Charging load prediction method and device based on dynamic game and space-time state evolution

Abstract

The invention provides a charging load prediction method and device based on dynamic game and spatio-temporal state evolution, and the method comprises the steps: building a simultaneous time recursion equation under a discrete spatio-temporal framework based on a multi-source impact factor parameter; the simultaneous time recursion equation is used for describing the joint evolution process of at least two different types of driver states under a discrete space-time framework, the influence of external factors on individual states is reflected, the game logic in a group is also reflected, the prediction result conforms to the real driver behavior logic, and the prediction accuracy is improved. A joint recursion relation is determined and obtained on the basis of the time recursion equation through iterative calculation on a discrete time axis until the joint recursion relation reaches a stable state, a curve of the total charging power of the target charging station changing along with time is extracted from the stable state, and a charging load prediction result is obtained. It is ensured that the prediction result is stable output after long-term evolution of the system and fits the load law in actual operation of the charging station.

Description

Technical Field

[0001] This invention relates to the field of power system optimization scheduling technology, and in particular to a charging load prediction method and device based on dynamic game theory and spatiotemporal state evolution. Background Technology

[0002] With the rapid growth in the number of new energy vehicles, vehicle-to-grid (V2G) interaction has become a key technological direction for improving the flexibility of the power system and promoting the consumption of renewable energy. Demand-side response, as the core mechanism of V2G, guides users to adjust their electricity consumption behavior through price signals, thereby achieving a balance between power grid supply and demand. Currently, several related technology patents have been filed, but significant shortcomings still exist.

[0003] CN120566606A proposes a vehicle-to-grid (V2G) interaction method for integrated photovoltaic-storage-charging-discharge power stations that considers demand-side response. While this patent also focuses on demand-side response, it suffers from two core flaws: first, price setting uses a fixed time-segmentation model, failing to consider the dynamic characteristics of intraday load fluctuations, resulting in low price signal transmission efficiency; second, demand response forecasting is based solely on historical load data, without incorporating dynamic adjustment factors based on user behavior, making the forecast accuracy insufficient for large-scale grid access scenarios. CN119582287A provides a V2G interaction control method and related equipment based on multi-agent collaboration in the power grid, but its limitation lies in the reliance on centralized decision-making for scheduling strategies. The current architecture has high requirements for communication bandwidth, which can easily lead to scheduling delays when electric vehicles are connected on a large scale. At the same time, the price adjustment range adopts a fixed threshold design and lacks a linkage mechanism with the real-time operation status of the power grid, making it difficult to adapt to complex power grid conditions. CN111798121A provides a distributed collaborative optimization method for energy management and scheduling of electric vehicles, which uses game theory. However, it has inherent defects: First, it assumes that the participants have complete information, which does not conform to the reality of user privacy protection and information asymmetry in actual vehicle-grid interaction. Second, as the number of participants increases, the computational complexity of solving the game equilibrium increases exponentially, making it difficult to guarantee real-time performance. Third, it does not consider the dynamic evolution characteristics of the strategies among the participants, which can easily lead to getting trapped in local optima.

[0004] In summary, existing technologies have significant shortcomings in terms of price guidance flexibility, collaborative optimization accuracy, and adaptability to complex scenarios. There is an urgent need for a charging load forecasting method that can dynamically adapt to grid conditions and take into account the interests of multiple stakeholders. Summary of the Invention

[0005] This invention provides a charging load prediction method and apparatus based on dynamic game theory and spatiotemporal state evolution to solve the problems mentioned in the background art.

[0006] A charging load prediction method based on dynamic game theory and spatiotemporal state evolution includes: S1: Divide the continuous prediction time range into discrete time units, obtain discrete space state units based on continuous physical state variables, and align the discrete time units and discrete space state units in time and space to obtain a discrete time-space framework. S2: Explicitly parameterize the factors affecting charging decisions to obtain multi-source influencing factor parameters; S3: Based on multi-source influence factor parameters, establish simultaneous time recursive equations in a discrete time-space framework. The simultaneous time recursive equations are used to describe the joint evolution process of at least two different types of driver states in a discrete time-space framework. S4: Iteratively calculate the joint recursive relationship based on the time recursive equation on the discrete time axis until the joint recursive relationship reaches a steady state. Extract the curve of the total charging power of the target charging station changing with time from the steady state to obtain the charging load prediction result.

[0007] Preferably, in step S3, based on multi-source influence factor parameters, a simultaneous time recursive equation is established within a discrete spatiotemporal framework, including: Based on multi-source influencing factor parameters, a recursive equation for the equivalent value of total energy, an equation for the number of drivers in the driving state, and an equation for the number of drivers in the charging state are established within a discrete spatiotemporal framework. By adding the equations for the number of drivers in driving mode and the number of drivers in charging mode to the recursive equation for the total energy equivalent value, we obtain a simultaneous time recursive equation.

