CN111262262A - Filter of energy storage equipment power optimization management system based on machine learning - Google Patents

Filter of energy storage equipment power optimization management system based on machine learning Download PDF

Info

Publication number
CN111262262A
CN111262262A CN202010156821.XA CN202010156821A CN111262262A CN 111262262 A CN111262262 A CN 111262262A CN 202010156821 A CN202010156821 A CN 202010156821A CN 111262262 A CN111262262 A CN 111262262A
Authority
CN
China
Prior art keywords
soc
filter
slds
machine learning
power
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.)
Granted
Application number
CN202010156821.XA
Other languages
Chinese (zh)
Other versions
CN111262262B (en
Inventor
胡俊杰
周华嫣然
周羿宏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
North China Electric Power University
Original Assignee
North China Electric Power University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by North China Electric Power University filed Critical North China Electric Power University
Priority to CN202010156821.XA priority Critical patent/CN111262262B/en
Publication of CN111262262A publication Critical patent/CN111262262A/en
Application granted granted Critical
Publication of CN111262262B publication Critical patent/CN111262262B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/28—Arrangements for balancing of the load in a network by storage of energy
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60L—PROPULSION OF ELECTRICALLY-PROPELLED VEHICLES; SUPPLYING ELECTRIC POWER FOR AUXILIARY EQUIPMENT OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRODYNAMIC BRAKE SYSTEMS FOR VEHICLES IN GENERAL; MAGNETIC SUSPENSION OR LEVITATION FOR VEHICLES; MONITORING OPERATING VARIABLES OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRIC SAFETY DEVICES FOR ELECTRICALLY-PROPELLED VEHICLES
    • B60L58/00—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles
    • B60L58/10—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles for monitoring or controlling batteries
    • B60L58/12—Methods or circuit arrangements for monitoring or controlling batteries or fuel cells, specially adapted for electric vehicles for monitoring or controlling batteries responding to state of charge [SoC]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/28—Arrangements for balancing of the load in a network by storage of energy
    • H02J3/32—Arrangements for balancing of the load in a network by storage of energy using batteries with converting means
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/60—Other road transportation technologies with climate change mitigation effect
    • Y02T10/70—Energy storage systems for electromobility, e.g. batteries

Abstract

The invention provides a filter of an energy storage equipment power optimization management system based on machine learning, which can filter an SOC instruction which does not meet the safety, filter an SOC instruction which has volatility, and filter an SOC instruction which does not meet the user satisfaction. After the filter is applied to generating a scheduling scheme based on a machine learning technology, only logic judgment and assignment statements need to be added in an original program during specific implementation, so that the occupied memory is small, the influence on the time for generating a result is small, and the obvious delay is not brought to the online running time of a power optimization management system; it is not dependent on specific machine learning mode and type of SLDs, is designed for multiple SLDs; the steps are relatively independent, and can be correspondingly adjusted according to the use requirement of the system.

