newly add upper and lower limit values of the lumped target charge quantity of the electric vehicle to the second stage, and Upper and lower limits of the deviation of the target charge amount tolerable to the individual electric car users.
As a further technical scheme, the method for establishing the intra-day real-time optimized scheduling layer based on model reinforcement learning comprises the following steps:
Constructing a real-time daily optimization scheduling problem as a Markov decision process, and defining states, actions and rewards;
Learning a state transition model of the building micro-grid system by utilizing LSTM, wherein the input of the state transition model is state and action, and the output is state variation;
Interacting the actor network for learning the deterministic strategy with the learned state transition model of the building micro-grid system in a mode of optimizing a limited prediction interval, and continuously updating actor and critic network model parameters by using accumulated rewards in the limited prediction interval obtained by interaction as an updating target of the critic network to finally obtain an optimal control strategy;
and after the daily load predicted value is obtained at set time intervals, the scheduling center obtains a daily real-time scheduling plan by utilizing actor network optimal control strategies.
In a second aspect, a building micro-grid robust optimal scheduling system is disclosed, comprising:
a source load uncertainty set construction module configured to generate a plurality of source load samples based on the day-ahead source load predicted value, representing possible realization values of the source load on the next day, thereby constructing a source load uncertainty set;
The day-ahead scheduling layer establishing module is configured to establish a day-ahead scheduling layer and comprises the steps of dividing a source load uncertainty set into a larger part of source load uncertainty set, a smaller part of source load uncertainty set and establishing a day-ahead three-stage robust optimal scheduling model, wherein the day-ahead three-stage robust optimal scheduling model comprises:
determining a scheduling plan under a source load predicted value nominal scene;
The second stage is to calculate the pursuit plan of the heating, ventilation, air conditioning and electric vehicle lumped load of the gas turbine, energy storage battery and load aggregator with different response speed with the aim of maximizing the minimum pursuit cost of the building micro-grid in the worst scene of the uncertainty concentration of the larger part of source load;
The second-stage pursuit plan is overlapped on the basis of the first-stage nominal plan, the third stage aims at maximizing the minimum pursuit cost of the building micro-grid in the condition of uncertainty concentration of small partial source load and worst, and calculates the pursuit plan of the lumped load of the energy storage battery and the electric vehicle of the load aggregator, which has higher response speed;
And the intra-day real-time optimal scheduling module is configured to establish an intra-day real-time optimal scheduling layer based on the model reinforcement learning on the basis of the pre-day scheduling layer and output an intra-day real-time optimal scheduling plan.
The one or more of the above technical solutions have the following beneficial effects:
In order to overcome the limitation of the traditional single-layer deterministic optimization scheduling method, the technical scheme of the invention creatively provides a multi-layer building micro-grid management framework. The framework integrates a data-driven uncertainty set modeling technology and a three-stage robust optimization scheduling strategy with multi-time scale characteristics, and ensures the comprehensive treatment of source load uncertainty in a day-ahead scheduling stage. Meanwhile, the intra-day real-time scheduling method based on model reinforcement learning is combined, and the response capability of the system to prediction deviation and the scheduling flexibility are further improved.
In order to accurately describe source load uncertainty, aiming at the correlation characteristic between source load prediction deviation and a predicted value, the technical scheme of the invention provides a modeling method for a data driving source load uncertainty set. Secondly, in a day-ahead scheduling layer, aiming at the multi-time scale characteristics of source load equipment and source load uncertainty of a building micro-grid, the invention provides a day-ahead scheduling layer based on three-stage robust optimal scheduling, firstly, a source load uncertainty set is decomposed into two parts of large fluctuation amplitude, long time scale, small fluctuation amplitude and short time scale, secondly, a nominal scheduling plan of all source load equipment is calculated in a first stage, a chase scheduling plan of the source load equipment with the long time scale characteristics and the short time scale characteristics is calculated in a second scheduling stage and a third scheduling stage respectively based on the source load uncertainty set with the long time scale and the short time scale characteristics, and finally day-ahead scheduling is summarized. In a real-time scheduling layer in the day, aiming at the random characteristic presented by the source load prediction deviation, the invention provides a real-time optimization scheduling model in the day based on model reinforcement learning, and the real-time scheduling result in the day is calculated by utilizing the scheduling result before the day so as to ensure the stable and efficient operation of a building micro-grid system.