[0008] Preferably, the process of establishing the recursive equation for the total energy equivalent value is as follows: Based on the equivalent total energy value at the initial moment, and combined with the state data at the initial moment, the update is performed as follows: First, subtract the power consumption of the driver in the initial driving state, and add the power replenishment of the driver in the initial charging state to obtain the equivalent value of the total power at the next moment. Establish a recursive equation for obtaining the equivalent value of the total power. Among them, the driver's battery status is converted by the expected conversion rate coefficient of battery status.

[0009] Preferably, the process of establishing the driver quantity equation for driving status is as follows: The driver's activity level is represented by subtracting the driver's desire to rest in the current time period from 1, and is combined with the current total battery equivalent value and the number of passenger orders as the target parameter; Based on the transformation of the objective parameters by a nonlinear response function that satisfies the condition that the first derivative is positive, the second derivative is negative, and has an asymptote, the equation for the number of drivers in the driving state is obtained.

[0010] Preferably, the process of establishing the equation for the number of drivers in the charging state is as follows: Establish an index factor for rest desire based on 1 minus the current value of drivers' rest desire in the current time period, and establish a function term related to the difference between the total number of active drivers in the surrounding area and the number of drivers charging at present, as well as related functions corresponding to multiple factors that affect the charging amount in the current time period. Multiplying the exponential factor, the function term, and the related function together, we obtain the equation for the number of drivers in the charging state.

[0011] Preferably, the parameters of driver rest desire, the curve of passenger order quantity, the total number of active drivers in the vicinity, and the internal parameters of the relevant functions are all obtained by dynamic calibration through gradient descent based on historical data.

[0012] Preferably, the factors affecting the charging amount in the current time period are as follows: The charging amount in the current time period is negatively correlated with the electricity price level in the current time period, and the charging amount in the current time period is positively correlated with the charging electricity price at multiple adjacent times after the current time period, which is taken as the first factor affecting the charging amount in the current time period. The number of drivers with insufficient battery power is taken as the maximum possible value of the number of charging drivers, and is used as the second factor affecting the charging volume in the current time period. The first and second factors are considered as multiple factors affecting the charging amount in the current time period.

[0013] Preferably, an index factor for rest desire is established based on 1 minus the current value of the driver's rest desire in the current time period, including: Get the current value of the driver's desire to rest in the current time period minus 1, and use it as the numerator. Get the reference values ​​of the driver's desire to rest in all other time periods minus 1 respectively. Perform a geometric mean on these reference values ​​and use it as the denominator to obtain the comprehensive product term. Perform an exponential operation on the comprehensive product term to obtain the exponential factor; The exponent is a trainable parameter used to control the impact of the comprehensive product term on the charging amount.

[0014] A charging load prediction device based on dynamic game theory and spatiotemporal state evolution includes: The spatiotemporal discrete module is used to divide the continuous prediction time range into discrete time units, obtain discrete spatial state units based on continuous physical state variables, and perform spatiotemporal alignment of discrete time units and discrete spatial state units to obtain a discrete spatiotemporal framework. The parameter definition module is used to explicitly parameterize the factors that affect charging decisions, and obtain the parameters of multi-source influencing factors. The recursive equation establishment module is used to establish simultaneous time recursive equations in a discrete time-space framework based on multi-source influence factor parameters. The simultaneous time recursive equations are used to describe the joint evolution process of at least two different types of driver states in a discrete time-space framework. The iterative prediction module is used to iteratively calculate the joint recursive relationship determined based on the time recursive equation on the discrete time axis until the joint recursive relationship reaches a steady state. From the steady state, the curve of the total charging power of the target charging station changing with time is extracted to obtain the prediction result of the charging load.

[0015] Compared with the prior art, the present invention has achieved the following beneficial effects: The joint recursive relationship is determined by iterative calculation on the discrete time axis based on the time recursive equation until the joint recursive relationship reaches a steady state. From the steady state, the curve of the total charging power of the target charging station changing with time is extracted to obtain the charging load prediction result. For example, the joint recursive relationship is determined by iterative calculation on the discrete time axis based on the time recursive equation. For example, starting from the initial state distribution at t=0, the recursive equation is iteratively calculated until t=47. The iteration is repeated until the stability condition is met (the fluctuation range of the total charging power of all spatial states is ≤4% within 3 consecutive time units). Finally, after 72 iterations, a steady state is reached. The total number of charging drivers in each time unit t is extracted. Combined with the average charging power of 60kW per electric vehicle, the total charging power is calculated, and a 24-hour charging load curve is synthesized.