Description

Filter of energy storage equipment power optimization management system based on machine learning
Technical Field
The invention belongs to the field of optimal scheduling of power systems, and particularly relates to a filter of an energy storage device power optimal management system based on machine learning.
Background
With the progress of computer computing power in recent years, intelligent computing technologies applying data driving models such as machine learning and the like are applied to optimization and computation of various power systems in a large scale by combining big data technologies. By mining information among mass data, the problem with rapidity and certain accuracy can be solved by applying a machine learning technology.
Nowadays, machine learning techniques are also primarily applied to power optimization management of Storage-like devices (SLDs). SLDs include conventional Energy Storage Systems (ESS) and other devices with energy storage properties, typically Electric Vehicles (EV). Because SLDs have strong charging and discharging flexibility, they are widely applied to the response of the demand side of the power system, and the purposes of saving power cost and optimizing load curve can be achieved by guiding the output of SLDs through the price of electricity.
The application of SLDs power optimization management based on a machine learning technology is mainly ESS, and EV is rare, and online power management of SLDs is realized by modeling an energy storage equipment power optimization management system and solving by using the machine learning technology. The solving mode of the machine learning technology is as follows: and off-line learning a mapping relation between the input quantity based on the historical data and the decision variable, and determining the power of the SLDs on line by using the mapping relation. The decision variable is typically the power or state of charge (SOC) of the SLDs. The offline learning process is established, so that the SLDs power optimization management system does not need to directly solve the bottom layer optimization problem during online decision making, but the long-term state of the system can be considered, the scheduling can be quickly executed, and the calculation burden of the system during online calculation is greatly reduced.
However, in the existing preliminary application, the decision performance is not yet perfect due to the following three aspects.
Firstly, the decision result based on machine learning has certain randomness, and the decision result may have certain fluctuation during online decision. Taking the SOC as an example of the decision variable, when the SOC curve has some slight fluctuation, the slight difference will directly affect the positive and negative properties and the magnitude of the power, thereby resulting in an unsatisfactory result of optimizing the power.
Secondly, the satisfaction of some safety constraints can not be guaranteed. The safety constraints of the system comprise upper and lower power limits, upper and lower SOC limits and the like, which are formed according to the learning of data in the historical optimal solution when power or SOC instructions are finally issued in the solving process of machine learning, although mandatory safety constraints are added in the optimal solving process of historical data, the learning effect of machine learning cannot achieve one hundred percent of recognition and meet the safety constraints, and although the performance of machine learning is better at present, the result basically does not deviate from the safety constraints, the possibility of boundary crossing still exists under the condition of lacking of the mandatory constraints.
Thirdly, some constraints related to user satisfaction cannot be guaranteed. User satisfaction constraints may generally include: the electric quantity demand is restricted, for example, the electric quantity meeting the driving demand is required to be reached after the electric automobile is charged; or some specific user-set personalized constraints such as the energy storage system power returning to 0.5 at the end of each day to ensure the next day of application with up-down adjustment margin. Similar to the reason of the point 2, through the learning of massive data, the decision result of machine learning may be able to satisfy the constraints, but a mandatory constraint is still lacked to guarantee. It is also important to satisfy such constraints considering that the demand and satisfaction of users in demand-side response projects are related to the enthusiasm of users for scheduling.
In summary, in order to ensure that the border crossings and fluctuations do not occur, the invention designs a filter for the energy storage device power optimization management system based on machine learning, so as to improve the effectiveness and reliability of the energy storage device power optimization management system based on machine learning.
Disclosure of Invention
The invention aims to provide a filter for restraining the fluctuation of decision results and avoiding the safety boundary crossing and the user satisfaction degree restraint boundary crossing aiming at the energy storage equipment power optimization management system based on machine learning, thereby improving the effectiveness and the reliability of the energy storage equipment power optimization management system based on machine learning.
A filter for a machine learning based energy storage class device power optimization management system, the filter filtering state of charge (SOC) commands by:
a, filtering an SOC instruction which does not meet the safety;
b, filtering the SOC instruction with volatility in the part;
and C, filtering the SOC instruction which does not meet the user satisfaction degree.
And step A, filtering the original output SOC into the SOC corresponding to the requirement of the safety constraint limit value. And step B, filtering the original output SOC into an SOC which is equal to the output SOC in the previous period. And step C, filtering the original output SOC into the SOC corresponding to the rated power charging.