Additional aspects of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a schematic overall flow chart of a method according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of a method according to an embodiment of the invention.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention.
Embodiments of the invention and features of the embodiments may be combined with each other without conflict.
Example 1
The embodiment discloses a robust optimal scheduling method for a building micro-grid, which can effectively cope with source load uncertainty in the building micro-grid and comprises the following steps:
step one, establishing a source load uncertainty set with correlation information based on data driving.
And capturing the conditional correlation of the prediction deviation and the prediction value by using the probability information of the source load historical data and the corresponding prediction deviation. And establishing a source load uncertainty set based on a conditional normal connection method according to the latest source load predicted value by utilizing the extracted correlation information, so that the source load uncertainty set has stronger data adaptability to the change of a predicted result.
In an embodiment, in order to effectively solve the problem of uncertainty of source load in the optimized dispatching of the building micro-grid, a robust optimized dispatching method is necessary. In this process, a set of source load uncertainties needs to be constructed first. However, conventional methods based on probabilistic statistics and optimization theory, while taking into account worst-case source load uncertainties, tend to be too conservative, may not be adequate to meet the flexibility and efficiency requirements of practical applications. Because of a certain correlation between the source load prediction bias and the predicted value, the correlation is characterized in that the prediction bias is smaller and distributed intensively at low prediction power, and the prediction bias is larger and distributed decentrally at high prediction power.
For this feature, the present embodiment proposes a data-driven source load uncertainty set modeling method. According to the method, the probability information of the source load historical data is utilized to capture the condition correlation between the prediction deviation and the prediction value, so that the adaptability of a source load uncertainty set to the change of the prediction result is improved, and the conservation is reduced.
First, collecting the historical true value of the source loadPredicted value predicted by source load prediction model based on depth learningThe method comprises the steps of collecting historical true values and corresponding predicted values of load power values of photovoltaic power generation, wind power generation, electric vehicle batteries, heating ventilation air conditioning equipment and the like, and obtaining combined accumulated distribution of source load prediction deviation and the predicted values thereof according to a normal connection theory.
(1)
Wherein, The method comprises the steps of predicting the joint cumulative distribution of the deviation and the predicted value of the source load; Is the source load prediction bias, where ;AndThe intermediate variables are normally connected and are obtained from the source load history true values and the corresponding predicted values respectively; And Respectively obtaining accumulated distribution functions from the source load historical true value and the corresponding source load predicted value; is a standard normal cumulative distribution function obtained from the source load history true value and the corresponding source load predicted value, Is an inverse function thereof; And Probability density functions obtained from the source load historical true value and the corresponding source load predicted value respectively; For normally connecting intermediate variables AndProbability density functions of a standard multivariate normal distribution of (c),A covariance matrix for the probability density function.
The source load data refer to load power values of photovoltaic power generation, wind power generation, an electric vehicle battery, heating ventilation air conditioning equipment and the like, wherein the photovoltaic power generation and the wind power generation are source sides, and the load power values of the electric vehicle battery, the heating ventilation air conditioning equipment and the like are load sides.
Definition of the definition,,,Vector composed of normal connected intermediate variables obtained from source load history real value and corresponding predicted value, andI.e. obeying the covariance matrix asIs a normal distribution of (c).
(2)
(3)
Wherein the method comprises the steps ofAndThe linear correlation coefficient and the spearman correlation coefficient of normal connected intermediate variables are used for extracting the correlation between the source load history true value and the predicted value,AndComputing the same as。
And secondly, establishing a conditional probability relation between the source load prediction deviation and the source load prediction value through a normal connection model.
(4)
Wherein, A conditional probability density function between the source load prediction bias and the source load prediction value; The distribution characteristics of the source load predicted values are described; is a conditional distribution of a multielement normal distribution, meets the following conditions Conditional correlation between source load predicted values and true values is described.
(5)
Wherein, As a mean value vector of the data set,In the form of a covariance matrix,Describing the distribution characteristics between the source charge realism values and the predicted values,Describing the distribution characteristics of the source load predictors,Describing the distribution characteristics of the source charge realism values.
A conditional correlation between a source load predictor and a true value is described, with which a number of source load samples can be generated based on the latest day-ahead source load predictor, representing possible realization values of the source load on the next day, thereby constructing a set of source load uncertainties. The specific process is as follows:
1) Obtaining the latest day-ahead source load predicted value to obtain ;
2) Acquisition by means of (5)Is used for the expected value and covariance matrix;
3) By aligning Sampling to generate a plurality ofIs then passed throughA sample of the source charge real value is obtained.
