Patent Application
20240059170
Electrochemical battery cells and battery packs constructed from such battery cells are used as direct current (DC) power supplies in a myriad of high-power battery electric systems. An electric vehicle is an exemplary type of battery electric system using a high-voltage propulsion battery pack constructed from an application-suitable number of cylindrical, prismatic, or pouch-style battery cells. The battery pack, which is connected to a DC voltage bus, ultimately powers one or more electric propulsion motors and associated power electronic components during battery discharging modes. During battery charging modes, a charging current is provided to the constituent battery cells of the battery pack from an offboard charging station. Thus, the charging process, when performed away from a user's home charging station, often entails locating an available charging station and scheduling a charging time.
While charging infrastructure continues to grow and evolve, drivers of electric vehicles face potential uncertainty regarding actual availability of a given charging station at a particular desired location or charging time. In order to reduce range anxiety, a driver may use an application (“app”) to help locate charging stations along a planned travel route, and to schedule a charging session at intervals along the way. However, existing “one-to-one” approaches for scheduling charging sessions, i.e., one electric vehicle seeking an open charging slot at a given charging station, may be suboptimal in terms of balancing energy supply and demand over a wider geographical area.
Disclosed herein are cloud-based methods and systems for performing dynamic optimal matching of electric vehicle (“EV”) charging demand (“EV demand”) with electric vehicle supply equipment (EVSE) charging supply levels (“EVSE supply”) within a geofenced area of interest (“AOI”). In particular, the computer-implemented solutions described in detail below provide for optimized bidirectional matching of multiple EV demands with available EVSE supplies within the geofenced AOI. Using the disclosed cloud modeling strategies, for instance, a driver of an EV may benefit from dynamic supply prediction over many different charging stations, and dynamic demand prediction over many different EVs. Dynamic bidirectional matching of multiple charging stations to multiple demands (“many-to-many”) as opposed to the aforementioned “one-to-one” matching is likewise enabled using a reinforced learning approach, e.g., using a state-of-charge (SoC) map and a charging station power map in a possible embodiment. The present teachings may enable a recommendation engine and automatic enabler for charging spot bookings, pre-conditioning activation, adaptive routing, charging station scheduling, load balancing, power sharing, and dynamic pricing, among other possible attendant benefits.
In a possible implementation, a method for dynamically matching an energy demand of a population of EVs with an energy supply of a population of charging stations within a geofenced perimeter includes receiving, via a cloud-based server, a set of EV information from each respective one of the EVs, and receiving, via the cloud-based server, a set of charging station information from each respective one of the charging stations;. The method further includes generating a state of charge (SoC) map and a charging station power map from the set of EV information and the set of charging station information, respectively, via the cloud-based server. In this embodiment, the method also includes predicting the energy supply and the energy demand as a predicted energy supply and a predicted energy demand, respectively, via the cloud-based server using the SoC map and the charging station power map. The EVs are then dynamically matched to at least one of the charging stations using a reward function, the predicted energy supply, and the predicted energy demand, including generating a rank-ordered listing for each respective one of the EVs and/or each respective one of the charging stations in a manner that maximizes an expected discounted future reward.
The geofenced perimeter may be a dynamically-adjustable polygon, with the method including using a pointer network to define the geofenced perimeter as a convex hull or minimum convex polygon.
Dynamically matching the EVs to at least one of the charging stations may include generating a table of ranked bidirectional matches, the table having a plurality of rewards including an estimated time of arrival (ETA) and an estimated time to charge (ETC) at a respective one of the charging stations. The plurality of rewards may include one or more of an estimated cost of charge (ECC), an estimated expected charge (ECC), or an estimated charging station profit (SCP).
In some aspects of the disclosure, dynamically matching the EVs to at least one of the charging stations includes calculating a combined reward value as a weighted function of the rewards, and maximizing the combined reward value across an area defined by the geofenced perimeter, e.g., using Pareto optimization. Dynamically matching the EVs to at least one of the charging stations may alternatively include performing a temporal difference (TD) model-free on-policy learning algorithm, e.g., Advantage Actor-Critic (A3C).
An aspect of the disclosure includes receiving, via the cloud-based server, contextual information characterizing or describing a vehicle type, a battery temperature, a battery age, a charging type, and local weather conditions for each respective one of the EVs, and a charging type and local weather conditions for each respective one of the charging stations.
Also disclosed herein is a cloud-based server having a processor and a computer-readable storage medium on which is recorded instructions executable by the processor. Execution of the instructions causes the processor to dynamically match an energy demand of a population of EVs with an energy supply of a population of charging stations within a geofenced perimeter by receiving a set of EV information from each respective one of the EVs, and receiving a set of charging station information from each respective one of the charging stations. generating a state of charge (SoC) map and a charging station power map from the set of EV information and the set of charging station information, respectively.