Step A comprises filtering instructions that do not satisfy SOC security constraints and filtering instructions that do not satisfy power security constraints; if the output SOC of any one of the energy storage devices (SLDs) exceeds the upper limit or the lower limit of the SOC, keeping the maximum or minimum SOC; if the output power of any SLDs exceeds the SOC upper limit or lower limit, the SLDs are kept at the upper limit or lower limit power.
The step B comprises the following steps: filtering discharge instructions at peak-valley electricity price off-peak time for SLDs with electric quantity requirements at the end of scheduling; filtering charging instructions at peak-valley electricity price peak time for SLDs with electricity quantity requirements at the end of scheduling on the premise of not influencing the electricity quantity requirements; SLDs that have power requirements at the end of the schedule are filtered of such discharge instructions if some discharge activity causes additional charge activity to occur during peak hours to compensate.
And step C, ensuring that the electric quantity demand constraint of the SLDs or some specific personalized constraints set by a user can be met, judging whether the SOC of the SLDs can reach the required level when the scheduling is finished, and if the SLDs are kept to be charged at the maximum charging power from the current time step, namely under the condition of full-speed charging, and cannot reach the required electric quantity level until the scheduling is finished, exiting the scheduling and keeping to be charged at the maximum charging rate from the current time step.
The three filtering steps of the filter are relatively independent, and can be correspondingly adjusted according to the use requirement of the system.
The filtration step C of the filter is carried out after steps a and B.
The invention has the beneficial effects that:
1. the filter designed by the invention is a modification of the scheduling scheme generated based on machine learning in the process, and after the scheduling scheme is generated based on the machine learning technology, only logic judgment and assignment statements need to be added in the original program during specific implementation, so that the occupied memory is small, the influence on the time for generating the result is small, and the obvious delay can not be brought to the online running time of the power optimization management system of the SLDs based on machine learning. The filter is operated for a plurality of times by a system containing 100 SLDs, and the time for finishing the filtration is not in millisecond level. This is important because a large reason for the application of machine learning techniques in SLDs power management systems is to save on-line scheduling time.
2. The adaptability is strong, and the filter is applicable no matter what machine learning technology the SLDs power management system based on machine learning is based on, what types of SLDs are managed, and the number of SLDs. Because the design process of the present filter does not depend on the specific machine learning mode and the type of SLDs, and is designed for a plurality of SLDs.
3. The filter is easy to adjust, and corresponding steps can be added and deleted according to the requirements of different systems. Although the design steps of the filter are specific, the steps are relatively independent, so that the filter can be adjusted according to the use requirement of the system.
Drawings
FIG. 1 is a block diagram illustrating the filtering steps of a filter of a power optimization management system for energy storage devices based on machine learning according to the present invention;
FIG. 2 shows a comparison of the energy storage scheduling SOC before and after the filtering of step A in an embodiment of the present invention;
FIG. 3 shows a power comparison of the online output results of the optimized dispatch simulation system with 100 electric vehicles before and after being filtered by step B and the offline optimization results using the conventional optimization algorithm according to the embodiment of the present invention;
FIG. 4 shows a comparison of SOC curves of 4 electric vehicles extracted in the example of FIG. 3 before and after filtering by step B and an off-line optimization result using a conventional optimization algorithm;
FIG. 5 shows a comparison of the output SOC of the electric vehicle power optimization management system based on machine learning according to the embodiment of the present invention before and after being filtered by step C.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
Machine learning-based energy Storage-like devices (SLDs) power optimization management system generally outputs SOC of ith SLDs at next time step (t +1) on line (at current time step t) with the aim of minimizing electricity consumption cost
Figure BDA0002404364980000051
Or the power (P) of the current time stepi(t)). However, since the output data based on machine learning greatly depends on the learning effect of machine learning on massive data, the generated output may have the following two problems. On the one hand, these outputs may violate some safety constraints of SLDs, including maximum battery power (S)max) Minimum battery charge (S)min) Maximum charge and discharge power
Figure BDA0002404364980000052
Or some SLDs user demand type constraints, such as power demand (S)dep). On the other hand, there may be some fluctuation between outputs due to the randomness of machine learning. For example, when the SOC curve fluctuates slightly due to machine learning randomness, there is a relationship between SOC and power:
Figure BDA0002404364980000053
wherein τ is the length of a time step，ηcAnd ηdCharge and discharge efficiencies, C, of SLDs devices, respectivelyiThe capacity of the ith SLDs device.
As can be seen from equation (1), if some slight fluctuation occurs in SOC, the power will be directly affected, thereby causing an error in decision making. When errors occur simultaneously with many SLDs, the total power error will be significant, which will adversely affect power costs and peak-to-valley regulation. In order to solve the problems, a filter needs to be designed so as to ensure the effectiveness and reliability of the energy storage type device power optimization management system based on machine learning. The filter designed by the invention is a modification of the scheduling scheme generated based on machine learning, and is applied after the scheduling scheme is generated based on the machine learning technology.
The present invention proposes a filter comprising three parts to address the above problems. These three parts are respectively directed to three different constraints. In the first part, it is ensured that no violations of the SLDs battery limit and the charging power limit occur. In the second section, the impact of fluctuations and the criteria for detecting and avoiding fluctuations are specified. And the third part ensures that the personalized requirements of SLDs users can be met.
Considering that a mutual conversion relation exists between the SOC and the power, the energy storage equipment power optimization management system based on machine learning can perform mutual conversion through an expression (1) no matter the output SOC or the output power, and considering that the SOC is stronger in continuity in a time domain and a filter is easy to detect and filter the SOC, the output SOC of the system is selected as a detection object.
The specific filtration method of the present invention is described below, as shown in FIG. 1.
Step A: SOC instructions that do not meet safety are filtered. This step is fundamental, and without this step, a control signal output by the system that lacks security is not feasible.
Step A comprises two parts, namely filtering the SOC instruction which does not meet the constraint of battery electric quantity safety. Considering that the battery capacities of SLDs have upper and lower limits, when the SOC does not satisfy equations (2) - (3), it is considered that the command does not satisfy the battery capacity safety constraint.
Figure BDA0002404364980000061
Figure BDA0002404364980000062
If it is not
Figure BDA0002404364980000063
Order to
Figure BDA0002404364980000064
If it is
Figure BDA0002404364980000065
Order to
Figure BDA0002404364980000066
That is, if the output SOC of any one of the SLDs exceeds the upper or lower battery charge level, it is kept at the maximum or minimum battery charge level.
Another part of step a is to filter SOC commands that do not meet the power safety constraints. The SLDs charge/discharge power also has a limit value due to limitations of line transmission power and the like, and considering the interconversion relationship (1) between the SOC and the power, when the SOC does not satisfy equations (4) to (5), it is considered that the command does not satisfy the power safety constraint.
Figure BDA0002404364980000067
Figure BDA0002404364980000068
If it is not
Figure BDA0002404364980000069
Order to
Figure BDA00024043649800000610
If it is not
Figure BDA00024043649800000611
Order to
Figure BDA00024043649800000612
That is, if any SLDs output power exceeds the SOC upper or lower limit, it is kept at the upper or lower limit power.
And B: the filtering portion has a fluctuating SOC command. The purpose of this step is to filter out fluctuations due to the randomness of the machine learning. To have the concept of fluctuation, it is assumed that there are two kinds of abnormal behaviors, i.e., a charging abnormal behavior and a discharging abnormal behavior, both of which may cause fluctuation. These abnormal behaviors have two negative effects. First, they may cause charging and discharging actions to occur at inappropriate times, which in turn may result in additional electrical costs. Second, even if these abnormal behaviors do not result in additional costs, they may affect peak-to-valley regulation if the cumulative abnormal behavior results in additional power peaks. In view of this, the principle of the invention is that once a charging or discharging behavior causes at least one negative effect as described above, the behavior is considered abnormal, and if abnormal behavior is avoided, the fluctuations will in turn be eliminated. The method specifically comprises the following steps:
SLDs that have power requirements at the end of the schedule are filtered for discharge instructions at peak-to-valley power rates off-peak. For such devices, the discharge behavior may be considered abnormal when the electricity price is not at the peak at peak-to-valley time. This is because the off-peak discharge will cause additional charging behavior to occur at peak times to compensate for this discharging behavior, since the SOC of SLDs must reach a certain level at the end of the schedule
Figure BDA0002404364980000071
The additional charging during the peak period of the electricity price will result in a higher electricity fee. After the abnormal command is detected, the solution of the invention is that the SOC of the next time step is the same as the SOC of the current time step. Let [ t)ps,tpe]Peak value representing peak-to-valley time-of-use electricity priceThe mathematical expression of this step is as follows:
off-peak periods of electricity prices when at peak-valley, i.e.
Figure BDA0002404364980000075
If it is not
Figure BDA0002404364980000072
Then this instruction is a discharge instruction, then
Figure BDA0002404364980000073
And filtering the charging instructions at the peak-valley electricity price and the peak on the premise of not influencing the electricity demand of SLDs with electricity demand when the dispatching is finished. During the peak period of the time-of-use electricity price, the charging is considered as abnormal behavior, and the SOC at the next time step is made to be the same as the SOC at the current time step. As this would result in higher costs than would otherwise be the case. However, it is considered that the power states of some SLDs must reach a certain level at the end of scheduling
Figure BDA0002404364980000074
Charging behavior may inevitably occur at peak times up to this level, and in view of this, such charging behavior will not be considered abnormal. The above criteria may be described as: charging at maximum charging power from the current SOC to the SOC demand must start from a time step during peak-valley time-peak hours of electricity prices. The mathematical expression for this step is as follows:
at t ∈ [ t ]ps,tpe]When, if
Figure BDA0002404364980000081
