CN117207803A - Electric automobile intelligent charging strategy selection method based on economic dispatch - Google Patents
Electric automobile intelligent charging strategy selection method based on economic dispatch Download PDFInfo
- Publication number
- CN117207803A CN117207803A CN202311096583.8A CN202311096583A CN117207803A CN 117207803 A CN117207803 A CN 117207803A CN 202311096583 A CN202311096583 A CN 202311096583A CN 117207803 A CN117207803 A CN 117207803A
- Authority
- CN
- China
- Prior art keywords
- charging
- representing
- cost
- gas turbine
- electric automobile
- 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.)
- Pending
Links
- 238000010187 selection method Methods 0.000 title claims abstract description 8
- 238000010248 power generation Methods 0.000 claims abstract description 64
- 238000000034 method Methods 0.000 claims abstract description 42
- 238000012423 maintenance Methods 0.000 claims description 29
- 238000004519 manufacturing process Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 14
- 238000009434 installation Methods 0.000 claims description 9
- XOFYZVNMUHMLCC-ZPOLXVRWSA-N prednisone Chemical compound O=C1C=C[C@]2(C)[C@H]3C(=O)C[C@](C)([C@@](CC4)(O)C(=O)CO)[C@@H]4[C@@H]3CCC2=C1 XOFYZVNMUHMLCC-ZPOLXVRWSA-N 0.000 claims description 6
- 238000012887 quadratic function Methods 0.000 claims description 6
- 238000004146 energy storage Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 abstract description 6
- 229910052799 carbon Inorganic materials 0.000 abstract description 6
- 238000004422 calculation algorithm Methods 0.000 description 35
- 230000006870 function Effects 0.000 description 34
- 238000010586 diagram Methods 0.000 description 11
- 230000009471 action Effects 0.000 description 10
- 230000007246 mechanism Effects 0.000 description 10
- 238000005457 optimization Methods 0.000 description 10
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 239000003795 chemical substances by application Substances 0.000 description 5
- 238000009826 distribution Methods 0.000 description 5
- 238000005265 energy consumption Methods 0.000 description 5
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 description 4
- 230000002787 reinforcement Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000006399 behavior Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 230000007774 longterm Effects 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 229910002092 carbon dioxide Inorganic materials 0.000 description 2
- 239000001569 carbon dioxide Substances 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000004821 distillation Methods 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000003912 environmental pollution Methods 0.000 description 2
- 230000002035 prolonged effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000029087 digestion Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Landscapes
- Charge And Discharge Circuits For Batteries Or The Like (AREA)
Abstract
The application discloses an intelligent charging strategy selection method for an electric automobile based on economic dispatch, relates to the technical field of electric automobiles, and aims at solving the problem that the electric automobile cannot be charged in time due to the fact that the influence of time-varying road conditions is not considered in the method in the prior art. The application also provides a proper charging strategy for the electric automobile, reduces the charging cost to the maximum extent and reduces the road congestion. The numerical result shows that the application has effectiveness and superiority in actual scenes, and achieves the aim of minimizing the power generation cost and the carbon emission. The application can realize lower charging cost of the electric automobile on the premise of minimizing the power dispatching cost and the carbon emission.
Description
Technical Field
The application relates to the technical field of electric automobiles, in particular to an intelligent charging strategy selection method for an electric automobile based on economic dispatch.
Background
In recent years, from different viewpoints, charging strategies of electric vehicles are also different. In the prior art, an electric vehicle joint charging dispatching optimization framework is provided, and a DRL and Binary Linear Programming (BLP) combined method is adopted to solve the problem, so that an optimal dispatching strategy is obtained. The prior art provides a model-free method based on safety deep reinforcement learning, and aims to find a charge and discharge scheduling strategy with constraint, so that a charge pile can meet the self-charge requirement and simultaneously minimize the charge cost of a user. In the prior art, a self-adaptive control algorithm for charging a plug-in electric automobile is provided, and the aim is to ensure that the charging behavior of the electric automobile does not bring burden to an electric power system in a low-voltage environment. Aiming at the problem of huge charging requirements caused by the increase of the number of plug-in electric vehicles, a real-time charging optimization scheduling method for the plug-in electric vehicles is provided in the prior art, and a parameterized aggregation charging model is established. In order to meet the explosive charging requirement of the electric automobile, the problems of electric automobile charging time optimization, electric automobile and charging pile pairing, charging pile pricing mechanism and the like are comprehensively considered in the prior art, a hierarchical game model is established, and an optimal charging time strategy of the electric automobile is obtained. However, the methods do not consider the influence of time-varying road conditions, which can lead to the failure of timely charging of the electric vehicle.
Disclosure of Invention
The purpose of the invention is that: aiming at the problem that the electric vehicle cannot be charged in time due to the influence of time-varying road conditions in the method in the prior art, the intelligent charging strategy selection method for the electric vehicle based on economic dispatch is provided.
The technical scheme adopted by the invention for solving the technical problems is as follows:
an electric automobile intelligent charging strategy selection method based on economic dispatch comprises the following steps:
step one: acquiring historical charging data of the charging piles, and predicting the power load demands of all the charging piles in the future 24 hours according to the historical charging data of the charging piles;
step two: based on the power load demands of all charging piles and the running state constraint of a motor unit in the future 24 hours, constructing a power plant economic dispatch strategy objective function, wherein the objective function of the power plant economic dispatch strategy is expressed as:
wherein,represents the total power generation cost of the ith wind power generation,/->Represents the ithTotal power generation cost of photovoltaic generator, +.>Representing the total power generation cost of the ith micro gas turbine,/->Indicating the CO released by the ith micro gas turbine 2 P represents the power generation of the unit, including +.>x represents the operating state of the micro gas turbine, T represents the time slot, K 1 ,K 2 ,K 3 ,K 4 The importance weights representing each component, which can be normalized to satisfy +.>
The operating state constraints of the motor group include:
output power constraints, power balance constraints, and micro gas turbine operating state constraints;
the output power constraint is specifically:
wherein,the upper limit of the output power of the wind power generator, the photovoltaic power generator and the micro gas turbine is respectively indicated by +.>Representing the lower limit of the output power of the micro gas turbine, < +.>Respectively representing the power output of the ith wind driven generator, the photovoltaic generator and the micro gas turbine;
the power balance constraint is specifically:
wherein L represents the number of generators, N represents the number of charging piles,representing the power load of the jth charging pile and the power transferred to the energy storage system by the jth charging pile respectively;
the running state constraint of the micro gas turbine is specifically as follows:
x i,t representing the operating state of the micro gas turbine;
step three: designing a first-stage rewarding function by using a power plant economic dispatching strategy objective function, training a first-stage DRL model, dispatching a power plant by using the trained first-stage DRL model, and obtaining a power dispatching schedule;
step four: based on the power dispatching schedule in the third step, obtaining the actual power supply quantity of each charging pile, and constructing an electric vehicle charging strategy objective function by combining the constraint of the electric vehicle in the charging process, wherein the electric vehicle charging strategy objective function is expressed as:
Wherein Δe represents the charge amount of the electric vehicle, y represents which charging pile the electric vehicle selects for charging, M represents the number of electric vehicles, K 5 ,K 6 Representing the importance weight of each component, which may be normalized to satisfyθ represents a cost conversion coefficient, +.>Indicating the degree of road congestion->Representing the charging cost of the electric automobile;
the constraint of the electric automobile in the charging process comprises the following steps:
battery charging constraints, battery power constraints, battery remaining capacity constraints, and charging pile selection constraints;
the battery charging constraint is specifically as follows:
the battery power constraint is specifically:
the battery residual capacity constraint is specifically as follows:
wherein,indicating the residual capacity of the kth electric automobile at the time t,/->Represents the minimum electric quantity required when the kth electric automobile runs to the jth charging pile, delta E k Represents the charge quantity, y of the kth electric automobile j,k Indicating that the kth electric automobile selects the jth charging pile for charging, < >>Indicating the remaining capacity of the charging pile +.>The maximum capacity at time t is shown;
the charging pile selection constraint is specifically as follows:
step five: designing a second-stage rewarding function by utilizing an electric vehicle charging strategy objective function, training a second-stage DRL model, and selecting an electric vehicle charging strategy by utilizing the trained second-stage DRL model;
Step six: and finishing intelligent charging strategy selection of the electric automobile by using the trained first-stage DRL model and the trained second-stage DRL model.