[0016] The beneficial effects of the above design scheme are as follows: By dividing the continuous prediction time range into discrete time units, discrete spatial state units are obtained based on continuous physical state variables. Spatiotemporal alignment of the discrete time units and discrete spatial state units yields a discrete spatiotemporal framework, ensuring that the evolution of the driver's state always occurs within a unified spatiotemporal dimension. This avoids prediction bias caused by the disconnect between time and state, transforming abstract spatiotemporal into a quantifiable and computable discrete set, providing a foundation for the subsequent establishment and iteration of recursive equations. By explicitly parameterizing the factors influencing charging decisions, multi-source influencing factor parameters are obtained. Explicit parameterization transforms fuzzy influencing factors into explicit quantitative parameters, making the impact of each factor on the prediction results traceable and adjustable. Based on multi-source influencing factor parameters, a discrete spatiotemporal framework is established... A simultaneous time recursive equation is used to describe the joint evolution of at least two different types of driver states within a discrete spatiotemporal framework. This equation reflects both the influence of external factors on individual states and the game logic within the group, ensuring that the prediction results align with the actual driver behavior logic. The joint recursive relationship is determined by iterative calculation based on the time recursive equation on the discrete time axis until the joint recursive relationship reaches a stable state. From the stable state, the curve of the total charging power of the target charging station changing over time is extracted to obtain the charging load prediction result. This ensures that the prediction result is a stable output after the long-term evolution of the system, closely matching the load patterns in the actual operation of the charging station. It directly matches the actual needs of charging station operation scheduling and grid load prediction, requiring no additional data conversion and can be directly used to guide time-of-use pricing and charging resource allocation.

[0017] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in this application.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a charging load prediction method based on dynamic game theory and spatiotemporal state evolution in an embodiment of the present invention; Figure 2 This is a structural diagram of a charging load prediction device based on dynamic game theory and spatiotemporal state evolution in an embodiment of the present invention. Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] Example 1: This embodiment of the invention provides a charging load prediction method based on dynamic game theory and spatiotemporal state evolution, such as... Figure 1 As shown, it includes: S1: Divide the continuous prediction time range into discrete time units, obtain discrete space state units based on continuous physical state variables, and align the discrete time units and discrete space state units in time and space to obtain a discrete time-space framework. S2: Explicitly parameterize the factors affecting charging decisions to obtain multi-source influencing factor parameters; S3: Based on multi-source influence factor parameters, establish simultaneous time recursive equations in a discrete time-space framework. The simultaneous time recursive equations are used to describe the joint evolution process of at least two different types of driver states in a discrete time-space framework. S4: Iteratively calculate the joint recursive relationship based on the time recursive equation on the discrete time axis until the joint recursive relationship reaches a steady state. Extract the curve of the total charging power of the target charging station changing with time from the steady state to obtain the charging load prediction result.

[0022] In this embodiment, physical state variables include, for example, driver battery level and time preference. In the driver battery level dimension, these variables include, for example, low battery level, medium battery level, and high battery level. In the time preference dimension, these variables include, for example, peak preference and off-peak preference. Discrete spatial state units include, for example, low battery level + peak preference and medium battery level + off-peak preference.

[0023] In this embodiment, the continuous prediction time range is divided into discrete time units, for example, 24 hours is divided into 48 discrete time units, and each unit has a granularity of 30 minutes.

[0024] In this embodiment, the discrete spatiotemporal framework divides the continuous prediction time range (24h) into 48 discrete time units of 30 minutes each, denoted as t∈{0,1,2,...,47}. Based on two continuous physical state variables—driver's battery charge SOC (0%~20% low, 20%~80% medium, 80%~100% high) and time preference (peak / valley)—it is divided into 6 types of discrete spatial state units. Spatiotemporal alignment is completed by mapping time unit t to spatial state s one-to-one, constructing a discrete spatiotemporal framework Ω={(t,s)|t=0~47,s=1~6}, which provides a unified computational carrier for the recursive equation.

[0025] In this embodiment, the multi-source influencing factor parameters include external incentive parameters (passenger order density), competitive environment parameters (price difference of surrounding stations), and driver decision parameters (price sensitivity, desire to rest).

[0026] In this embodiment, the joint recursive relationship is determined by iterative calculation on the discrete time axis based on the time recursive equation until the joint recursive relationship reaches a steady state. From the steady state, the curve of the total charging power of the target charging station changing with time is extracted to obtain the charging load prediction result. For example, the joint recursive relationship is determined by iterative calculation on the discrete time axis based on the time recursive equation. For example, starting from the initial state distribution at t=0, the recursive equation is iteratively calculated until t=47. The iteration is repeated until the stability condition is met (the fluctuation range of the total charging power of all spatial states is ≤4% within 3 consecutive time units). Finally, after 72 iterations, a steady state is reached. The total number of charging drivers in each time unit t is extracted. Combined with the average charging power of 60kW per electric vehicle, the total charging power is calculated, and a 24-hour charging load curve is synthesized.

[0027] The beneficial effects of the above design scheme are as follows: By dividing the continuous prediction time range into discrete time units, discrete spatial state units are obtained based on continuous physical state variables. Spatiotemporal alignment of the discrete time units and discrete spatial state units yields a discrete spatiotemporal framework, ensuring that the evolution of the driver's state always occurs within a unified spatiotemporal dimension. This avoids prediction bias caused by the disconnect between time and state, transforming abstract spatiotemporal into a quantifiable and computable discrete set, providing a foundation for the subsequent establishment and iteration of recursive equations. By explicitly parameterizing the factors influencing charging decisions, multi-source influencing factor parameters are obtained. Explicit parameterization transforms fuzzy influencing factors into explicit quantitative parameters, making the impact of each factor on the prediction results traceable and adjustable. Based on multi-source influencing factor parameters, a discrete spatiotemporal framework is established... A simultaneous time recursive equation is used to describe the joint evolution of at least two different types of driver states within a discrete spatiotemporal framework. This equation reflects both the influence of external factors on individual states and the game logic within the group, ensuring that the prediction results align with the actual driver behavior logic. The joint recursive relationship is determined by iterative calculation based on the time recursive equation on the discrete time axis until the joint recursive relationship reaches a stable state. From the stable state, the curve of the total charging power of the target charging station changing over time is extracted to obtain the charging load prediction result. This ensures that the prediction result is a stable output after the long-term evolution of the system, closely matching the load patterns in the actual operation of the charging station. It directly matches the actual needs of charging station operation scheduling and grid load prediction, requiring no additional data conversion and can be directly used to guide time-of-use pricing and charging resource allocation.

[0028] Example 2: Based on Example 1, this embodiment of the invention provides a charging load prediction method based on dynamic game theory and spatiotemporal state evolution. In step S3, based on multi-source influencing factor parameters, a simultaneous time recursive equation is established within a discrete spatiotemporal framework, including: Based on multi-source influencing factor parameters, a recursive equation for the equivalent value of total energy, an equation for the number of drivers in the driving state, and an equation for the number of drivers in the charging state are established within a discrete spatiotemporal framework. By adding the equations for the number of drivers in driving mode and the number of drivers in charging mode to the recursive equation for the total energy equivalent value, we obtain a simultaneous time recursive equation.

[0029] The beneficial effects of the above design scheme are: by setting separate equations for the equivalent value of total power and the two driver states, the core indicators are quantified in a targeted manner, avoiding variable confusion, improving the accuracy of single variable calculation, forming a complete logical chain after being combined, and ensuring coherent data transmission. This provides a self-consistent framework for iterative calculation on the discrete time axis, ensuring the steady-state convergence of the system and supporting the accuracy of charging load prediction.

[0030] Example 3: Based on Example 2, this embodiment of the invention provides a charging load prediction method based on dynamic game theory and spatiotemporal state evolution. The process of establishing the recursive equation for the equivalent value of total power is as follows: Based on the equivalent total energy value at the initial moment, and combined with the state data at the initial moment, the update is performed as follows: First, subtract the power consumption of the driver in the initial driving state, and add the power replenishment of the driver in the initial charging state to obtain the equivalent value of the total power at the next moment. Establish a recursive equation for obtaining the equivalent value of the total power. The purpose of converting the driver's battery status during driving is to reflect the battery loss during driving by using the expected conversion rate coefficient of battery status.

[0031] In this embodiment, the formula for the recursive equation of the total energy equivalent value is as follows: ; in, for The equivalent total charge at any given moment. for The equivalent total charge at any given moment. for The driver's battery consumption at any given moment while driving. for The driver's battery level is constantly being replenished as needed.

[0032] In this embodiment, the recursive equation for the total energy equivalent value realizes the recursive evolution of the total energy equivalent value over time.

[0033] The beneficial effects of the above design scheme are as follows: Based on the equivalent value of the total power at the initial moment and combined with the real-time update of the driver's status, the power evolution is directly linked to the driver's behavior, avoiding abstract power calculations that are detached from the actual state. This makes the equivalent value of the total power more realistically reflect the actual power level of the driver group. By adjusting the power consumption of the driver in the driving state through the expected conversion rate coefficient of power state, it adapts to the fluctuation characteristics of power loss during actual driving. The power replenishment of the driver in the charging state is directly included in the recursion, and the positive gain of charging behavior on the total power is quantified in real time. This allows the equivalent value of the total power to be dynamically adjusted with the charging action, ensuring a strong correlation between power changes and charging behavior. This provides accurate power basis for subsequent driver status decisions and ensures the accuracy of the entire charging load prediction model.

[0034] Example 4: Based on Example 2, this embodiment of the invention provides a charging load prediction method based on dynamic game theory and spatiotemporal state evolution. The process of establishing the driver quantity equation for the driving state is as follows: The driver's activity level is represented by subtracting the driver's desire to rest in the current time period from 1, and is combined with the current total battery equivalent value and the number of passenger orders as the target parameter; Based on the transformation of the objective parameters by a nonlinear response function that satisfies the condition that the first derivative is positive, the second derivative is negative, and has an asymptote, the equation for the number of drivers in the driving state is obtained.

[0035] In this embodiment, the driver quantity equation based on driving status is used to quantify the impact of order demand and driver status on driving behavior.

[0036] In this embodiment, the formula for the number of drivers in the driving state is as follows: ; in, This represents the number of drivers in the driving state at time t. This represents the driver's desire to rest at time t, and characterizes the driver's activity level. This represents the number of passenger orders at time t. ( ) represents a nonlinear response function that satisfies the condition that the first derivative is positive, the second derivative is negative, and it has an asymptote.

[0037] The beneficial effects of the above design scheme are as follows: The driver's activity level is represented by subtracting the driver's rest desire from 1, directly linking it to the driver's work-rest preferences. This ensures that the quantification of driving intention aligns with real physiological and behavioral patterns. Combining the current total battery equivalent value and the number of passenger orders as target parameters, it comprehensively captures the core variables driving drivers to choose to drive, considering both objective conditions and external demands. This avoids the bias caused by a single factor. By transforming the target parameters through a nonlinear response function that satisfies a positive first derivative, a negative second derivative, and an asymptote, an equation for the number of drivers in driving status is obtained. Using a specific nonlinear response function, the positive first derivative reflects the positive driving force of the variable, the negative second derivative reflects the diminishing marginal effect, and the asymptote limits the upper limit of the number, adapting to the actual scenario of a limited number of drivers, avoiding calculation results exceeding a reasonable range, and improving the accuracy of the equation.

[0038] Example 5: Based on Example 2, this embodiment of the invention provides a charging load prediction method based on dynamic game theory and spatiotemporal state evolution. The process of establishing the equation for the number of drivers in the charging state is as follows: Establish an index factor for rest desire based on 1 minus the current value of drivers' rest desire in the current time period, and establish a function term related to the difference between the total number of active drivers in the surrounding area and the number of drivers charging at present, as well as related functions corresponding to multiple factors that affect the charging amount in the current time period. Multiplying the exponential factor, the function term, and the related function together, we obtain the equation for the number of drivers in the charging state.

[0039] In this embodiment, the function term reflects the current number of drivers available for charging.

[0040] In this embodiment, the index factor reflects the difference in the impact of rest intention on charging choice at different times.

[0041] In this embodiment, the formula for calculating the number of drivers in the charging state is as follows: ; in, This represents the total number of active drivers in the surrounding area at time t. This represents the current number of charging drivers at time t. This represents the relevant functions, including charging pile usage saturation based on whether there are empty orders, orders, weather conditions, traffic conditions, holidays, and other surrounding charging stations.

[0042] The beneficial effects of the above design scheme are as follows: It establishes an index factor for rest desire based on 1 minus the current value of drivers' rest desire in the current time period, directly quantifying the impact of drivers' work-rest preferences on charging choices, making the equation conform to real behavioral patterns; it establishes a function term related to the difference between the total number of active drivers in the surrounding area and the number of drivers currently charging, accurately reflecting the remaining number of drivers who can participate in charging, avoiding the number of charging drivers exceeding the actual supply range; and it multiplies multiple factors affecting the charging volume in the current time period, covering core factors of electricity prices, and improving the accuracy of calculating the number of drivers in charging status.

[0043] Example 6: Based on Example 5, this embodiment of the invention provides a charging load prediction method based on dynamic game theory and spatiotemporal state evolution. The parameters of driver rest desire, the curve of passenger order quantity, the total number of active drivers in the surrounding area, and the internal parameters of related functions are all obtained by dynamic calibration through gradient descent based on historical data.

[0044] In this embodiment, the specific process of dynamic calibration via gradient descent involves obtaining measured data from all time periods over the past 7 days for the target charging station, using the mean square error between predicted and actual loads as the loss function, employing mini-batch gradient descent with a learning rate of 0.01 and 500 iterations to calibrate parameters such as driver rest desire, passenger order quantity curves, total number of active drivers in the vicinity, and the internal parameters of relevant functions. Specifically, for example... Let represent the total number of active drivers in the surrounding area at time t, specifically calculated as a*d(t), where d(t) represents the initial total number of active drivers in the surrounding area. Here, 'a' is calibrated dynamically using gradient descent, thereby achieving [the desired result]. The calibration principle for this parameter is the same as that for the other parameters.

[0045] In this embodiment, the driver's desire to rest is obtained through dynamic calibration of gradient descent based on historical data, which is strictly distinguished from the traditional static rule model and endows the model with adaptive and self-learning AI attributes.

[0046] In this embodiment, the relevant functions are, for example, function terms, related functions, etc., covering all the functions required to establish the simultaneous time recursive equations, including empty order runs, order runs, weather conditions, traffic conditions, holidays, and the saturation of charging pile usage in other surrounding stations.

[0047] The beneficial effects of the above design scheme are: based on the historical data of the target station, the rest desire parameter is made to fit the actual work and rest patterns of the drivers serving the station, and can respond in real time to the long-term changes in the drivers' work and rest habits. At the same time, the curve of passenger order quantity and the internal parameters of related functions are all dynamically calibrated based on historical data through gradient descent, avoiding the rigidity problem of general fixed values, improving the adaptability of parameters to the scenario, supporting continuous calibration with the accumulation of historical data, and ensuring the stability of the model's long-term prediction effect.

[0048] Example 7: Based on Example 5, this embodiment of the invention provides a charging load prediction method based on dynamic game theory and spatiotemporal state evolution. The specific factors affecting the charging amount in the current time period are as follows: The first factor influencing the charging amount in the current time period is based on the fact that the charging amount in the current time period is negatively correlated with the electricity price level in the current time period, and positively correlated with the charging electricity price at multiple adjacent times after the current time period. The number of drivers with insufficient battery power is taken as the maximum possible value of the number of charging drivers, and is used as the second factor affecting the charging volume in the current time period. The first and second factors are considered as multiple factors affecting the charging amount in the current time period.

[0049] The beneficial effects of the above design scheme are as follows: The first factor takes into account the impact of electricity prices at the current and subsequent adjacent times, accurately captures the spatiotemporal shift behavior of drivers due to price differences, and conforms to the real decision-making law. The second factor uses the number of drivers with insufficient electricity as the upper limit of charging volume, avoids the prediction from exceeding the actual number of drivers who can charge, and ensures the rationality of the results. The two factors cover price incentives and supply constraints respectively, and synergistically quantify the core impact, so that the prediction is both in line with behavioral logic and does not deviate from the actual scenario.

[0050] Example 8: Based on Example 5, this embodiment of the invention provides a charging load prediction method based on dynamic game theory and spatiotemporal state evolution, establishing an index factor for rest desire based on 1 minus the current value of the driver's rest desire in the current time period, including: Get the current value of the driver's desire to rest in the current time period minus 1, and use it as the numerator. Get the reference values ​​of the driver's desire to rest in all other time periods minus 1 respectively. Perform a geometric mean on these reference values ​​and use it as the denominator to obtain the comprehensive product term. Perform an exponential operation on the comprehensive product term to obtain the exponential factor; The exponent is a trainable parameter used to control the impact of the comprehensive product term on the charging amount.

[0051] The beneficial effects of the above design scheme are: by comparing the current activity level with that at other times, it objectively reflects the relative differences in current work and rest preferences, which fits the comparative logic of drivers' charging decisions; the geometric mean can smooth out extreme values, more objectively reflect the average level at other times, avoid deviations caused by abnormal data; the trainable index parameters can dynamically adjust the intensity of influence, adapt to the work and rest characteristics of drivers at different stations, avoid the rigidity of fixed coefficients, and improve the fit and accuracy of charging status calculation.

[0052] Example 9: This embodiment of the invention provides a charging load prediction device based on dynamic game theory and spatiotemporal state evolution, such as... Figure 2 As shown, it includes: The spatiotemporal discrete module is used to divide the continuous prediction time range into discrete time units, obtain discrete spatial state units based on continuous physical state variables, and perform spatiotemporal alignment of discrete time units and discrete spatial state units to obtain a discrete spatiotemporal framework. The parameter definition module is used to explicitly parameterize the factors that affect charging decisions, and obtain the parameters of multi-source influencing factors. The recursive equation establishment module is used to establish simultaneous time recursive equations in a discrete time-space framework based on multi-source influence factor parameters. The simultaneous time recursive equations are used to describe the joint evolution process of at least two different types of driver states in a discrete time-space framework. The iterative prediction module is used to iteratively calculate the joint recursive relationship determined based on the time recursive equation on the discrete time axis until the joint recursive relationship reaches a steady state. From the steady state, the curve of the total charging power of the target charging station changing with time is extracted to obtain the prediction result of the charging load.

[0053] In this embodiment, physical state variables include, for example, driver battery level and time preference. In the driver battery level dimension, these variables include, for example, low battery level, medium battery level, and high battery level. In the time preference dimension, these variables include, for example, peak preference and off-peak preference. Discrete spatial state units include, for example, low battery level + peak preference and medium battery level + off-peak preference.

[0054] In this embodiment, the continuous prediction time range is divided into discrete time units, for example, 24 hours is divided into 48 discrete time units, and each unit has a granularity of 30 minutes.

[0055] In this embodiment, the multi-source influencing factor parameters include external incentive parameters (passenger order density), competitive environment parameters (price difference of surrounding stations), and driver decision parameters (price sensitivity, desire to rest).

[0056] In this embodiment, the joint recursive relationship is determined by iterative calculation on the discrete time axis based on the time recursive equation until the joint recursive relationship reaches a steady state. From the steady state, the curve of the total charging power of the target charging station changing with time is extracted to obtain the charging load prediction result. For example, the joint recursive relationship is determined by iterative calculation on the discrete time axis based on the time recursive equation. For example, starting from the initial state distribution at t=0, the recursive equation is iteratively calculated until t=47. The iteration is repeated until the stability condition is met (the fluctuation range of the total charging power of all spatial states is ≤4% within 3 consecutive time units). Finally, after 72 iterations, a steady state is reached. The total number of charging drivers in each time unit t is extracted. Combined with the average charging power of 60kW per electric vehicle, the total charging power is calculated, and a 24-hour charging load curve is synthesized.

[0057] The beneficial effects of the above design scheme are as follows: By dividing the continuous prediction time range into discrete time units, discrete spatial state units are obtained based on continuous physical state variables. Spatiotemporal alignment of the discrete time units and discrete spatial state units yields a discrete spatiotemporal framework, ensuring that the evolution of the driver's state always occurs within a unified spatiotemporal dimension. This avoids prediction bias caused by the disconnect between time and state, transforming abstract spatiotemporal into a quantifiable and computable discrete set, providing a foundation for the subsequent establishment and iteration of recursive equations. By explicitly parameterizing the factors influencing charging decisions, multi-source influencing factor parameters are obtained. Explicit parameterization transforms fuzzy influencing factors into explicit quantitative parameters, making the impact of each factor on the prediction results traceable and adjustable. Based on multi-source influencing factor parameters, a discrete spatiotemporal framework is established... A simultaneous time recursive equation is used to describe the joint evolution of at least two different types of driver states within a discrete spatiotemporal framework. This equation reflects both the influence of external factors on individual states and the game logic within the group, ensuring that the prediction results align with the actual driver behavior logic. The joint recursive relationship is determined by iterative calculation based on the time recursive equation on the discrete time axis until the joint recursive relationship reaches a stable state. From the stable state, the curve of the total charging power of the target charging station changing over time is extracted to obtain the charging load prediction result. This ensures that the prediction result is a stable output after the long-term evolution of the system, closely matching the load patterns in the actual operation of the charging station. It directly matches the actual needs of charging station operation scheduling and grid load prediction, requiring no additional data conversion and can be directly used to guide time-of-use pricing and charging resource allocation.

[0058] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this application and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A charging load prediction method based on dynamic game theory and spatiotemporal state evolution, characterized in that, include: S1: Divide the continuous prediction time range into discrete time units, obtain discrete space state units based on continuous physical state variables, and align the discrete time units and discrete space state units in time and space to obtain a discrete time-space framework. S2: Explicitly parameterize the factors affecting charging decisions to obtain multi-source influencing factor parameters; S3: Based on multi-source influence factor parameters, establish simultaneous time recursive equations in a discrete time-space framework. The simultaneous time recursive equations are used to describe the joint evolution process of at least two different types of driver states in a discrete time-space framework. S4: Iteratively calculate the joint recursive relationship based on the time recursive equation on the discrete time axis until the joint recursive relationship reaches a steady state. Extract the curve of the total charging power of the target charging station changing with time from the steady state to obtain the charging load prediction result.

2. The charging load prediction method based on dynamic game theory and spatiotemporal state evolution according to claim 1, characterized in that, In S3, based on multi-source influence factor parameters, a simultaneous time recursive equation is established within a discrete spatiotemporal framework, including: Based on multi-source influencing factor parameters, a recursive equation for the equivalent value of total energy, an equation for the number of drivers in the driving state, and an equation for the number of drivers in the charging state are established within a discrete spatiotemporal framework. By adding the equations for the number of drivers in driving mode and the number of drivers in charging mode to the recursive equation for the total energy equivalent value, we obtain a simultaneous time recursive equation.

3. The charging load prediction method based on dynamic game theory and spatiotemporal state evolution according to claim 2, characterized in that, The process of establishing the recursive equation for the equivalent value of total electrical energy is as follows: Based on the equivalent total energy value at the initial moment, and combined with the state data at the initial moment, the update is performed as follows: First, subtract the power consumption of the driver in the initial driving state, and add the power replenishment of the driver in the initial charging state to obtain the equivalent value of the total power at the next moment. Establish a recursive equation for obtaining the equivalent value of the total power. Among them, the driver's battery status is converted by the expected conversion rate coefficient of battery status.

4. The charging load prediction method based on dynamic game theory and spatiotemporal state evolution according to claim 2, characterized in that, The process of establishing the driver quantity equation for driving status is as follows: The driver's activity level is represented by subtracting the driver's desire to rest in the current time period from 1, and is combined with the current total battery equivalent value and the number of passenger orders as the target parameter; Based on the transformation of the objective parameters by a nonlinear response function that satisfies the condition that the first derivative is positive, the second derivative is negative, and has an asymptote, the equation for the number of drivers in the driving state is obtained.

5. The charging load prediction method based on dynamic game theory and spatiotemporal state evolution according to claim 2, characterized in that, The process of establishing the equation for the number of drivers in the charging state is as follows: Establish an index factor for rest desire based on 1 minus the current value of drivers' rest desire in the current time period, and establish a function term related to the difference between the total number of active drivers in the surrounding area and the number of drivers charging at present, as well as related functions corresponding to multiple factors that affect the charging amount in the current time period. Multiplying the exponential factor, the function term, and the related function together, we obtain the equation for the number of drivers in the charging state.

6. The charging load prediction method based on dynamic game theory and spatiotemporal state evolution according to claim 5, characterized in that, The parameters of driver rest desire, the curve of passenger order quantity, the total number of active drivers in the vicinity, and the internal parameters of the relevant functions are all obtained by dynamic calibration through gradient descent based on historical data.

7. The charging load prediction method based on dynamic game theory and spatiotemporal state evolution according to claim 5, characterized in that, The specific factors affecting the charging amount in the current time period are as follows: The first factor influencing the charging amount in the current time period is based on the fact that the charging amount in the current time period is negatively correlated with the electricity price level in the current time period, and positively correlated with the charging electricity price at multiple adjacent times after the current time period. The number of drivers with insufficient battery power is taken as the maximum possible value of the number of charging drivers, and is used as the second factor affecting the charging volume in the current time period. The first and second factors are considered as multiple factors affecting the charging amount in the current time period.

8. The charging load prediction method based on dynamic game theory and spatiotemporal state evolution according to claim 5, characterized in that, An exponential factor for rest desire is established, based on 1 minus the current value of the driver's rest desire in the current time period, including: Get the current value of the driver's desire to rest in the current time period minus 1, and use it as the numerator. Also get the reference values ​​of the driver's desire to rest in all other times minus 1. Perform a geometric mean on these reference values ​​and use it as the denominator to obtain the comprehensive product term. Perform an exponential operation on the comprehensive product term to obtain the exponential factor; The exponent is a trainable parameter used to control the impact of the comprehensive product term on the charging amount.

9. A charging load prediction device based on dynamic game theory and spatiotemporal state evolution, used to implement the charging load prediction method based on dynamic game theory and spatiotemporal state evolution as described in claim 1, characterized in that, include: The spatiotemporal discrete module is used to divide the continuous prediction time range into discrete time units, obtain discrete spatial state units based on continuous physical state variables, and perform spatiotemporal alignment of discrete time units and discrete spatial state units to obtain a discrete spatiotemporal framework. The parameter definition module is used to explicitly parameterize the factors that affect charging decisions, and obtain the parameters of multi-source influencing factors. The recursive equation establishment module is used to establish simultaneous time recursive equations in a discrete time-space framework based on multi-source influence factor parameters. The simultaneous time recursive equations are used to describe the joint evolution process of at least two different types of driver states in a discrete time-space framework. The iterative prediction module is used to iteratively calculate the joint recursive relationship determined based on the time recursive equation on the discrete time axis until the joint recursive relationship reaches a steady state. From the steady state, the curve of the total charging power of the target charging station changing with time is extracted to obtain the prediction result of the charging load.

Images