I.e. the current charge command for a peak period of electricity price, if any
Figure BDA0002404364980000082
Then
Figure BDA0002404364980000083
SLDs with capacity requirements at the end of the schedule filter such discharge instructions if some discharge activity causes additional charge activity to occur during peak hours to compensate. During peak hours of the time of use electricity prices, certain discharge behaviors are considered abnormal if they cause additional charging behaviors to occur during the peak hours to compensate for the SOC reaching a certain level at the end of the schedule. Although this event does not incur additional cost due to charging and discharging occurring during the same power rate, the additional peak-to-valley power still affects peak-to-valley regulation. Therefore, time-shared power, in which the discharge behavior causes additional charge behavior to occur at the peak time, is considered abnormal. The mathematical expression for this step is as follows:
at t ∈ [ t ]ps,tpe]When, if
Figure BDA0002404364980000084
That is, a discharge command currently reaching a peak period of electricity price, if any
Figure BDA0002404364980000085
I.e., this discharge behavior will result in additional charging behavior during peak hours to compensate for the capacity demand at the end of its schedule, then
Figure BDA0002404364980000086
And C: and filtering the SOC instruction which does not meet the user satisfaction degree. It should be ensured that the power requirements of SLDs can be met, including: the electric quantity demand is restricted, for example, the electric quantity meeting the driving demand is required to be reached after the electric automobile is charged; or some specific user-set personalized constraints such as the energy storage system power returning to 0.5 at the end of each day to ensure the next day of application with up-down adjustment margin. This step fully represents a respectful of the user's wishes. This step becomes the last step because, if this step exchanges the order with the preceding steps, the SOC scheduling scheme that satisfies the user requirement obtained after this step ends may not satisfy the user requirement because of filtering adjustment such as eliminating fluctuation.
SOC instructions that do not meet user satisfaction are detected and full speed charging is scheduled. To achieve this, the present invention needs to determine whether the SOC of the SLDs can reach its required level when the schedule is finished, and if the SLDs cannot reach its required charge level until the schedule is finished in the case where the SLDs are kept charging at the maximum charging power from the current time step, i.e., full-speed charging, the SLDs will exit the schedule and keep charging at the maximum charging rate from the current time step. The formula for this step is expressed as:
if it is not
Figure BDA0002404364980000087
Then
Figure BDA0002404364980000088
I.e., full speed charging. Wherein
Figure BDA0002404364980000091
Indicating the time step when the schedule is finished.
The filter of the invention can add or delete corresponding steps according to the requirements of different systems. Although the filtering steps of the filter of the present invention are specific, the steps are relatively independent, and thus can be adjusted according to the use requirements of the system.
The advantages of the filter according to the invention are illustrated below by means of specific examples.
By filtering the SOC instruction which does not meet the SOC and the upper and lower power limits in the step A, the SLDs power optimization management result based on machine learning can be ensured to meet the safety constraint. Fig. 2 shows a comparison of SOC before and after the action of step a for a certain energy storage scheduling SOC. It can be seen that, due to the "empirical" decision of machine learning, the unfiltered slope of the SOC curve in the dashed box is too large, and this SOC control result will be out of limit for power. After the filtering in the step A, the slope of the SOC curve at the stage returns to the value corresponding to the rated charging power.
And B, correcting some fluctuations influencing the scheduling effect, and avoiding the adverse influence of the fluctuations on the scheduling. To verify this, an optimized dispatching system including 100 electric vehicles was simulated, and an online output result of the system using machine learning and an offline power optimization result (i.e., a global optimal solution) using a conventional optimization algorithm were obtained, respectively, as shown in fig. 3. And comparing online output results based on machine learning before and after filtering in the step B, and extracting SOC curves of 4 electric vehicles and total power curves of all vehicle superposed loads, as shown in FIG. 4. It can be seen from fig. 3 that the variability of the total power curve is greatly improved as the step B filters the variability. As can be seen from FIG. 4, the SOC curve after filtering in step B is more flat and more consistent with the global optimal solution.
And C, constraining the scheduling condition which does not meet the user satisfaction degree, so that the SLDs power optimization management result based on machine learning can be ensured to meet the user constraint. Fig. 5 shows the output SOC of the electric vehicle power optimization management system based on machine learning, and if the charging SOC requirement of the electric vehicle is set to 0.85, the electric vehicle cannot reach the electric quantity of 0.85 when leaving the system due to the overlong discharge time scheduled in the original optimization management output result. After filtering in the step C, it can be seen that the EV continuous discharge is detected to be incapable of meeting the electric quantity requirement, and the electric vehicle can finally meet the charging requirement by stopping scheduling and full-speed charging in advance.
The present invention is not limited to the above embodiments, and any changes or substitutions that can be easily made by those skilled in the art within the technical scope of the present invention are also within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (7)