Based onTime sampling acquired sampleAnd the latest day-ahead predictive value(Is the day-ahead scheduling period), a set of source load uncertainties is established.
(6)
Wherein the sampled vectorThe method has the distribution characteristics of prediction deviation under different day-ahead predicted values, so that the constructed source load uncertainty set has strong data adaptability to the change of the predicted result, and the uncertainty of the source load can be described more accurately.In order to predict the magnitude of the fluctuation of the deviation,Is a variable which is 0 to 1,AndTogether describe the degree of conservation of the source charge uncertainty set.
And step two, establishing a day-ahead scheduling layer based on three-stage robust optimal scheduling.
Firstly, dividing a source load uncertainty set into two parts of large fluctuation amplitude, a long time scale and a short time scale, and reducing conservation, wherein the large fluctuation amplitude and the small fluctuation amplitude refer to the large fluctuation amplitude of the prediction deviation, and the long and short time scales refer to the long and short time scale characteristics of source load side equipment. The method comprises the steps of establishing a three-stage robust optimization scheduling model in the first stage, wherein the first stage aims at minimizing the running cost of a building micro-grid under a source load predicted value, calculating a nominal scheduling plan for all source load devices, the second stage aims at maximizing the minimum cost of the building micro-grid in a source load uncertain set with large fluctuation amplitude and long time scale based on the nominal scheduling plan, and obtaining a top-down scheduling plan of the source load devices with long time scale characteristics, and the third stage aims at maximizing the minimum cost of the building micro-grid in the source load uncertain set with small fluctuation amplitude and short time scale based on the nominal scheduling plan and the top-down scheduling plan of the second stage. Finally, summarizing the scheduling plans of the three stages, and sending the summarized scheduling plans to an underlying intra-day real-time optimized scheduling layer as a pre-day scheduling plan.
In a real-time two-layer building micro-grid management framework in the day-ahead and day with the cooperation of multiple time scales, a robust optimal scheduling method is applied in the day-ahead stage to pre-process possible source load fluctuation. In the day-ahead dispatching layer, aiming at the multi-time scale characteristic of source load side equipment and source load uncertainty of a building micro-grid, the characteristic cannot be effectively utilized by the traditional day-ahead two-stage robust optimization dispatching. Firstly, dividing a source load uncertainty set into two parts of large fluctuation amplitude, long time scale, small fluctuation amplitude and short time scale, and reducing conservation. And secondly, establishing a three-stage robust optimal scheduling model before the day.
In a building micro-grid, the response speed of source load side equipment shows obvious multi-time scale characteristics. For example, gas turbines and hvac equipment have relatively slow response speeds due to thermal inertia and regulatory process limitations, with long time scale characteristics. Meanwhile, the energy storage battery, the electric automobile battery, the photovoltaic battery and the wind turbine unit show quick response capability, can quickly adapt to load change no matter in low power or high power output, and have the characteristic of short time scale. The accuracy of the source load prediction is inversely proportional to the prediction time scale, namely the longer the prediction time is, the larger the prediction deviation is, the shorter the prediction time is, the smaller the deviation is, and the management strategy of the building micro-grid needs to adjust the processing of the source load prediction deviation according to the response speed of the equipment. Specifically, for slower responding devices, strategies may be designed to handle larger prediction bias, while for faster responding devices, the response may be fast and adjusted to accommodate smaller prediction bias. Based on the method, the source load uncertainty set is further subdivided into two parts, wherein one part is in the case of large fluctuation amplitude and long time scale, and the other part is in the case of small fluctuation amplitude and short time scale. This subdivision helps to more accurately manage the uncertainty of the source load side equipment, optimize the overall scheduling strategy of the building micro-grid, and equations (7) and (8) are larger and smaller sets of source load uncertainties, respectively.