Further, the total power generation cost of the ith wind power generatorExpressed as:
wherein,representing the operation and maintenance costs of the ith wind power generator, G w Representing the operational and maintenance cost factor of the wind power generator, < >>Representing the initial depreciation cost of the ith wind turbine generator system, C wins Representing the installation cost of the unit capacity of the wind generating set, k w Representing the capacity coefficient of the wind generating set, n w Indicating the service life of the wind generating set, wherein I indicates annual rate,/day rate>Representing the production cost of the ith wind power generator set,/->And the output power of the ith wind generating set is represented, and a, b and c are the production cost quadratic function coefficients.
Further, the total power generation cost of the ith photovoltaic generatorExpressed as:
wherein,representing the production cost of the ith photovoltaic generator set, < >>Indicating the output power of the ith photovoltaic generator set,/->Representing the total power generation cost of the ith photovoltaic generator, < >>Representing the operation and maintenance costs of the ith photovoltaic generator, G s Representing the operational and maintenance cost factor of the photovoltaic generator, < > >Representing the initial depreciation cost of the ith photovoltaic generator set, C sins Representing installation cost, k of unit capacity of photovoltaic generator set s Representing the capacity coefficient of the photovoltaic generator set, n s And the service life of the photovoltaic generator set is represented.
Further, the total power generation cost of the ith micro gas turbineExpressed as:
wherein,representing the production cost, x of the ith micro gas turbine i,t Indicating the operating state of the micro gas turbine, when x=0 indicates the unit is turned off, x=1 indicates the unit is operated, and the air intake valve is opened>Indicating the output of the ith micro gas turbine, < >>Representing initial depreciation cost of ith micro gas turbine, C mins Representing the installation cost, k, of the unit capacity of the micro gas turbine unit m Representing the capacity coefficient of a micro gas turbine, n m Indicating the service life of the micro gas turbine,representing the operating and maintenance costs of the ith micro gas turbine, G m Representing the operating and maintenance costs of the micro gas turbine, < >>Representing the start-stop cost of the ith micro gas turbine, f msc Indicating the cost required for start-up and shut-down of a micro gas turbine.
Further, the ith micro gas turbine releases CO 2 Cost of (2)Expressed as:
wherein,representing CO produced by an ith micro gas turbine 2 Emission amount alpha 1 ,β 1 ,γ 1 Representing CO 2 Pollution emission coefficient of>Indicating the treatment of CO released by the ith micro gas turbine 2 Cost of->Representing the cost conversion factor.
Further, the electric vehicle charging cost is expressed as:
wherein,represents the charging cost of the kth electric automobile, t a And t d Charging pile for respectively indicating arrival and departure of electric automobileTime of DeltaE of (2) k Represents the charge capacity of the kth electric automobile, < >>Represents the energy price of the jth charging pile at time t, y j,k And the kth electric automobile is indicated to select the jth charging pile for charging.
Further, the road congestion degree is expressed as:
wherein,indicating the waiting time of the kth electric car, < +.>Indicating the waiting time required by the electric car to select the jth charging stake, +.>Represents the running time of the kth electric automobile, d jk Representing the distance between the kth electric car and the jth charging post,/for>Represents the average speed of the road from the kth electric vehicle to the jth charging pile at time t,/for the kth electric vehicle>Representing the kth electric vehicleWaiting time, < >>Indicating the waiting time required by the electric car to select the jth charging stake, +.>Indicating the degree of road congestion caused by the kth electric vehicle,/- >The larger the value of (a) indicates the higher the congestion degree of the road, alpha 2 And beta 2 Representing the weight coefficient, fixed value t 0 The time required for the electric automobile to travel to the charging pile at the maximum allowable speed is represented by a fixed value t 1 Indicating the maximum tolerable waiting time for the electric vehicle.
Further, in the first step, the power load requirements of all charging piles predicted for the next 24 hours are performed by using an index model.
Further, the first-stage prize function is expressed as:
wherein,representing prediction model to predict the pre-prediction of the jth charging stakeMeasuring the amount of electricity generation>Respectively representing the actual output power of the ith wind power, the ith photovoltaic and the ith micro gas turbine to the jth charging pile at the t moment, ζ epsilon [0,1 ]]Representing the discount coefficient.
Further, the second-stage bonus function is expressed as:
wherein,e k,t 、d jk 、and +.>Respectively representing the electric quantity of the electric automobile, the distance from the electric automobile to the charging pile, the minimum electric quantity required by the electric automobile to travel to the charging pile and the maximum capacity of the battery of the electric automobile, < >>And +.>Respectively represent the energy price of the charging pile, the residual capacity of the charging pile and the queuing waiting time required by the charging pile, +.> Represents the average speed of the road, eta e [0,1 ] ]Representing the discount coefficient.
The beneficial effects of the application are as follows:
the application combines the charging cost of the user and the road congestion degree to design the optimal charging strategy so as to relieve traffic jam and improve the user satisfaction. The application also provides a proper charging strategy for the electric automobile, reduces the charging cost to the maximum extent and reduces the road congestion. The numerical result shows that the application has effectiveness and superiority in actual scenes, and achieves the aim of minimizing the power generation cost and the carbon emission. The application can realize lower charging cost of the electric automobile on the premise of minimizing the power dispatching cost and the carbon emission.
In addition, from the aspect of economic dispatch, the application designs an intelligent charging strategy of the electric automobile in the intelligent traffic system. Promote the consumption of renewable energy sources, optimize renewable energy sources and traditional energy power generation, and enable the energy distribution proportion of each charging pile to be reasonable.
Drawings
FIG. 1 is a diagram of an intelligent charging strategy architecture of an electric vehicle considering economic dispatch;
FIG. 2 is a schematic diagram of an intelligent charging double-layer optimization framework of an electric vehicle in consideration of power economy dispatching in an urban traffic system;
FIG. 3 is a schematic diagram of the overall framework of an index model;
FIG. 4 is a network structure diagram of a two-stage scheduling algorithm;
FIG. 5 is a graph of the predicted result of the total load of the charging pile for the next 24 hours;
FIG. 6 is a schematic diagram of the convergence process of EDSAC algorithm at different learning rates;
FIG. 7 is a graph showing the comparison of the usage rate of renewable energy and conventional energy in one day of the first charging pile;
FIG. 8 is a schematic diagram of the convergence process of the EVSAC algorithm at different discount rates;
FIG. 9 is a schematic diagram of test results of the EVSAC algorithm;
FIG. 10 is a graph comparing reward training results based on different DRL algorithms;
FIG. 11 is a schematic diagram showing the relationship between the energy consumption and the road congestion in a single day of the first charging pile;
FIG. 12 is a schematic diagram showing the relationship between energy consumption and energy price fluctuation in a single day of a first charging pile;
FIG. 13 is a schematic diagram showing the relationship between charging behavior and energy price fluctuation in a single day of an electric vehicle;
FIG. 14 is a diagram illustrating a relationship between a charging behavior and a road congestion level in a single day of an electric vehicle;
fig. 15 is a flowchart of the algorithm of the present application.
Detailed Description
It should be noted that, in particular, the various embodiments of the present disclosure may be combined with each other without conflict.
The first embodiment is as follows: referring to fig. 1, a specific description is given of an electric vehicle intelligent charging strategy selection method based on economic dispatch according to the present embodiment, including the following steps:
step one: acquiring historical charging data of the charging piles, and predicting the power load demands of all the charging piles in the future 24 hours according to the historical charging data of the charging piles;
step two: based on the power load demands of all charging piles and the running state constraint of a motor unit in the future 24 hours, constructing a power plant economic dispatch strategy objective function, wherein the objective function of the power plant economic dispatch strategy is expressed as:
wherein,represents the total power generation cost of the ith wind power generation,/->Representing the total power generation cost of the ith photovoltaic generator, < >>Representing the total power generation cost of the ith micro gas turbine,/->Indicating the CO released by the ith micro gas turbine 2 P represents the power generation of the unit, including +.>x represents the operating state of the micro gas turbine, T represents the time slot, K 1 ,K 2 ,K 3 ,K 4 The importance weights representing each component, which can be normalized to satisfy +.>
The operating state constraints of the motor group include:
output power constraints, power balance constraints, and micro gas turbine operating state constraints;
The output power constraint is specifically:
wherein,the upper limit of the output power of the wind power generator, the photovoltaic power generator and the micro gas turbine is respectively indicated by +.>Representing the lower limit of the output power of the micro gas turbine, < +.>Respectively representing the power output of the ith wind driven generator, the photovoltaic generator and the micro gas turbine;
the power balance constraint is specifically:
wherein L represents the number of generators, N represents the number of charging piles,representing the power load of the jth charging pile and the power transferred to the energy storage system by the jth charging pile respectively;
the running state constraint of the micro gas turbine is specifically as follows:
x i,t representing the operating state of the micro gas turbine;
step three: designing a first-stage rewarding function by using a power plant economic dispatching strategy objective function, training a first-stage DRL model, dispatching a power plant by using the trained first-stage DRL model, and obtaining a power dispatching schedule;
step four: based on the power dispatching schedule in the third step, obtaining the actual power supply quantity of each charging pile, and constructing an electric vehicle charging strategy objective function by combining the constraint of the electric vehicle in the charging process, wherein the electric vehicle charging strategy objective function is expressed as:
Wherein Δe represents the charge amount of the electric vehicle, y represents which charging pile the electric vehicle selects for charging, M represents the number of electric vehicles, K 5 ,K 6 Representing the importance weight of each component, which may be normalized to satisfyθ represents a cost conversion coefficient, +.>Indicating the degree of road congestion->Representing the charging cost of the electric automobile;
the constraint of the electric automobile in the charging process comprises the following steps:
battery charging constraints, battery power constraints, battery remaining capacity constraints, and charging pile selection constraints;
the battery charging constraint is specifically as follows:
the battery power constraint is specifically:
the battery residual capacity constraint is specifically as follows:
wherein,indicating the residual capacity of the kth electric automobile at the time t,/->Represents the minimum electric quantity required when the kth electric automobile runs to the jth charging pile, delta E k Represents the charge quantity, y of the kth electric automobile j,k Indicating that the kth electric automobile selects the jth charging pile for charging, < >>Indicating the remaining capacity of the charging pile +.>The maximum capacity at time t is shown;
the charging pile selection constraint is specifically as follows:
step four: and designing a second-stage reward function by using the electric vehicle charging strategy objective function, training a second-stage DRL model, and selecting the electric vehicle charging strategy by using the trained second-stage DRL model.
The application firstly describes the economic dispatching problem of the intelligent power grid, and then introduces the electric vehicle charging strategy dispatching problem in the intelligent traffic system. The system architecture is shown in fig. 1, and consists of a smart grid and an intelligent transportation system. Firstly, historical load data of the charging piles are transmitted to a power grid server, analysis is carried out, the electric quantity required by the charging piles in the next day is predicted, and the electric quantity scheduled to each charging pile is determined. Secondly, the electric automobile selects a proper charging pile to carry out a charging task according to the self demand. Finally, the excess energy generated is stored in an Energy Storage System (ESS).
Figure 2 depicts a two-layer optimization framework in accordance with the present application. The lower layer is an economic dispatching part in the intelligent power grid, and the considered units are a wind generating set, a photovoltaic generating set and a micro gas turbine generating set. By distributing a reasonable proportion of the electric power consisting of renewable energy and traditional energy to each charging pile, the digestion of renewable energy is promoted to achieve minimization of power generation cost and carbon emission. After the power grid is scheduled, the upper-layer electric automobile selects a proper charging pile to carry out a charging task, so that the charging cost and the road congestion degree of the upper-layer electric automobile are reduced. In order to obtain better performance, the bottom-layer power grid indirectly controls the charging tasks of the electric vehicles by dispatching reasonable energy to each charging pile, and the upper-layer electric vehicles feed energy consumption data back to the power grid, so that the power grid can better dispatch sufficient renewable energy sources, and environmental pollution is reduced.
The energy source of the present application is generated by wind energy, photovoltaic and micro gas turbines. The goal of economic dispatch is to minimize the total cost of power generation, including the cost of wind power generation, photovoltaic power generation, micro gas turbine power generation, and the disposal of CO emitted by micro gas turbines 2 Is not limited by the cost of (a).
In order to fully examine the cost of wind power generation, the present application considers three key factors, including production costs, initial depreciation costs, operation and maintenance costs.
Production cost: the production cost of wind power generation is generally expressed as a quadratic function of the output power, expressed as:
wherein,production cost for ith wind power generation set, < >>And a, b and c are the output power of the ith wind generating set and are the production cost quadratic function coefficients.
Initial depreciation cost-after a certain period of use, the wind driven generator itself generates an initial depreciation cost, expressed as:
wherein,for the initial depreciation cost of the ith wind turbine generator system, C wins The installation cost k is the unit capacity of the wind generating set w Is the capacity coefficient of the wind generating set, n w The service life of the wind generating set is prolonged, and I is annual fee rate.
Running and maintenance costs: the wind power generator requires regular maintenance during operation, where operation and maintenance costs will be incurred, expressed as:
Wherein,representing the operation and maintenance costs of the ith wind power generator, G w Representing the operational and maintenance cost coefficients of the wind turbine.
Thus, the total cost of wind power generation can be described as:
wherein the method comprises the steps ofIs the ith wind power generationTotal power generation cost of the motor.
Photovoltaic power generation
Similar to wind power generation, the total cost of photovoltaic power generation consists of production costs, initial depreciation costs, operation and maintenance costs.
Production cost: production cost of photovoltaic power generation is generally expressed as a quadratic function of output power, expressed as:
wherein,production cost for ith photovoltaic generator set, < >>The output power of the ith photovoltaic generator set.
Initial depreciation cost-after a period of use, the photovoltaic generator will produce an initial depreciation cost expressed as:
wherein,c, the initial depreciation cost of the ith photovoltaic generator set sins Installation cost k for unit capacity of photovoltaic generator set s Is the capacity coefficient of the photovoltaic generator set, n s The service life of the photovoltaic generator set is prolonged.
Running and maintenance costs: the photovoltaic generator requires regular maintenance during operation, which will result in operating and maintenance costs, expressed as:
wherein,representing the operation and maintenance costs of the ith photovoltaic generator, G s Representing the operational and maintenance cost coefficients of the photovoltaic generator.
Thus, the total cost of photovoltaic power generation can be described as:
wherein the method comprises the steps ofIs the total power generation cost of the ith photovoltaic generator.
Miniature gas turbine power generation
Due to the instability of renewable energy power generation, micro gas turbine power generation is indispensable, which brings about production cost, start-stop cost, operation maintenance cost and initial depreciation cost of micro gas turbine power generation.
The production cost is as follows: the production cost of micro gas turbine power generation is typically expressed as a quadratic function of the output power, expressed as:
wherein,representing the production cost, x of the ith micro gas turbine i,t Indicating the operating state of the micro gas turbine, when x=0 indicates the unit is turned off, x=1 indicates the unit is operated, and the air intake valve is opened>The output of the ith micro gas turbine is shown.
Start-stop cost: when the micro gas turbine generator is switched from on to off or off to on, a certain cost will be incurred, expressed as:
wherein,representing the start-stop cost of the ith micro gas turbine, f msc Indicating the cost required for start-up and shut-down of a micro gas turbine.
Initial depreciation cost: the initial depreciation cost of micro gas turbine power generation is defined as:
Wherein,representing initial depreciation cost of ith micro gas turbine, C mins Representing the installation cost, k, of the unit capacity of the micro gas turbine unit m Representing the capacity coefficient of a micro gas turbine, n m Indicating the life of the micro gas turbine.
Running and maintenance costs: the operating and maintenance costs of a micro gas turbine are defined as:
wherein,representing the operating and maintenance costs of the ith micro gas turbine, G m Is a coefficient of operational and maintenance costs for micro gas turbines.
Thus, the total cost of micro gas turbine power generation is:
wherein the method comprises the steps ofIs the total power generation cost of the ith micro gas turbine.
In addition, micro gas turbine power generation can cause certain pollution to the environment, such as carbon dioxide emissions, expressed as:
wherein,representing CO produced by an ith micro gas turbine 2 Emission amount alpha 1 ,β 1 ,γ 1 Respectively represent CO 2 Is a pollution emission coefficient of (a).
Thus, the treatment of CO released by micro gas turbines 2 The cost of (2) is given by:
wherein,indicating the treatment of CO released by the ith micro gas turbine 2 Is not limited by the cost of (a).Representing a cost conversion factor, the purpose of which is to convert carbon dioxide emissions into cost.
Power economy scheduling constraints
The smart grid economic dispatch needs to meet various constraints including output power constraints, power balance constraints, and micro gas turbine operating state constraints, as follows.
Output power constraint: in the power generation process, the power generation power of the unit cannot exceed the upper limit and the lower limit, and the power generation power is expressed as follows:
wherein the method comprises the steps ofThe upper limit of the output power of the wind power generator, the photovoltaic power generator and the micro gas turbine is respectively indicated by +.>Representing the lower limit of the output power of the micro gas turbine.
Power balance constraint: the power balance constraint is the most fundamental requirement for the generator power generation process, expressed as:
where L is the number of generators and N represents the number of charging piles.Representing the power output of the ith wind power generator, photovoltaic power generator and micro gas turbine, respectively,/->Representing the power load of the jth charging pile and the power transferred to the energy storage system by the jth charging pile, respectively。
Operating state constraints for micro gas turbines: in the process of generating power by a micro gas turbine, a generator can only be positioned at a certain moment
In the start or stop state, expressed as:
based on the above description, the goal of the power economy dispatch is to minimize the overall cost of generating the genset while taking into account the genset output and operating state constraints. Thus, the objective function of the first stage of determining the optimal economic dispatch strategy may be defined as follows:
s.t.～
wherein P represents the power generation of the unit and comprises x represents the operating state of the micro gas turbine, and x is i,t The same applies. T represents a time slot, K 1 ,K 2 ,K 3 ,K 4 Is the importance weight of each component, which can be normalized to satisfy
Electric automobile charging strategy problem in intelligent traffic system
When the intelligent power grid economic dispatching is completed, each charging pile correspondingly receives electric energy with different proportions consisting of renewable energy sources and traditional energy sources, and in an intelligent traffic system, the aim of the application is to reduce the charging cost and the road congestion degree of a user to the greatest extent while meeting the charging requirement of an electric automobile.
Cost of charging
The charging cost of an electric car can be expressed as:
wherein,represents the charging cost of the kth electric automobile, t a And t d Respectively represent the time of the electric automobile reaching and leaving the charging pile, delta E k Represents the charge capacity of the kth electric automobile, < >>And the energy price of the jth charging pile at time t is represented, wherein the energy price is determined by a third-party operator of the charging pile, and the energy price of each charging pile in the hour is slightly different for promoting the electric automobile to select the charging pile with the lower energy price for charging. y is j,k And the kth electric automobile is indicated to select the jth charging pile for charging.
Travel time
The application defines the running time of the electric vehicle by the distance and the speed of the electric vehicle from the charging pile, and is expressed as follows:
wherein,is the running time of the kth electric automobile, d jk Representing the distance between the kth electric car and the jth charging post,/for>The average speed of the road from the kth electric vehicle to the jth charging pile at time t is represented.
Charge latency
The present application assumes that each charging post is installed on both sides of the road, and when an electric car arrives at the charging post, if there is another electric car in front of it, it must wait for charging. At this time, the waiting queue of the electric automobile affects the throughput of the road, and the waiting time is defined as:
wherein,indicating the waiting time of the kth electric car, < +.>Indicating the waiting time required by the electric vehicle to select the j-th charging pile.
Degree of road congestion
When an electric car is driving to a charging pile or waiting for charging, roads may be congested because many electric cars need to be charged. The present application indirectly reflects the road congestion degree through the running time and the charging waiting time of the kth electric automobile, and usesAnd (3) representing. The lower the values of the travel time and the charge waiting time, the larger the road traffic flow, the smaller the road congestion degree.
Wherein the method comprises the steps ofIndicating the degree of road congestion caused by the kth electric vehicle. Alpha 2 And beta 2 The weight coefficients are the sum of the two calculated values for the normalization equation. Fixed value t 0 The time required for the electric vehicle to travel to the charging pile at the maximum allowable speed is represented. Fixed value t 1 Indicating the maximum tolerable waiting time for the electric vehicle.The larger the value of (c) indicates the higher the congestion degree of the road.
Constraint condition for charging schedule of electric automobile
The electric vehicle charging needs to meet various constraints including battery charging constraints, battery power constraints, battery remaining capacity constraints, and charging stake selection constraints. The specific case is as follows.
Constraint of battery charging: the charging capacity of the kth electric automobile cannot exceed the remaining capacity of the charging pile, and is expressed as:
battery power constraint: the maximum capacity of the kth electric automobile at the time t, which cannot be exceeded by the battery electric quantity at the time t, is represented by
The method comprises the following steps:
battery remaining capacity constraint: the remaining battery power of the kth electric automobile should be able to meet that it reaches the charging stake, expressed as:
wherein,indicating the residual capacity of the kth electric automobile at the time t,/->Representing the minimum amount of power required when the kth electric vehicle travels to the jth charging stake.
Charging pile selection constraint: the electric automobile can only select one charging pile to carry out charging tasks at each moment, and the charging tasks are expressed as follows:
based on the above description, the goal of the electric vehicle charging schedule is to minimize the charging cost and the road congestion of the user while taking into account multiple constraints of the electric vehicle during charging. Thus, the objective function of the second stage to determine the optimal electric vehicle charging strategy may be defined as:
s.t.～
wherein delta E represents the charging capacity of the electric automobile, y represents the charging pile selected by the electric automobile for charging, and the charging pile and y are the same as each other j,k The meaning is the same. M is the number of electric vehicles, K 5 ,K 6 Is the importance weight of each component, which can be normalized to satisfyθ represents a cost conversion coefficient, the purpose of which is to convert the degree of congestion of the road into cost.
Because of the integer constraint sums, both the power economy dispatch and electric vehicle charging problems herein are NP-hard.
The application firstly predicts the load demand of the charging pile for 24 hours in the future by using an index deep learning model. Secondly, in order to effectively solve the NP difficult problem, the application provides two DRL algorithms on the basis of the SAC algorithm to solve the two optimization problems.
Power load prediction based on Informir model
With the popularization of electric vehicles, a charging pile is an important component of an electric vehicle charging facility, and the electric load demand prediction is important to the operation and planning of an electric power system. In order to better realize accurate prediction of the electricity load of the charging pile, an index-based deep learning method is adopted.
As a deep learning model, an index [26] There are many advantages in predicting the power load demand of the charging pile the next day. First, the charging pile load data generally contains a large amount of time-series information, such as an hour level or a minute level. As a sequence-to-sequence model, an index model can capture the temporal relationship in time series data and can take into account the influence of historical data when making predictions. Secondly, the index model adopts a self-attention mechanism, and can automatically learn the long-term dependence in the data, so that the index model can better process the sequence dependence in the charging pile load data. In addition, the inpform model also adopts a multi-layer encoder and decoder structure, so that multi-layer abstraction and reconstruction can be carried out on an input sequence, and characteristic information in data can be better captured. Finally, the model is flexible in model parameter setting and training method, and can be customized according to the characteristics and the requirements of the charging pile data, so that a more accurate prediction result is obtained. In summary, based on the characteristics of the charging pile load data, the power load demand of the charging pile on the next day is predicted by adopting an index model.
input of an index model
Fig. 3 shows the overall framework of the former model. The input of the model consists of two parts, X feed_en And X feed_de 。X feed_en Represented is the input of the encoder, typically a historical observation sequence. X is X feed_de Represented is the input of a decoder, represented by X token And X 0 Composition is prepared. Wherein X is token Representing sequences of known length before the predicted target sequence begins. X is X 0 Representing placeholders of the target sequence.
Model self-attention mechanism
First, the input form of the traditional self-attention mechanism is (Query, key), followed by a scaled dot product, namely:
wherein Q is a Query vector, K is a Key vector, V is a Value vector, d k Is the dimension of the projection matrix and,representing the transpose of the matrix,For scaling the attention weights to better control the gradient. The probability form of the attention coefficient of the first Query is:
wherein,and k (q) l ,k g ) Selecting an asymmetric index kernel->
The self-attention mechanism uses a dot product operation of quadratic time complexity to calculate the probability p (q l ,k g ) And the calculation requiresSpatial complexity of the size, where L Q ,L K The input lengths of Query and Key are represented, respectively, and therefore this is a major obstacle to improving predictive ability. In addition, in probability research, probability distribution of self-attention mechanism has a certain potential sparsity.
In order to measure the sparsity of the Query, the application uses the Kullback-Leibler divergence [27] . The sparsity evaluation formula for the first Query is:
the first of these is the Log-Sum-Exp of all keys, and the second is their arithmetic mean. Based on the above evaluation method, the self-attention formula of probspark can be obtained, namely:
wherein,is a sparse matrix of the same size as Q and it contains only the first u Query's under sparse assessment M (Q, M), u being of a size determined by the sampling parameters. This allows the ProbSparse to self-note that only every Query-Key needs to be madeDot product calculation of the size, thereby replacing classical self-attention mechanisms.
Encoder of an index model
The encoder is a structure provided for extracting long-term correlations of long-sequence inputs. Because of the probabilistic sparsity of the self-attention mechanism, there is a redundant combination of Value in the feature map of the encoder. Thus, the distillation operation assigns higher weights to the dominant features and generates a focus self-attention feature map in the next layer and a focus self-attention feature map in the next layer. The distillation procedure from g layer to g+1 layer was as follows:
wherein the method comprises the steps ofAnd->Respectively represent the sequences of g+1 layer and g layer, and +. >By->And (5) generating. MaxPool (·) represents maximum pooling, conv1d represents one-dimensional convolution, and ELU (·) is the activation function. The timing problem proposed by the inputer adopts a self-attention mechanism to reduce the length of each layer of input sequence of the decoder by half, thereby greatly saving the memory use and calculation time of the encoder.
Decoder of an index model
Fig. 3 uses a standard decoder architecture, which is formed by stacking two identical multi-headed attention layers. However, generating inferences serves to mitigate the slowdown of long-term predictions. The present application provides the following vectors to the decoder:
wherein Concat (-) is a join function that willAnd->Are linked into a long sequence. ProbSparse self-attention uses masked multi-headed attention in the calculation. The masked multi-headed attention mechanism does not make each location aware of the next location, thereby avoiding autoregressive problems. The final output is obtained through a fully connected layer and whether the prediction is univariate or multivariate determines its output dimensions. The method uses a generating structure in the decoder structure of the encoder, and can generate all the prediction sequences at one time, thereby greatly shortening the time of prediction decoding.
Algorithm based on deep reinforcement learning
DRL has been widely used in the fields of automobile traffic, economic dispatch, etc. In this context, in order to solve these two optimization problems, the present application devised a DRL-based algorithm that can adaptively learn the optimal strategy and does not require any a priori knowledge of the uncertainty. Since the DRL is based on a Markov Decision Process (MDP), the present application first converts the sum problem into MDP form.
Smart grid economic dispatch MDP
System state: the present application takes 24 hours a day as a period, and for time t e (0, 1,2,., 23), the system state is defined as:
wherein the method comprises the steps ofAnd the predicted power generation amount of the j-th charging pile is predicted by the prediction model.
The actions are as follows: the intelligent agent makes corresponding actions according to the current system state:
wherein,the power actually output by the ith wind power, the ith photovoltaic and the ith micro gas turbine to the jth charging pile at the t moment is respectively shown.
Bonus function: the system gives the agent an immediate prize immediately after the agent makes an action, which is also the objective function of the optimization problem, namely:
after the T time has elapsed, the system will acquire the total jackpot:
wherein ζ ε [0,1] represents the discount coefficient.
Electric automobile charging schedule MDP in intelligent traffic system
System state: taking into consideration the fluctuation of the energy price and the average speed of the road within 24 hours, letThe application divides the system state into three parts of electric automobile, charging pile and road traffic, and the concrete steps are as follows:
wherein,the electric quantity of the electric automobile, the distance from the electric automobile to the charging pile, the minimum electric quantity required by the electric automobile to travel to the charging pile and the maximum capacity of the battery of the electric automobile are respectively represented.
And respectively representing the energy price of the charging pile, the residual capacity of the charging pile and the queuing waiting time required by the charging pile.
Representing the average speed of the road.
The actions are as follows: according to the current system state, the electric automobile performs an action defined as:
bonus function: when the electric vehicle completes an action, the system feeds back an instant reward to the electric vehicle, wherein the reward is also an objective function of the optimization problem:
after T, the system will acquire the total jackpot:
wherein η ε [0,1] represents the discount coefficient.
Two-stage DRL algorithm
In the power economy dispatching and electric vehicle charging dispatching process, since the state transition of the system is high-dimensional and continuously variable, a DRL algorithm based on SAC is designed for obtaining an optimal dispatching strategy.
In addition, SAC generates actions based on maximum entropy theory under off-polarity, so that the defect of high complexity of off-polarity reinforcement learning sampling is overcome, and the problem of weak convergence of reinforcement learning is solved.
Based on SAC frame, the application respectively uses deep neural network to define soft state value function network V ψ (s t ) Soft Q-function network Q θ (s t ,a t ) And policy function network pi φ (a t |s t ). Furthermore, a target network with soft state value functions implicitly defined in SAC
To train the soft state value function network parameters, MSE is used to minimize the residual error expressed as:
wherein the state is from an experience bufferThe action is generated according to the current strategy.
Similarly, the objective function Q of the soft Q function θ (s t ,a t ) MSE is also used to minimize the soft Bellman residual, which is expressed as:
wherein,an estimate of the soft Q objective function is shown.
Policy pi of the application φ (a t |s t ) Is to minimize the Kullback-Leible divergence between the two distributions, expressed as:
in the first stage, the present application proposes an economical scheduling (EDSAC) algorithm based on SAC. As shown in part a of fig. 4, the agent collects status information from the charging piles, then gives an action set using the deep neural network, performs the action, then receives rewards or penalties given by the system, and finally tries to learn an optimal scheduling strategy, thereby distributing different proportions of electric energy to each charging pile. Part b of fig. 4 shows that the entire neural network is made up of the main criticism network, the target criticism network, the actor network, and the experience replay buffer. The specific procedure is shown in fig. 15.
The time complexity of FIG. 15 is relatively low, consisting essentially of 2 with Z 1 Soft Q network with full connectivity layer, 2 with Z 2 Value network with full connection layer and 1 with Z 3 All-connected layerPolicy network decisions of (2). The specific time complexity calculation form is as follows:
wherein,respectively z 1 A soft Q network layer, z 2 Personal value network layer and z-th 3 Number of elements of the policy network layer.
After economic dispatch of the smart grid, each charging pile receives renewable energy sources and traditional energy sources in different proportions. In the second stage, the application provides an electric vehicle charging scheduling algorithm (EVSAC) based on SAC. As shown in fig. 4 (c), the electric vehicle in the area is an agent that acquires state information from the charging pile and tries to learn an optimal charging strategy. The specific algorithm steps are similar to algorithm 1, the inputs are the output of algorithm 1, the road condition and the electric quantity of the electric automobile, and the output of the EVSAC algorithm is to select a certain charging pile and how much electric quantity is charged. The state, action, and manifestation of the rewarding function of the EVSAC algorithm have been presented in MDP form in section 4.2.2.
In a simulation experiment, for the economic dispatch optimization problem of the first stage, the application is provided with 5 wind power generators, 5 photovoltaic generators and 5 micro gas turbines. The real data set contains photovoltaic power generation, wind power generation and load data from the elia group. For the electric automobile charging scheduling problem of the second stage, a total of 4 charging piles belong to the same operator, and 2 cross roads are arranged. The energy price and the average speed of the road respectively obey the truncated normal distribution with different mean values and standard deviations, and the rest variables are all obeyed the uniform distribution. The specific parameter settings are shown in table 1.
Table 1 experimental parameter settings
For the EDSAC algorithm, the present application trains the model using three types of networks, namely a soft state value network, a soft Q network, and a policy network, respectively. The soft state value network has the structure (3, 256,1, relu (x), adam), the soft Q network has the structure (11, 256,1, relu (x), adam), the policy network has the structure (3,256,256,8,adam). For the EVSAC algorithm, the soft state value network has the structure of (17, 128,1, relu (x), adam), the soft Q network has the structure of (22, 128,1, relu (x), adam), and the policy network has the structure of (17,128,128,5,>adam). The size of the experience replay buffer is set to 10 6 The size of the batch of samples is set to 64./>
All simulation experiments in the application are run on a 4-core i5 central processing unit, a 12GB memory and a GTX960M display card, and the versions of Python and Pytorch are respectively 3.8 and 1.8.
In order to reduce the problems of energy surplus and environmental pollution, the application needs to predict the power load demands of all charging piles within 24 hours in the future, so as to provide enough time for scheduling decisions. Table 2 shows a comparison of evaluation indexes of three prediction models for predicting load data of a charging pile for 24 hours in the future. "MSE" means mean square error for evaluating the degree of variation of data. "RMSE" represents the root mean square error, which is a measure of the degree of dispersion of the deviation between the predicted data and the actual data. "MAE" is the mean absolute error that reflects the predictive performance of the mean magnitude of the model predictive error. "MAPE" refers to the mean absolute percent error that reflects the overall average performance of the predictive model. The smaller the values of the above four indices, the smaller the prediction error of the model.
Fig. 5 is a result of total load prediction of the charging pile for 24 hours in the future. As can be seen from fig. 5, the predictive curve of the index model fits a more "Real" curve than the LSTM, RNN model.
Power economy dispatch results
FIG. 6 shows normalized average jackpots at different learning rates (lr). When the number of training rounds reaches 4000, both the circular curve and the triangular curve converge to the same prize value. However, the convergence speed of a circular curve is much faster than that of a triangular curve. When lr=10 -4 While the rewards in the first 1500 rounds of training exhibited an upward trend, the average jackpot exhibited a significant downward trend as the number of training rounds increased. Overall, when lr=10 -5 The EDSAC algorithm performs optimally when it is used. Thus, different network learning rates have a great impact on the convergence trend of rewards.
Fig. 7 shows the result of optimizing the renewable energy source of the first charging pile and the conventional energy source, wherein delta is the renewable energy utilization rate. As can be seen from fig. 7, the square curve is above the triangular curve for most of the 24 hours of the day. For example, 9 to 10 a.m., renewable energy utilization even exceeds 75% due to the increased charging demand of electric vehicles. In general, the power generation utilization rate of renewable energy sources in a single day of the first charging pile is higher than that of traditional energy sources.
Electric automobile charging scheduling result
Fig. 8 depicts the normalized average jackpot for the proposed EVSAC algorithm at different discount rates. The average jackpot at the three discount rates showed a tendency to increase gradually as the number of training rounds was followed by 1500 rounds. However, when the number of training rounds is between 1500 and 4000, the reward curves for γ=0.79 and γ=0.89 decrease significantly. When the number of training rounds reaches 8000 rounds, the average jackpot at all three discount rates reaches a convergence state, but the average jackpot at γ=0.99 is greater than the average jackpot at the other two discount rates. From the overall effect, when γ=0.99, the performance of the EVSAC algorithm proposed by the present application is optimal. Thus, different discount rates have a great impact on the convergence of rewards.
Fig. 9 shows the test results of the EVSAC algorithm. The histogram represents the generalization ability of the algorithm over 10 sets of test data sets. The dashed line is the optimal average jackpot value for the model during training, which is 0.95. As can be seen from fig. 9, the average jackpot result for the 10 sets of test data sets was substantially around 0.95, so that the model of the present application achieved good generalization performance.
Fig. 10 shows a comparison of normalized average jackpot when two different algorithms are used to solve the scheduling problem in the same electric vehicle charging scheduling environment. As can be seen from fig. 10, the EVSAC algorithm proposed by the present application is superior to the PPO algorithm in terms of convergence speed and final prize value.
Fig. 11 and 12 illustrate the influence factors of the energy consumption in one day of the first charging pile, wherein the histogram represents the energy consumption per time of the first charging pile, the square curve in fig. 11 represents the degree of congestion of the road, and the triangle curve in fig. 12 represents the energy price. As can be seen from fig. 11 and 12, during the period of 6:00 to 8:00, road congestion is gradually increased, the charging demand of the electric vehicle is reduced, and the energy price is at a minimum. And 13:00, the photovoltaic power generation amount is maximum, the charging requirement of the electric automobile is gradually increased, and the energy price creates the record of the highest day of the single day. During the period of 16:00-18:00, the road congestion degree suddenly increases to more than 0.7 due to the rush hour.
In fig. 13 and 14, the histogram represents the electric quantity of the electric vehicle, the three hatched histograms represent the charged electric quantity of the electric vehicle at the charging stake, the triangle curve in fig. 13 represents the energy price, the square curve in fig. 14 represents the road congestion degree at different times of the day, and the broken line represents the minimum electric quantity required for the electric vehicle to reach the charging stake. As can be seen from fig. 13 and 14, from 0 to 13, the electric quantity of the electric vehicle gradually decreases, approaching the threshold value. Therefore, the electric vehicle must be charged by selecting the charging stake. Through the EVSAC algorithm provided by the application, the electric automobile is finally selected to be charged at 14:00. Although the current energy price is not the lowest price of the current day, the congestion degree of the electric automobile on the road is about 0.3 at this time, which is not congested. On the premise that the electric automobile can travel to the charging pile at the fastest speed, the electric automobile can be charged at a lower energy price. Therefore, the EVSAC algorithm can provide a proper charging strategy for electric vehicles and intelligent transportation systems.
In the application, from the aspect of economic dispatch, an intelligent charging strategy of the electric automobile in the intelligent traffic system is designed. Firstly, in order to promote the consumption of renewable energy sources, an EDSAC algorithm is designed, and renewable energy sources and traditional energy source power generation are optimized, so that the energy source distribution proportion of each charging pile is reasonable. Secondly, according to the charging strategy provided by the EVSAC, the electric automobile can select the most suitable charging pile to carry out a charging task according to the charging requirement of the electric automobile. Numerical results show that the algorithm provided by the application has effectiveness and superiority in actual scenes, and the aim of minimizing the power generation cost and the carbon emission is fulfilled. In addition, the algorithm provides a proper charging strategy for the electric automobile, and reduces the charging cost to the maximum extent and the road congestion.
It should be noted that the detailed description is merely for explaining and describing the technical solution of the present application, and the scope of protection of the claims should not be limited thereto. All changes which come within the meaning and range of equivalency of the claims and the specification are to be embraced within their scope.
Claims (10)
1. An electric automobile intelligent charging strategy selection method based on economic dispatch is characterized by comprising the following steps:
Step one: acquiring historical charging data of the charging piles, and predicting the power load demands of all the charging piles in the future 24 hours according to the historical charging data of the charging piles;
step two: based on the power load demands of all charging piles and the running state constraint of a motor unit in the future 24 hours, constructing a power plant economic dispatch strategy objective function, wherein the objective function of the power plant economic dispatch strategy is expressed as:
wherein,represents the total power generation cost of the ith wind power generation,/->Representing the total power generation cost of the ith photovoltaic generator, < >>Representing the total power generation cost of the ith micro gas turbine,/->Indicating the CO released by the ith micro gas turbine 2 P represents the power generation of the unit, including +.>x represents the operating state of the micro gas turbine, T represents the time slot, K 1 ,K 2 ,K 3 ,K 4 The importance weights representing each component, which can be normalized to satisfy +.>
The operating state constraints of the motor group include:
output power constraints, power balance constraints, and micro gas turbine operating state constraints;
the output power constraint is specifically:
wherein,the upper limit of the output power of the wind power generator, the photovoltaic power generator and the micro gas turbine is respectively indicated by +.>Representing the lower limit of the output power of the micro gas turbine, < +. >Respectively representing the power output of the ith wind driven generator, the photovoltaic generator and the micro gas turbine;
the power balance constraint is specifically:
wherein L represents the number of generators, N represents the number of charging piles,representing the power load of the jth charging pile and the power transferred to the energy storage system by the jth charging pile respectively;
the running state constraint of the micro gas turbine is specifically as follows:
x i,t representing the operating state of the micro gas turbine;
step three: designing a first-stage rewarding function by using a power plant economic dispatching strategy objective function, training a first-stage DRL model, dispatching a power plant by using the trained first-stage DRL model, and obtaining a power dispatching schedule;
step four: based on the power dispatching schedule in the third step, obtaining the actual power supply quantity of each charging pile, and constructing an electric vehicle charging strategy objective function by combining the constraint of the electric vehicle in the charging process, wherein the electric vehicle charging strategy objective function is expressed as:
wherein Δe represents the charge amount of the electric vehicle, y represents which charging pile the electric vehicle selects for charging, M represents the number of electric vehicles, K 5 ,K 6 Representing the importance weight of each component, which may be normalized to satisfy θ represents a cost conversion coefficient, +.>Indicating the degree of road congestion->Representing the charging cost of the electric automobile;
the constraint of the electric automobile in the charging process comprises the following steps:
battery charging constraints, battery power constraints, battery remaining capacity constraints, and charging pile selection constraints;
the battery charging constraint is specifically as follows:
the battery power constraint is specifically:
the battery residual capacity constraint is specifically as follows:
wherein,indicating the residual capacity of the kth electric automobile at the time t,/->Represents the minimum electric quantity required when the kth electric automobile runs to the jth charging pile, delta E k Represents the charge quantity, y of the kth electric automobile j,k Indicating that the kth electric automobile selects the jth charging pile for charging, < >>Indicating the remaining capacity of the charging pile +.>The maximum capacity at time t is shown;
the charging pile selection constraint is specifically as follows:
step five: designing a second-stage rewarding function by utilizing an electric vehicle charging strategy objective function, training a second-stage DRL model, and selecting an electric vehicle charging strategy by utilizing the trained second-stage DRL model;
step six: and finishing intelligent charging strategy selection of the electric automobile by using the trained first-stage DRL model and the trained second-stage DRL model.
2. The method for selecting the intelligent charging strategy of the electric automobile based on economic dispatch of claim 1, wherein the total power generation cost of the ith wind power generator is as followsExpressed as:
wherein,representing the operation and maintenance costs of the ith wind power generator, G w Representing the operational and maintenance cost factor of the wind power generator, < >>Representing the initial depreciation cost of the ith wind turbine generator system, C wins Representing the installation cost of the unit capacity of the wind generating set, k w Representing the capacity coefficient of the wind generating set, n w Representing a wind power generatorThe service life of the group, I represents annual rate,/->Representing the production cost of the ith wind power generator set,/->And the output power of the ith wind generating set is represented, and a, b and c are the production cost quadratic function coefficients.
3. The method for selecting the intelligent charging strategy of the electric automobile based on economic dispatch according to claim 2, wherein the total power generation cost of the ith photovoltaic generator is as followsExpressed as:
wherein,representing the production cost of the ith photovoltaic generator set, < >>Indicating the output power of the ith photovoltaic generator set,/->Representing the total power generation cost of the ith photovoltaic generator, < >>Representing the operation and maintenance costs of the ith photovoltaic generator, G s Representing the operational and maintenance cost factor of the photovoltaic generator, < >>Representing the initial depreciation cost of the ith photovoltaic generator set, C sins Representing installation cost, k of unit capacity of photovoltaic generator set s Representing the capacity coefficient of the photovoltaic generator set, n s And the service life of the photovoltaic generator set is represented.
4. The method for selecting an intelligent charging strategy for an electric vehicle based on economic dispatch of claim 3, wherein the i-th micro gas turbine has a total power generation cost ofExpressed as:
wherein,representing the production cost, x of the ith micro gas turbine i,t Indicating the operating state of the micro gas turbine, when x=0 indicates the unit is turned off, x=1 indicates the unit is operated, and the air intake valve is opened>Indicating the output power of the ith micro gas turbine,representing initial depreciation cost of ith micro gas turbine, C mins Representing the installation cost, k, of the unit capacity of the micro gas turbine unit m Representing the capacity coefficient of a micro gas turbine, n m Indicating the service life of the micro gas turbine,representing the operating and maintenance costs of the ith micro gas turbine, G m Representing the operating and maintenance costs of the micro gas turbine, < >>Representing the start-stop cost of the ith micro gas turbine, f msc Indicating the cost required for start-up and shut-down of a micro gas turbine.
5. The method for selecting an intelligent charging strategy for an electric vehicle based on economic dispatch of claim 4, wherein the ith step is characterized byCO released by micro gas turbines 2 Cost of (2)Expressed as:
wherein,representing CO produced by an ith micro gas turbine 2 Emission amount alpha 1 ,β 1 ,γ 1 Representing CO 2 Pollution emission coefficient of>Indicating the treatment of CO released by the ith micro gas turbine 2 Cost of->Representing the cost conversion factor.
6. The method for selecting the intelligent charging strategy of the electric automobile based on economic dispatch of claim 5, wherein the charging cost of the electric automobile is expressed as:
wherein,indicating the charge cost of the kth electric vehicle,t a and t d Respectively represent the time of the electric automobile reaching and leaving the charging pile, delta E k Represents the charge capacity of the kth electric automobile, < >>Represents the energy price of the jth charging pile at time t, y j,k And the kth electric automobile is indicated to select the jth charging pile for charging.
7. The method for selecting an intelligent charging strategy for an electric vehicle based on economic dispatch as claimed in claim 6, wherein the road congestion level is expressed as:
wherein,indicating the waiting time of the kth electric car, < +. >Indicating the waiting time required by the electric car to select the jth charging stake, +.>Represents the running time of the kth electric automobile, d jk Representing the distance between the kth electric car and the jth charging post,/for>Represents the average speed of the road from the kth electric vehicle to the jth charging pile at time t,/for the kth electric vehicle>Indicating the waiting time of the kth electric car, < +.>Representing the waiting time required by the electric automobile to select the j-th charging pile, D k cwd Indicating the degree of road congestion caused by the kth electric vehicle,/->The larger the value of (a) indicates the higher the congestion degree of the road, alpha 2 And beta 2 Representing the weight coefficient, fixed value t 0 The time required for the electric automobile to travel to the charging pile at the maximum allowable speed is represented by a fixed value t 1 Indicating the maximum tolerable waiting time for the electric vehicle.
8. The method for selecting the intelligent charging strategy of the electric automobile based on economic dispatch of claim 7, wherein the step one predicts the power load requirements of all charging piles for the future 24 hours by using an index model.
9. The method for selecting an intelligent charging strategy for an electric vehicle based on economic dispatch of claim 8, wherein the first-stage rewarding function is expressed as:
Wherein,representing the predicted power generation amount of the j-th charging pile predicted by the prediction model,/and>respectively representing the actual output power of the ith wind power, the ith photovoltaic and the ith micro gas turbine to the jth charging pile at the t moment, ζ epsilon [0,1 ]]Representing the discount coefficient.
10. The method for selecting an intelligent charging strategy for an electric vehicle based on economic dispatch of claim 9, wherein the second-stage rewarding function is expressed as:
wherein,e k,t 、d jk 、and +.>Respectively representing the electric quantity of the electric automobile, the distance from the electric automobile to the charging pile, the minimum electric quantity required by the electric automobile to travel to the charging pile and the maximum capacity of the battery of the electric automobile, < >> And +.>Respectively represent the energy price of the charging pile, the residual capacity of the charging pile and the queuing waiting time required by the charging pile, +.> Represents the average speed of the road, eta e [0,1 ]]Representing the discount coefficient.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311096583.8A CN117207803A (en) | 2023-08-29 | 2023-08-29 | Electric automobile intelligent charging strategy selection method based on economic dispatch |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311096583.8A CN117207803A (en) | 2023-08-29 | 2023-08-29 | Electric automobile intelligent charging strategy selection method based on economic dispatch |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117207803A true CN117207803A (en) | 2023-12-12 |
Family
ID=89034451
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311096583.8A Pending CN117207803A (en) | 2023-08-29 | 2023-08-29 | Electric automobile intelligent charging strategy selection method based on economic dispatch |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117207803A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117689188A (en) * | 2024-02-04 | 2024-03-12 | 江西驴充充物联网科技有限公司 | Big data-based user charging strategy optimization system and method |
| CN118144619A (en) * | 2024-05-09 | 2024-06-07 | 山东浪潮智慧能源科技有限公司 | Low-carbon emission charging method and device for electric automobile and storage medium |
-
2023
- 2023-08-29 CN CN202311096583.8A patent/CN117207803A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117689188A (en) * | 2024-02-04 | 2024-03-12 | 江西驴充充物联网科技有限公司 | Big data-based user charging strategy optimization system and method |
| CN117689188B (en) * | 2024-02-04 | 2024-04-26 | 江西驴充充物联网科技有限公司 | Big data-based user charging strategy optimization system and method |
| CN118144619A (en) * | 2024-05-09 | 2024-06-07 | 山东浪潮智慧能源科技有限公司 | Low-carbon emission charging method and device for electric automobile and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117207803A (en) | Electric automobile intelligent charging strategy selection method based on economic dispatch | |
| Sun et al. | A customized voltage control strategy for electric vehicles in distribution networks with reinforcement learning method | |
| WO2023070293A1 (en) | Long-term scheduling method for industrial byproduct gas system | |
| CN112686571B (en) | Comprehensive intelligent energy optimization scheduling method and system based on dynamic adaptive modeling | |
| Goh et al. | An assessment of multistage reward function design for deep reinforcement learning-based microgrid energy management | |
| CN116247648A (en) | Deep reinforcement learning method for micro-grid energy scheduling under consideration of source load uncertainty | |
| CN114648176B (en) | Wind-solar power supply consumption optimization method based on data driving | |
| CN115544899B (en) | Water plant water intake pump station energy-saving scheduling method based on multi-agent deep reinforcement learning | |
| CN112132379B (en) | Economic-considered new energy cross-region consumption evaluation method and storage medium | |
| CN113809780B (en) | Micro-grid optimal scheduling method based on improved Q learning punishment selection | |
| CN107730048A (en) | A kind of random robust Optimization Scheduling of wind-powered electricity generation electric automobile association system | |
| Guo et al. | Energy management of intelligent solar parking lot with EV charging and FCEV refueling based on deep reinforcement learning | |
| CN114169593A (en) | Platform area electric vehicle ordered charging method based on load prediction and deep reinforcement learning | |
| CN117350424A (en) | Economic dispatching and electric vehicle charging strategy combined optimization method in energy internet | |
| Yang et al. | Data-driven optimal dynamic dispatch for Hydro-PV-PHS integrated power systems using deep reinforcement learning approach | |
| Xia et al. | Two‐layer optimal charging strategy for electric vehicles in old residential areas | |
| CN118381118A (en) | Data-driven Wasserstein fuzzy set-based active distribution network distribution robust day-ahead scheduling method | |
| CN112510690B (en) | Optimal scheduling method and system considering wind-fire-storage combination and demand response reward and punishment | |
| CN114169916A (en) | Market member quotation strategy making method suitable for novel power system | |
| CN117543581A (en) | Virtual power plant optimal scheduling method considering electric automobile demand response and application thereof | |
| CN117932417A (en) | Hierarchical consistency wind speed prediction method based on multi-time scale features | |
| Hong et al. | 6G based intelligent charging management for autonomous electric vehicles | |
| CN112101624A (en) | ArIMA-based electric vehicle random charging demand prediction and scheduling method | |
| CN116979579A (en) | Electric automobile energy-computing resource scheduling method based on safety constraint of micro-grid | |
| CN116780627A (en) | Micro-grid regulation and control method in building park |
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 |