1. A filter for a machine learning based energy storage class device power optimization management system, the filter filtering state of charge (SOC) commands by:
a, filtering an SOC instruction which does not meet the safety;
b, filtering the SOC instruction with volatility in the part;
and C, filtering the SOC instruction which does not meet the user satisfaction degree.
2. The filter of the energy storage device power optimization management system based on machine learning according to claim 1, wherein: the step A filters the original output SOC into the SOC corresponding to the requirement of the safety constraint limit value; the step B filters the original output SOC into an SOC which is equal to the output SOC in the previous period; and C, filtering the original output SOC into the SOC corresponding to the rated power charging.
3. The filter of the energy storage device power optimization management system based on machine learning according to claim 2, wherein: the step A comprises filtering instructions which do not meet the battery power safety constraint and filtering instructions which do not meet the power safety constraint; if the output SOC of any one of the energy storage devices (SLDs) exceeds the upper limit or the lower limit of the battery capacity, keeping the maximum or the minimum battery capacity; if the output power of any SLDs exceeds the SOC upper limit or lower limit, the SLDs are kept at the upper limit or lower limit power.
4. The filter of the energy storage device power optimization management system based on machine learning according to claim 2, wherein the step B comprises: filtering discharge instructions at peak-valley electricity price off-peak time for SLDs with electric quantity requirements at the end of scheduling; filtering charging instructions at peak-valley electricity price peak time for SLDs with electricity quantity requirements at the end of scheduling on the premise of not influencing the electricity quantity requirements; SLDs that have power requirements at the end of the schedule are filtered of such discharge instructions if some discharge activity causes additional charge activity to occur during peak hours to compensate.
5. The filter of a power optimization management system for energy storage devices based on machine learning as claimed in claim 2, wherein the step C ensures that the power demand constraint of SLDs or some specific personalized constraint set by the user can be satisfied, it needs to be determined whether the SOC of SLDs can reach its required level when the schedule is finished, if the SLDs can not reach its required power level until the schedule is finished under the condition that the SLDs are kept charged at the maximum charging power from the current time step, i.e. full-speed charging, then the SLDs will exit the schedule and keep being charged at the maximum charging rate from the current time step.
6. The filter of the energy storage device power optimization management system based on machine learning according to any one of claims 1-5, wherein: the three filtering steps of the filter are relatively independent, and can be correspondingly adjusted according to the use requirement of the system.
7. The filter of the energy storage device power optimization management system based on machine learning according to any one of claims 1-5, wherein: the filtration step C of the filter is carried out after steps a and B.
CN202010156821.XA 2020-03-09 2020-03-09 Filter of energy storage equipment power optimization management system based on machine learning Active CN111262262B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010156821.XA CN111262262B (en) 2020-03-09 2020-03-09 Filter of energy storage equipment power optimization management system based on machine learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010156821.XA CN111262262B (en) 2020-03-09 2020-03-09 Filter of energy storage equipment power optimization management system based on machine learning