(7)
(8)
、Larger and smaller source load uncertainty sets, respectively; And The greater and lesser source load uncertainty concentrations, respectivelyThe moment source load value vector is the load power value vector of photovoltaic power generation, wind power generation, electric automobile batteries, heating ventilation air conditioning equipment and the like; Is that The predicted value vector of the source load at moment, namely the predicted value vector of the load power of photovoltaic power generation, wind power generation, electric automobile batteries, heating ventilation air conditioning equipment and the like; the fluctuation amplitude for describing the source load prediction deviation in the larger and smaller source load uncertainty sets is respectively in the range of ;For 0-1 variable, in the larger and smaller source load uncertainty sets, respectively, for determining the uncertainty setWhether positive and negative source load prediction deviation exists at any moment; Uncertainty parameters in the larger and smaller source load uncertainty sets represent the total time period number of the source load prediction deviation, and the value range is ,;AndThe conservation degree of the source load uncertainty set is described together, and the larger the value is, the worse the source load uncertainty scene is, so that the more conservative the scheduling plan is.
The optimization objective of the three-stage robust optimization scheduling model before day is given by the following equation (9). In the first stage, we determine a scheduling plan under the source load predictor nominal scenario, whose optimization objectives, constraints, and decision variables are defined by equations (10) - (12), respectively.
(9)
(10)
(11)
(12)
Wherein, Optimizing the targets for the first, second and third stages,Is a first, a second and a third stage decision vector,For the second and third stages in larger and smaller part of source charge uncertainty setIs a source load vector in (a).
First stage operation costComprises a gas turbine start-stop cost, a gas turbine operation and maintenance cost, an energy storage battery operation and maintenance cost and a power purchase cost,Is the cost of single start-stop of the gas turbine,AndThe unit operation and maintenance cost for the gas turbine and the energy storage battery,Is thatThe electricity selling price of the power grid at any moment,Is the first stageThe starting and stopping state of the gas turbine is carried out at the moment,Is the first stageThe output power of the gas turbine at the moment,Is the first stageThe charging and discharging power of the energy storage battery is always equal,Is the first stagePurchase power to the power grid at any time; Time scales are scheduled for the day before.
The first stage constraints include power balance constraints, gas turbine operating constraints, grid constraints, energy storage battery operating constraints, and load aggregator constraints, including power balance constraints and gas turbine operating constraints in equation (11) above,Is the first stageThe output power of the photovoltaic power generation is at the moment,Is the first stageThe wind power generation output power at the moment,Is the first stageThe electric power is used by the load at the moment; is the first stage Heating ventilation and air conditioning of a moment load aggregator and lumped load of an electric automobile. Equation (12) is a first stage decision vector.
And in the second stage, the minimum pursuit cost of the building micro-grid in the worst scene of the uncertainty concentration of the larger part of source load is used as a target, and the pursuit plans of the lumped loads of the gas turbines, the energy storage batteries, the heating ventilation air conditioner of the load aggregator and the electric automobile with different response speeds are calculated. Equations (13) - (15) below are second stage optimization objectives, new constraints and decision variables, respectively, relative to the first stage.
(13)
(14)
(15)
Wherein, the second stage recoils the costComprises the following cost of a gas turbine, the following cost of an energy storage battery and the compensation cost of a building micro-grid on the lumped load adjustment behavior of a heating ventilation air conditioner of a load aggregator,Is thatThe unit power compensation cost of the building micro-grid to the total load of the heating ventilation air conditioner of the load aggregator is realized at any time,、、、In the second stage for each device of source loadThe second stage is to utilize the dispatching capability of the load aggregation heating, ventilation and air conditioning aggregate load, the second stage will not exactly meet the initial temperature setting of the user, so the formula (14) is the newly added constraint of the second stage relative to the first stage,The method is characterized in that the method comprises the steps of (1) a first stage decision variable is adopted in the formula (15), and the first stage decision variable is adopted in the formula (15), wherein the first stage decision variable is used for adjusting the starting and stopping states of a gas turbine and the purchase power of a power grid.
And on the basis that the second-stage chase plan is overlapped with the first-stage nominal plan, the third-stage chase plan for calculating the lumped load of the energy storage battery and the load aggregator electric automobile with higher response speed is calculated by taking the minimum chase cost of the maximum building micro-grid under the condition of the uncertainty concentration worst of the small partial source load as a target, and the following formulas (16) - (18) are respectively the third-stage optimization target, the newly added constraint relative to the second-stage and the decision variable.