Additionally, execution of the instructions causes the processor to predict the energy supply and the energy demand as a predicted energy supply and a predicted energy demand, respectively, via the cloud-based server using the SOC map and the charging station power map, and to dynamically match the EVs to at least one of the charging stations using a reward function, the predicted energy supply, and the predicted energy demand. This includes generating a rank-ordered listing for each respective one of the EVs and/or each respective one of the charging stations in a manner that maximizes an expected discounted future reward.
Another aspect of the disclosure includes a method for dynamically matching the energy demand of the population of EVs with an energy supply of the population of charging stations within a geofenced perimeter. This embodiment of the method includes receiving, via a cloud-based server, a set of EV information from each respective one of the EVs, and receiving, via the cloud-based server, a set of charging station information from each respective one of the charging stations. The method further includes using a pointer network to define the geofenced perimeter as a dynamically-adjustable polygon, e.g., as a convex hull or minimum convex polygon, generating an SoC map and a charging station power map from the set of EV information and the set of charging station information, respectively, via the cloud-based server, and predicting the energy supply and the energy demand as a predicted energy supply and a predicted energy demand, respectively, via the cloud-based server using the SoC map and the charging station power map.
Additionally, the method in this embodiment includes dynamically matching the EVs to at least one of the charging stations using a reward function, the predicted energy supply, and the predicted energy demand, including generating a rank-ordered listing for each respective one of the EVs and/or each respective one of the charging stations as a table of ranked bidirectional matches in a manner that maximizes an expected discounted future reward, the table having a plurality of rewards including an estimated time of arrival (ETA) and an estimated time to charge (ETC) at a respective one of the charging stations and one or more of an estimated cost of charge (ECC), an estimated expected charge (ECC), or an estimated charging station profit (SCP).
The above features and advantages, and other features and attendant advantages of this disclosure, will be readily apparent from the following detailed description of illustrative examples and modes for carrying out the present disclosure when taken in connection with the accompanying drawings and the appended claims. Moreover, this disclosure expressly includes combinations and sub-combinations of the elements and features presented above and below.
The accompanying drawings, which are incorporated into and constitute a part of this specification, illustrate implementations of the disclosure which, taken together with the description, serve to explain the principles of the disclosure.
The appended drawings are not necessarily to scale, and may present a somewhat simplified representation of various preferred features of the present disclosure as disclosed herein, including, for example, specific dimensions, orientations, locations, and shapes. Details associated with such features will be determined in part by the particular intended application and use environment.
The present disclosure is susceptible of embodiment in many different forms. Representative examples of the disclosure are shown in the drawings and described herein in detail as non-limiting examples of the disclosed principles. To that end, elements and limitations described in the Abstract, Introduction, Summary, and Detailed Description sections, but not explicitly set forth in the claims, should not be incorporated into the claims, singly or collectively, by implication, inference, or otherwise.
For purposes of the present description, unless specifically disclaimed, use of the singular includes the plural and vice versa, the terms “and” and “or” shall be both conjunctive and disjunctive, and the words “including,” “containing,” “comprising,” “having,” and the like shall mean “including without limitation.” Moreover, words of approximation such as “about,” “almost,” “substantially,” “generally,” “approximately,” etc., may be used herein in the sense of “at, near, or nearly at,” or “within 0-5% of,” or “within acceptable manufacturing tolerances,” or logical combinations thereof. As used herein, a component that is “configured to” perform a specified function is capable of performing the specified function without alteration, rather than merely having potential to perform the specified function after further modification. In other words, the described hardware, when expressly configured to perform the specified function, is specifically selected, created, implemented, utilized, programmed, and/or designed for the purpose of performing the specified function.
Referring to the drawings, wherein like reference numbers refer to like features throughout the several views,
As appreciated in the art, energy supply and demand are not static parameters in the context of day-to-day vehicular operations. Rather, supply and demand vary in real-time based on factors such as time of day and the particular number of EVs 10 requiring use of a given charging station 12. The present strategy thus relies on two-way communication between the cloud-based server 50 and the various EVs 10 and charging stations 12 to optimally match supply and demand in real time. Applicable boundary constraints may be applied to effectively limit consideration and computing resources to evaluation of an area defined by the geofenced perimeter 14, 140.
As part of the present methodology, the various EVs 10 located within the geofenced perimeter 14, 140 provide signal inputs to the cloud-based server 50 of
In the performance of the present method, the cloud-based server 50 shown in
Referring to
As part of block B102, the cloud-based server 50 receives the various SoCs and constructs the noted SoC map. As contemplated herein, the SoC map may be a data file of current locations and SoCs of the EVs 10 located within or traveling through the geofenced perimeter 14, 140 of
Block B103 includes receiving, via the cloud-based server 50, a set of charging station information from each respective one of the charging stations 12 and generating an EVSE charging station power map (“CS-M”) using the charging station information, as a counterpart to the above-described EV SoC map constructed in block B102. That is, the various charging stations 12 of
At block B104, a demand prediction model (“DPM”) of the cloud-based server 50 may receive contextual information from one or more available sources 52. As contemplated herein, “contextual information” may include weather conditions, vehicle type (manufacturer, model, year, etc.), battery temperature and age, and/or charging system types and ages. Such data may be provided by the individual EVs 10 of
The method 100 as contemplated herein includes predicting the energy supply and the energy demand within the geofenced area 14, 140 via the cloud-based server 50 using the SoC map and the charging station power map. Demand prediction occurs for the area defined by the geofenced perimeter 14, 140, e.g., using machine learning as exemplified below. Outputs from block B104 may include a predicted number of charging stations 12 needed at a future time point, i.e., Ct: t+T, and a predicted charging demand at the future time point, i.e., Dt: t+T, as described in further detail below.
Block B106 of
As shown in
Turning back to
In general, block B106 of
Data-driven Optimal Demand-Supply Matching for EV Charging: a possible approach for performing block B106 of
An objective of the cloud-based server 50 is to efficiently assign an EV 10 of
r
t=ωscp×SCP−ωeta×ETA−ωetc×ETC−ωecc×ECC
with the various weighting terms ωscp, ωeta, ωetc, and ωecc reflecting a level of importance of each criterion, summing to 1 without a loss of generality.
Due to the conflicting nature between objective functions, duality is used to convert minimization (ETA, ETC, and ECC) into maximization to form the overall reward. At each time step (t), that is, the dynamic matching function performed at block B106 obtains a representation of the environment, sat, and a reward, rt. Based on this information, an action at is taken to direct EVs 10 to charging stations 12 such that the expected discounted future reward is maximized, i.e.,:
where 0<γ<1 is a time discounting factor providing a penalty to uncertainty of future awards. The cloud-based server 50 may then maximize the expected discounted future reward. This preference-based multi-objective optimization procedure involves having a preference vector or a weighting scheme that is highly subjective. Pareto optimization as appreciated in the art may also be used to find multiple tradeoff solutions and choose one using higher-level information from the user.
Referring briefly to
Pointer Networks (Ptr-Net)-based Convex Hull Dynamic Geofence: the present teachings may be applied to the fixed static geofenced perimeter 14 of
Input: a sequence of vectors x=(X1, . . . , xn)
Ptr-Net Outputs: a sequence of integer indices, c=(c1, . . . , cm) and 1≥ci≥n.
Encoder hidden states: (hi, . . . , hn)
Decoder hidden states: (si, . . . , sm) where si is the output gate after cell activation in the decoder.
The Ptr-Net applies additive attention between states and then normalizes by the softmax function to model the output conditional probability:
y
i
=p(ci|c1, . . . , x)=softmax(score (st, hi))=softmax(vaT tan h(Wa[st; hi]))
Asynchronous Advantage Actor-Critic (A3C): as understood in the art, the following symbols are commonly used in the above-noted A3C algorithm:
Exemplary A3C pseudocode appears as follows:
dθ←dθ+∇
{acute over (θ)}log(π{acute over (θ)}(a, s))(R=V{acute over (ω)}(s))
dω←dω+2∇{acute over (ω)}(R−V{acute over (ω)}(s))(R−V{acute over (ω)}(s))
The above-described solutions provide systems and methods for optimized dynamic bidirectional matching of multiple demands of charging EVs 10 with available power supplies of multiple charging stations 12, as shown in
As will be appreciated by those skilled in the art, the present disclosure provides a data-driven demand prediction approach to matching supply and demand within the geofenced area. Optimal matching between EV demand and charging station supply uses a temporal difference model-free on-policy learning algorithm, e.g., A3C, to learn the optimal pairings of EVs 10 and charging stations 12 within the geofenced area. A reward is thus obtained for each time step. The cloud-based server 50 then executes an action to match the EVs 10 with charging stations 12 such that the expected discounted future reward is maximized. Taken as a whole, the proposed solutions are intended to help reduce range anxiety and enable optimal charge scheduling from the perspective of the various EV drivers as well as charging station/infrastructure operators. These and other attendant benefits will be appreciated by those skilled in the art in view of the foregoing disclosure.
The detailed description and the drawings or figures are supportive and descriptive of the present teachings, but the scope of the present teachings is defined solely by the claims. While some of the best modes and other embodiments for carrying out the present teachings have been described in detail, various alternative designs and embodiments exist for practicing the present teachings defined in the appended claims. Moreover, this disclosure expressly includes combinations and sub-combinations of the elements and features presented above and below.