Publications (2)

Publication Number Publication Date
CN111262262A true CN111262262A (en) 2020-06-09
CN111262262B CN111262262B (en) 2022-11-29

Family

ID=70953013

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010156821.XA Active CN111262262B (en) 2020-03-09 2020-03-09 Filter of energy storage equipment power optimization management system based on machine learning

Country Status (1)

Country Link
CN (1) CN111262262B (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060055374A1 (en) * 2002-12-11 2006-03-16 Shinji Fujihara Battery charged condition computing device and battery charged condition computing method
CN104716693A (en) * 2013-12-13 2015-06-17 华为技术有限公司 Energy management method, controller and system for distributed energy storage batteries
CN105244869A (en) * 2015-10-13 2016-01-13 国网山东省电力公司电力科学研究院 Dynamic random scheduling control method for power distribution network containing micro-grid
CN107968420A (en) * 2017-12-06 2018-04-27 重庆大学 Energy-storage system and its energy-optimised management method based on distributed extremum seeking algorithm
CN108964050A (en) * 2018-08-26 2018-12-07 燕山大学 Micro-capacitance sensor dual-layer optimization dispatching method based on Demand Side Response
CN110011342A (en) * 2019-04-04 2019-07-12 华北电力大学 A kind of cluster electric car charge-discharge electric power optimum management method

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060055374A1 (en) * 2002-12-11 2006-03-16 Shinji Fujihara Battery charged condition computing device and battery charged condition computing method
CN104716693A (en) * 2013-12-13 2015-06-17 华为技术有限公司 Energy management method, controller and system for distributed energy storage batteries
CN105244869A (en) * 2015-10-13 2016-01-13 国网山东省电力公司电力科学研究院 Dynamic random scheduling control method for power distribution network containing micro-grid
CN107968420A (en) * 2017-12-06 2018-04-27 重庆大学 Energy-storage system and its energy-optimised management method based on distributed extremum seeking algorithm
CN108964050A (en) * 2018-08-26 2018-12-07 燕山大学 Micro-capacitance sensor dual-layer optimization dispatching method based on Demand Side Response
CN110011342A (en) * 2019-04-04 2019-07-12 华北电力大学 A kind of cluster electric car charge-discharge electric power optimum management method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
戴志辉等: "基于实时电价机制的微网多时间尺度能量管理研究", 《华北电力大学学报(自然科学版)》 *
熊书华等: "实时电价下户用储能智能化管理", 《电力系统及其自动化学报》 *

Also Published As

Publication number Publication date
CN111262262B (en) 2022-11-29

Similar Documents

Publication Publication Date Title
CN102496991B (en) Backup lithium ion battery pack management method and management system thereof
CN107565607B (en) Micro-grid multi-time scale energy scheduling method based on real-time electricity price mechanism
JP5350942B2 (en) Supply / demand control device for power system, supply / demand control program, and recording medium thereof
CN105684257A (en) Microgrid energy management system and method for controlling operation of a microgrid
CN110979085B (en) Method, device and equipment for regulating and controlling charging of electric automobile
CN102185375A (en) Electric vehicle energy efficiency evaluation and supervision system and method thereof for intelligent power grid
CN110867881A (en) Energy storage control method and system for optical storage and charging system
CN111917113B (en) Power grid load allowance calculation system and method and charging pile access power distribution method
Meng et al. A control strategy for battery energy storage systems participating in primary frequency control considering the disturbance type
CN111262262B (en) Filter of energy storage equipment power optimization management system based on machine learning
CN110866647A (en) User side energy storage control method, device, equipment and storage medium
CN115166546A (en) Online capacity checking method and device for storage battery of distributed direct-current power supply system for station
CN115360738A (en) Electric automobile primary frequency modulation control method considering controllable domain constraint
CN114194060A (en) Method and device for determining charging station
CN114400689A (en) Power dispatching method and device based on V2G charging station
CN109256790B (en) Energy storage system configuration method and device and storage medium
CN112736894A (en) Two-stage unit combination modeling method considering randomness of wind power and electric automobile
CN112613229A (en) Energy management method and model training method and device for hybrid power equipment
CN110880777A (en) Method and device for evaluating emergency peak regulation standby capacity of energy storage at power grid side
CN116613796A (en) Method, system, equipment and medium for charging and discharging polymer of electric automobile
JP2020162304A (en) Battery control system
CN110962665B (en) Scattered electric vehicle charging coordination method based on local measurement voltage amplitude
Liu et al. Operation-Area-Constrained Adaptive Primary Frequency Support Strategy for Electric Vehicle Clusters
US20230069004A1 (en) Electric vehicle charging station management method using blockchain
EP4250511A1 (en) V2g system, prohram and method

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