(16)
(17)
(18)
Wherein, the cost of the pursuit in the third stageComprises the following cost of an energy storage battery and the compensation cost of a building micro-grid to the lumped load adjustment behavior of the electric automobile of a load aggregator,Is thatThe unit power compensation cost of the building micro-grid to the lumped load of the electric automobile of the load aggregator is realized at any time,、、In the third stage for each device of source loadThe third stage is to utilize the dispatching capability of the load aggregation electric automobile to aggregate load, and the third stage will not exactly meet all the target charge amount, so the formula (17) is to increase the constraint of the third stage relative to the second stage,、An upper limit value and a lower limit value of a lumped target charge amount of the electric vehicle;、 Is the first Upper and lower limits of the deviation of the target charge amount tolerable to the individual electric car users. Equation (18) is a second stage decision variable, and the third stage will not adjust the output power of the gas turbine and the electric power for the lumped load of the heating ventilation air conditioner.
And thirdly, establishing a real-time daily optimal scheduling layer based on model reinforcement learning.
Firstly, constructing a real-time daily optimization scheduling problem as a Markov decision process, learning a building micro-grid system state transition model by utilizing LSTM, and secondly, learning a real-time daily optimization scheduling optimal strategy based on actor-critic network and fusing a model predictive control algorithm by utilizing the daily front scheduling plan obtained in the step two as a reference to compensate the actual source load prediction deviation.
Because the source load prediction deviation is reduced along with the shortening of the prediction time, the deviation of a day-ahead scheduling plan occurs when the day-ahead scheduling plan operates in the day, and therefore different source load side devices need to be adjusted to ensure the stable operation of the building micro-grid. Because of the random characteristic presented by the source load prediction deviation, the traditional method based on mathematical programming is insufficient to cope with the random characteristic, so that the running deviation cannot be effectively eliminated, the reinforcement learning is based on the randomness of the Markov decision process to the environment effectively.
Firstly, the real-time daily optimization scheduling problem is constructed into a Markov decision process, and states, actions and rewards are defined. The state vector is defined as the following formula (19).
(19)
Wherein, The schedule time variable is optimized for real-time during the day,The schedule time scale is optimized for real-time throughout the day,,,Is thatThe real-time prediction value of the time source lotus in the day,The initial electric quantity value is lumped for the electric automobile; Is that The energy stored by the energy storage battery is stored at any time,Results are scheduled for the day before.
Defining motion vectorsIs of the formula (20):
(20)
Wherein, For each device of source loadAnd the corresponding power adjustment amount at the moment.
The defined reward function is the following formula (21):
(21)
Wherein, And (3) withRespectively isPunishment cost of electricity purchase violations of a unit power grid and punishment of reduction of unit photovoltaic power generation,Is thatThe unit power compensation cost of the building micro-grid to the lumped load of the electric automobile of the load aggregator is realized at any time,Is thatThe unit power compensation cost of the building micro-grid to the total load of the heating ventilation air conditioner of the load aggregator is realized at any time,Is thatAnd the electricity selling price of the power grid at moment.
Learning building micro-grid system state transition model by LSTMThe input to the state transition model is the stateAnd actionsOutput is state change quantity。
(22)
Fusion model predictive control algorithm concept, and learning deterministic strategyThe actor network and the learned building micro-grid system state transition model interact in a mode of optimizing a limited prediction interval, accumulated rewards in the limited prediction interval obtained by interaction are used as an updating target of the critic network, network model parameters actor and critic are continuously updated, and finally an optimal control strategy is obtained。
(23)
(24)
(25)
Wherein, For critic networks, equation (23) is the cumulative rewards within the predicted interval of the H-step model prediction, equation (24) is the loss function of critic networks, and equation (25) is the deterministic policy gradient of actor networks.
Finally, after obtaining the predicted daily load value every 5 minutes, the dispatching center uses actor networkAnd obtaining a real-time scheduling plan in the day.
(26)
Wherein, For real-time optimization of the scheduling plan during the day,For a day-ahead dispatch plan,And outputting a real-time optimal scheduling plan in the day for actor networks.
The reinforcement learning state vector (formula 19) includes the day-ahead schedule of the second stage, and the day-ahead schedule obtained by formula 26 is the sum of the day-ahead schedule and the result obtained by the day-ahead real-time optimization.
In the embodiment, a more accurate uncertainty set is established based on data driving, and the embodiment establishes a source load uncertainty set based on a data driving method, captures the condition correlation between a prediction error and a prediction value by utilizing probability information and prediction deviation of historical data, so that the source load uncertainty set can be more flexibly adapted to the change of a source load prediction result, the uncertainty set is more accurate, and the conservation type of the uncertainty set established by the traditional method is reduced.
In the embodiment, the robustness of the system is improved based on the day-ahead dispatching of three-stage robust optimization, a source load uncertainty set is decomposed into two parts with large fluctuation amplitude and small fluctuation amplitude aiming at the multi-time scale characteristics of source load side equipment and source load uncertainty of a building micro-grid, and a day-ahead three-stage robust optimal dispatching model is established based on the source load uncertainty set, so that compared with the traditional day-ahead two-stage robust optimal dispatching, the method and the system utilize the multi-time scale characteristics of the source load side equipment and the source load uncertainty of the building micro-grid, greatly reduce conservation and improve the robustness.
In the embodiment, the scheduling deviation is eliminated based on the intra-day real-time optimization scheduling of the model reinforcement learning, and the scheduling efficiency is improved. Because the source load prediction deviation is reduced along with the shortening of the prediction time scale, the deviation of the day-ahead scheduling plan occurs when the day-ahead scheduling plan operates in the day, and therefore different source load side devices need to be adjusted to ensure the stable operation of the building micro-grid. Because of the random characteristic presented by the source load prediction deviation, the traditional method based on mathematical programming is insufficient to cope with the random characteristic, so that the running deviation cannot be effectively eliminated, the reinforcement learning is based on the randomness of the Markov decision process to the environment effectively.
Example two
It is an object of the present embodiment to provide a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, which processor implements the steps of the above method when executing the program.
Example III
An object of the present embodiment is to provide a computer-readable storage medium.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the above method.
Example IV
The purpose of this embodiment is to provide a robust optimization scheduling system for a micro-grid of a building, including:
a source load uncertainty set construction module configured to generate a plurality of source load samples based on the day-ahead source load predicted value, representing possible realization values of the source load on the next day, thereby constructing a source load uncertainty set;
The day-ahead scheduling layer establishing module is configured to establish a day-ahead scheduling layer and comprises the steps of dividing a source load uncertainty set into a larger part of source load uncertainty set, a smaller part of source load uncertainty set and establishing a day-ahead three-stage robust optimal scheduling model, wherein the day-ahead three-stage robust optimal scheduling model comprises:
determining a scheduling plan under a source load predicted value nominal scene;
The second stage is to calculate the pursuit plan of the heating, ventilation, air conditioning and electric vehicle lumped load of the gas turbine, energy storage battery and load aggregator with different response speed with the aim of maximizing the minimum pursuit cost of the building micro-grid in the worst scene of the uncertainty concentration of the larger part of source load;
The second-stage pursuit plan is overlapped on the basis of the first-stage nominal plan, the third stage aims at maximizing the minimum pursuit cost of the building micro-grid in the condition of uncertainty concentration of small partial source load and worst, and calculates the pursuit plan of the lumped load of the energy storage battery and the electric vehicle of the load aggregator, which has higher response speed;
And the intra-day real-time optimal scheduling module is configured to establish an intra-day real-time optimal scheduling layer based on the model reinforcement learning on the basis of the pre-day scheduling layer and output an intra-day real-time optimal scheduling plan.
Example five
It is an object of the present embodiments to provide a computer program product comprising instructions which, when run on a computer, cause the computer to perform the methods and functions involved in any of the embodiments described above.
The steps involved in the apparatus of the above embodiment correspond to those of the first embodiment of the method, and the detailed description of the embodiment refers to the relevant description of the first embodiment. The term "computer-readable storage medium" shall be taken to include a single medium or multiple media that includes one or more sets of instructions, and shall also be taken to include any medium that is capable of storing, encoding or carrying a set of instructions for execution by the processor and that cause the processor to perform any one of the methodologies of the present invention.
It will be appreciated by those skilled in the art that the modules or steps of the invention described above may be implemented by general-purpose computer means, alternatively they may be implemented by program code executable by computing means, whereby they may be stored in storage means for execution by computing means, or they may be made into individual integrated circuit modules separately, or a plurality of modules or steps in them may be made into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.
Claims (5)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411266152.6A CN118798584B (en) | 2024-09-11 | 2024-09-11 | A method, system, medium and device for robust optimization scheduling of building microgrid |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411266152.6A CN118798584B (en) | 2024-09-11 | 2024-09-11 | A method, system, medium and device for robust optimization scheduling of building microgrid |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118798584A CN118798584A (en) | 2024-10-18 |
| CN118798584B true CN118798584B (en) | 2025-02-07 |
Family
ID=93033949
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411266152.6A Active CN118798584B (en) | 2024-09-11 | 2024-09-11 | A method, system, medium and device for robust optimization scheduling of building microgrid |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118798584B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119024705B (en) * | 2024-10-29 | 2025-01-24 | 山东理工大学 | A robust load control method based on model prediction and multi-time scales |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019165701A1 (en) * | 2018-02-28 | 2019-09-06 | 东南大学 | Random robust coupling optimization scheduling method for alternating-current and direct-current hybrid micro-grids |
| CN114336749A (en) * | 2021-12-30 | 2022-04-12 | 国网北京市电力公司 | A distribution network optimization method, system, device and storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110571789B (en) * | 2018-06-06 | 2020-10-20 | 南京工程学院 | Electric heating air network three-stage scheduling method based on wind power uncertainty under data driving |
| CN114389262B (en) * | 2022-01-21 | 2024-01-09 | 合肥工业大学 | Regional power grid dispatching method based on robust optimization in elastic environment |
| CN115659096A (en) * | 2022-09-28 | 2023-01-31 | 南京信息工程大学 | Micro-grid multi-time scale energy scheduling method and device considering source load uncertainty |
| CN115912506A (en) * | 2022-11-30 | 2023-04-04 | 国网江苏省电力有限公司常州供电分公司 | Multi-time-stage economic dispatching method and system for micro-grid based on new energy uncertainty |
-
2024
- 2024-09-11 CN CN202411266152.6A patent/CN118798584B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019165701A1 (en) * | 2018-02-28 | 2019-09-06 | 东南大学 | Random robust coupling optimization scheduling method for alternating-current and direct-current hybrid micro-grids |
| CN114336749A (en) * | 2021-12-30 | 2022-04-12 | 国网北京市电力公司 | A distribution network optimization method, system, device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118798584A (en) | 2024-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Megahed et al. | Energy management in zero-energy building using neural network predictive control | |
| JP2023129546A (en) | System and method for optimal control of energy storage system | |
| CN111682536B (en) | Random-robust optimization operation method for virtual power plant participating in dual market before day | |
| CN118798584B (en) | A method, system, medium and device for robust optimization scheduling of building microgrid | |
| CN118412863A (en) | Optical storage integrated electric energy scheduling method, system, electronic equipment and storage medium | |
| Wang et al. | Optimal scheduling for integrated energy system considering scheduling elasticity of electric and thermal loads | |
| CN118899831A (en) | Energy management control method and device for optimizing operation of source, grid, load and storage | |
| CN116384039A (en) | Intelligent power grid energy optimization efficient management method based on model prediction | |
| CN118783550A (en) | A multi-time scale optimization scheduling method and system for hydrogen-electricity coupled microgrid | |
| CN115544856A (en) | A Day-Ahead Optimal Scheduling Method for Electric-Heat Integrated Energy System | |
| CN106786702B (en) | A full-state model predictive energy dispatch method for hybrid energy storage systems | |
| CN108683211A (en) | A kind of virtual power plant combined optimization method and model considering distributed generation resource fluctuation | |
| Ren et al. | Bayesian network based real-time charging scheduling of electric vehicles | |
| CN117833351A (en) | A new energy power system operation risk online assessment method and system | |
| Cheng et al. | Reinforcement learning-based scheduling of multi-battery energy storage system | |
| Li et al. | A two-stage optimal operation strategy for community integrated energy system considering dynamic and static energy characteristics | |
| Horichenko et al. | Research and Software Implementation of Intelligent Method of Energy Consumption Control. | |
| Tian et al. | An Attention-Based Reinforcement Learning Approach For Electric Vehicle Charging Control | |
| CN118504417B (en) | A reinforcement learning optimization scheduling method and system considering scheduling experience | |
| CN119848552B (en) | Hydrogen energy transregional space-time prediction method and system based on transfer learning | |
| CN119444502B (en) | Comprehensive energy system planning optimization method and system based on deep reinforcement learning | |
| Rani et al. | Energy Management System for Electric Vehicle Charging Station using Deep Learning and Model Predictive Control | |
| Sheryar | Reinforcement learning for building energy system control in multi-family buildings | |
| Yang et al. | Optimal Operation of Integrated Energy System Considering Carbon Emission: A Case Study with Deep Reinforcement Learning | |
| Xie et al. | An Adaptive Load Baseline Prediction Method for Power Users as Virtual Energy Storage Elements |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |