Electric vehicle demand response ratio optimization method and system considering dual uncertainty

Abstract

The invention discloses an electric vehicle demand response ratio optimization method and system considering dual uncertainty. A dual-layer ratio optimization model comprehensively considering source load uncertainty and response uncertainty is constructed. The upper layer takes a power grid operator as a main body, in order to consider the uncertainty of the source load, a random optimization method is adopted to construct a demand response ratio optimization model, the IBDR ratio, the energy electricity price and the excitation electricity price are optimized and transmitted to the lower layer EVA, and the IBDR ratio is defined as the ratio of the EVA load participating in the IBDR to all the EVA loads. And the lower layer takes EVA as a main body, puts forward a DRIN (Demand Response Inference Network) based on data driving, reasones an EVA response strategy aiming at the PBDR and the IBDR, outputs a mean value and a variance of the EVA response strategy, represents the EVA demand response strategy considering the response uncertainty by using the mean value, and returns the EVA demand response strategy to an upper layer power grid operator. And repeatedly iterating the upper layer and the lower layer until the maximum number of iterations is reached or the total cost change of the power grid operator is less than a threshold value, and finally outputting the IBDR ratio and the total cost of the power grid operator.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system optimization, and in particular relates to an electric vehicle demand response ratio optimization method and system considering dual uncertainty. Background Art

[0002] As the proportion of new energy generation on the power generation side gradually increases, the penetration rate of new energy such as wind power and photovoltaic power increases, and its randomness and volatility increase the pressure on the power system to maintain the balance between supply and demand. Simply relying on the regulation capacity of traditional thermal power units will be difficult to meet the power gap during peak hours in the future. As of June 2024, China's installed capacity of new energy has reached 1.18 billion kilowatts. On the demand side, the scale of flexible loads such as electric vehicles (EV), air conditioners, and data centers has also continued to expand in recent years, and the regulation mode of the new power system is shifting from "source follows load" to "source-load interaction". At the same time, with the gradual advancement of the construction of a unified national power market, power grid operators can coordinate the flexible resources of multiple sides of the system through various market means such as electric energy, backup, and demand response, and ensure the stable operation of the system while minimizing operating costs. my country clearly proposed in the "Action Plan for Accelerating the Construction of a New Power System (2024-2027)": to achieve a demand-side response capacity of 5% or more of the maximum power load in typical areas. For power grid operators, how to fully tap the potential of demand response regulation and give play to the regulatory role of demand response mechanisms under limited operating costs is of great research significance.

[0003] EV resources generally have small adjustable capacity and scattered locations, making them difficult to be directly regulated by grid operators. Existing studies often use electric vehicle aggregators (EVAs) for unified management. Demand response (DR) can generally be divided into price-based demand response (PBDR) and incentive-based demand response (IBDR). PBDR generally refers to the use of energy prices such as time-of-use electricity prices, real-time electricity prices, and peak electricity prices by grid operators to influence the electricity consumption behavior of loads, thereby guiding loads to shift electricity consumption periods. IBDR generally refers to the use of incentive electricity prices such as incentive discounts and incentive electricity prices by grid operators to guide users to reduce or interrupt electricity loads. The demand response potential of EVA is limited under a single demand response mechanism, and the coordination of multiple demand response mechanisms is an effective way to stimulate the demand response potential of EVA. In terms of considering the coordinated optimization of multiple demand response mechanisms, existing technologies have only studied the optimal dispatch of power systems under different demand response mechanisms, but often analyze the impact of demand response on the power system under the premise of fixing the load scale participating in demand response. Under the circumstances of limited scale of adjustable resources and limited overall operating costs, the impact on the costs of grid operators was not analyzed from the perspective of demand response type and scale changes. Moreover, only the uncertainty of renewable energy output and load forecasting was considered, and there was a lack of modeling and consideration of the response uncertainty of users participating in demand response.

[0004] In view of the above problems, the present invention uses EVA as the modeling object of the demand-side adjustable load and constructs a two-layer demand response ratio optimization model considering dual uncertainties. The upper layer considers the uncertainty of source and load and proposes a method based on stochastic optimization to optimize the IBDR ratio and energy price and incentive price. The lower layer considers the uncertainty of response and proposes a data-driven demand response inference network (DRIN) to perform probabilistic modeling on the response characteristics of EVA participating in PBDR and IBDR. Summary of the invention

[0005] The present invention aims to solve the technical problems existing in the background technology, and its purpose is to provide an optimization method and system for electric vehicle demand response ratio considering dual uncertainty, and to construct a double-layer ratio optimization model that comprehensively considers source-load uncertainty and response uncertainty. The upper-level power grid operator optimizes the IBDR ratio and electricity prices under different scenarios, proposes a dynamic pricing mechanism for electricity prices, and introduces the demand response imbalance into the energy electricity price and incentive electricity price pricing process to reflect the impact of demand response conditions at different times on the pricing of power grid operators and EVA response strategies. The lower-level EVA optimizes the response strategy to provide a reference for power grid operation and planning. A data-driven demand response modeling method is proposed. The transferable load characteristics are modeled as physical information loss to guide model training, and the response strategy of EVA under PBDR and IBDR is inferred using probabilistic modeling, and the mean and variance of the EVA response strategy are output.

[0006] In order to solve the technical problem, the technical solution of the present invention is:

[0007] A method for optimizing electric vehicle demand response ratio considering dual uncertainty, the method comprising:

[0008] The target control curve and basic energy electricity price are obtained, and a two-layer ratio optimization model that comprehensively considers source-load uncertainty and response uncertainty is constructed, including an upper-layer demand response ratio optimization model and a lower-layer demand response strategy reasoning model based on the demand response inference network DRIN. After the maximum number of iterations is reached or the total cost change of the power grid operator is less than the threshold, the incentive-based demand response IBDR ratio and the corresponding electricity price operating cost are obtained.

[0009] Furthermore, the constructed two-layer matching optimization model comprehensively considers source-load uncertainty and response uncertainty. The upper layer is based on the power grid operator. In order to consider the source-load uncertainty, a random optimization method is used to construct a demand response matching optimization model, optimize the IBDR matching, energy electricity price, and incentive electricity price and pass them to the lower layer EVA, where the IBDR matching is defined as the proportion of EVA load participating in IBDR to the total EVA load; the lower layer is based on EVA, and uses the data-driven demand response inference network DRIN to construct an EVA response strategy for PBDR and IBDR, output the mean and variance of the EVA response strategy, and use the mean to represent the EVA demand response strategy considering the response uncertainty, which is returned to the upper layer power grid operator. The upper and lower layers are iterated repeatedly until the maximum number of iterations is reached or the total cost of the power grid operator changes less than a threshold, and finally the IBDR matching and the total cost of the power grid operator are output.

[0010] Further, the upper layer demand response ratio optimization model is constructed, specifically including:

[0011] The upper-level demand response ratio optimization model aims to minimize the total cost of the grid operator, including power generation cost, incentive cost, penalty for renewable energy abandonment, and penalty for demand response deviation; the decision variables are energy price, incentive price, and IBDR ratio;

[0012]

[0013] Where: F DR represents the upper optimization objective function; r represents the proportion of EVA load participating in IBDR in the total EVA load, that is, the IBDR ratio; S represents the number of typical scenarios; p s represents the probability of a typical scenario; T represents the total optimization period in a day; C DR represents the total EVA load scale; a, b, c represent the power generation cost coefficients of conventional units respectively; Indicates the maximum output of the new energy unit at time t under scenario s; Respectively represent the output of thermal power units and new energy units; represents the energy price; represents the incentive electricity price; pen It indicates the penalty price for curtailment of renewable energy and insufficient regulation; Indicates DR deviation;

[0014] The constraints of the upper-level demand response ratio optimization model are as follows:

[0015]

[0016] Where: Indicates the maximum output of conventional units; It represents the per unit value of EVA load participating in PBDR; It represents the per unit value of EVA load participating in IBDR; It represents the target control curve of the total EVA load, including both EVA participating in PBDR and EVA participating in IBDR; Indicates non-adjustable load.

[0017] Furthermore, considering the impact of demand response deviation at different times on electricity prices, a dynamic pricing mechanism for electricity prices is constructed. The marginal energy price of the power grid when there is no demand response is used as the basic energy price. The deviation between the actual load curve and the target control curve at different times is used as the basic energy price. Corrected energy price When there is a positive load deviation at time t, that is, the actual load is greater than the target load, the energy price is increased to guide the EVA participating in PBDR and the EVA participating in IBDR to reduce the load at time t; when there is a negative load deviation at time t, the opposite is true, and the incentive price is increased. The pricing mechanism takes the average clearing price of the spot market as:

[0018]

[0019] Where: E represents the energy price sensitivity coefficient; represents the basic energy electricity price; Indicates the maximum / minimum limit of energy price.

[0020] Furthermore, a demand response strategy reasoning model based on the demand response reasoning network DRIN is constructed, which specifically includes:

[0021] Using the data-driven demand response inference network DRIN, the response strategy of EVA participating in PBDR and IBDR is constructed; the attention mechanism and the time series convolution network TCN are introduced. The attention mechanism is used to analyze the importance and correlation between multiple input data, and TCN is used to analyze the correlation between time series features at different times; TCNs of different sizes are used to design a multi-time series feature extraction network to gradually extract the fluctuation characteristics of short, medium and long time series;

[0022] The input data set X and the real output data set Y are shown in the following formula. The data dimensions of the samples on the dth day are 24*12 and 24*1 respectively. The input data set includes DR type, DR instruction, energy price, incentive price, power energy upper and lower bounds, baseline power, and baseline energy. The real output data set includes deterministic demand response strategy.

[0023]

[0024] Y={P d} (35)

[0025] Where: Indicates the type of EVA participating in demand response on day d, 0 represents PBDR and 1 represents IBDR; It represents the IBDR regulation signal on day d, with a value between -1 and 1; represents the energy price; represents the incentive electricity price; P d,max / min Indicates the EVA power upper / lower limit per unit value; Indicates the per unit value of the reference load; Indicates the average load per unit value; E d,max / min Indicates the EVA energy upper / lower limit per unit value; Indicates the EVA reference energy per unit value; ΔE d P represents the per-unit value of PBDR energy transition caused by EV entering and leaving the grid; d It represents the per unit value of EVA load after demand response;

[0026] DRIN converts the transferability constraint into physical information loss. The smaller the physical information loss, the smaller the transferability error. This makes the PBDR strategy output by DRIN conform to the actual transferability load characteristics. The total network loss function L DRIN As shown in formula (12):

[0027]

[0028] Where: L MLE represents the maximum likelihood function loss; L PI Indicates physical information loss; μ d Represents the mean of the normal distribution that the response strategy follows; σ d represents the standard deviation of the corresponding normal distribution;

[0029] After N trainings, the last network weight is saved for inferring the EVA demand response strategy. DRIN Based on the difference in EVA energy boundary, it is determined whether EVA participates in PBDR or IBDR, and the corresponding response strategy is output. The EVA response strategy output by DRIN is further corrected based on the EVA energy boundary, and the final response strategy mean μd is returned to the upper-level grid operator.

[0030] An electric vehicle demand response ratio optimization system considering dual uncertainty, the system comprising:

[0031] Acquisition unit: used to obtain the target control curve and basic energy price;

[0032] Model building unit: used to build a two-layer ratio optimization model that comprehensively considers source load uncertainty and response uncertainty;

[0033] The model building unit comprises a first building subunit and a second building subunit;

[0034] The first construction subunit is used to construct an upper-layer demand response ratio optimization model, and the second construction subunit is used to construct a lower-layer demand response strategy reasoning model based on a demand response reasoning network DRIN;

[0035] Judgment unit: When the maximum number of iterations is reached or the total cost change of the power grid operator is less than the threshold, the incentive-based demand response IBDR ratio and the corresponding electricity price operating cost are obtained.

[0036] Furthermore, the first construction subunit specifically includes: constructing the upper layer demand response ratio optimization model:

[0037] The upper-level demand response ratio optimization model aims to minimize the total cost of the grid operator, including power generation cost, incentive cost, penalty for renewable energy abandonment, and penalty for demand response deviation; the decision variables are energy price, incentive price, and IBDR ratio;

[0038]

[0039] Where: F DR represents the upper optimization objective function; r represents the proportion of EVA load participating in IBDR in the total EVA load, that is, the IBDR ratio; S represents the number of typical scenarios; p s represents the probability of a typical scenario; T represents the total optimization period in a day; C DR represents the total EVA load scale; a, b, c represent the power generation cost coefficients of conventional units respectively; Indicates the maximum output of the new energy unit at time t under scenario s; Respectively represent the output of thermal power units and new energy units; represents the energy price; represents the incentive electricity price; pen It indicates the penalty price for curtailment of renewable energy and insufficient regulation; Indicates DR deviation;

[0040] The constraints of the upper-level demand response ratio optimization model are as follows:

[0041]

[0042]

[0043] Where: Indicates the maximum output of conventional units; It represents the per unit value of EVA load participating in PBDR; It represents the per unit value of EVA load participating in IBDR; It represents the target control curve of the total EVA load, including both EVA participating in PBDR and EVA participating in IBDR; Indicates non-adjustable load;

[0044] Considering the impact of demand response deviation at different times on electricity prices, a dynamic pricing mechanism for electricity prices is constructed. The marginal energy price of the power grid when there is no demand response is used as the basic energy price. The deviation between the actual load curve and the target control curve at different times is used as the basic energy price. Corrected energy price When there is a positive load deviation at time t, that is, the actual load is greater than the target load, the energy price is increased to guide the EVA participating in PBDR and the EVA participating in IBDR to reduce the load at time t; when there is a negative load deviation at time t, the opposite is true, and the incentive price is increased. The pricing mechanism takes the average clearing price of the spot market as:

[0045]

[0046] Where: E represents the energy price sensitivity coefficient; represents the basic energy electricity price; Indicates the maximum / minimum limit of energy price.

[0047] Furthermore, the first construction subunit specifically includes: constructing a lower-layer demand response strategy reasoning model based on a demand response reasoning network DRIN:

[0048] Using the data-driven demand response inference network DRIN, the response strategy of EVA participating in PBDR and IBDR is constructed; the attention mechanism and the time series convolution network TCN are introduced. The attention mechanism is used to analyze the importance and correlation between multiple input data, and TCN is used to analyze the correlation between time series features at different times; TCNs of different sizes are used to design a multi-time series feature extraction network to gradually extract the fluctuation characteristics of short, medium and long time series;

[0049] The input data set X and the real output data set Y are shown in the following formula. The data dimensions of the samples on the dth day are 24*12 and 24*1 respectively. The input data set includes DR type, DR instruction, energy price, incentive price, power energy upper and lower bounds, baseline power, and baseline energy. The real output data set includes deterministic demand response strategy.

[0050]

[0051] Y={P d} (47)

[0052] Where: Indicates the type of EVA participating in demand response on day d, 0 represents PBDR and 1 represents IBDR; It represents the IBDR regulation signal on day d, with a value between -1 and 1; represents the energy price; represents the incentive electricity price; P d,max / min Indicates the EVA power upper / lower limit per unit value; Indicates the per unit value of the reference load; Indicates the average load per unit value; E d,max / min Indicates the EVA energy upper / lower limit per unit value; Indicates the EVA reference energy per unit value; ΔE d P represents the per-unit value of PBDR energy transition caused by EV entering and leaving the grid; d It represents the per unit value of EVA load after demand response;

[0053] DRIN converts the transferability constraint into physical information loss. The smaller the physical information loss, the smaller the transferability error. This makes the PBDR strategy output by DRIN conform to the actual transferability load characteristics. The total network loss function L DRIN As shown in formula (12):

[0054]

[0055] Where: L MLE represents the maximum likelihood function loss; L PI Indicates physical information loss; μ d Represents the mean of the normal distribution that the response strategy follows; σ d represents the standard deviation of the corresponding normal distribution;

[0056] After N trainings, the last network weight is saved for inferring the EVA demand response strategy. DRIN Based on the difference in EVA energy boundary, it is determined whether EVA participates in PBDR or IBDR, and the corresponding response strategy is output. The EVA response strategy output by DRIN is further corrected based on the EVA energy boundary, and the final response strategy mean μd is returned to the upper-level grid operator.

[0057] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any one of the above-mentioned methods for optimizing the ratio of electric vehicle demand response considering dual uncertainty is implemented.

[0058] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any one of the above-mentioned methods for optimizing the ratio of electric vehicle demand response considering dual uncertainty.

[0059] Compared with the prior art, the advantages of the present invention are:

[0060] 1) The ratio optimization model used in the present invention can effectively characterize the optimal IBDR ratio for power grid operators when the total EVA load capacity is fixed (i.e., the proportion of EVA participating in IBDR to the total EVA scale), providing a reference for power grid planning and operation. 2) Considering the uncertainty of EVA response based on DRIN can improve the rationality of the optimization process and reduce the additional costs incurred by power grid operators due to failure to consider the uncertainty of EVA response. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1. Demand response reasoning network structure diagram; (a) multi-time series feature extraction network, (b) demand response reasoning network;

[0062] Figure 2 , demand response ratio optimization flow chart. DETAILED DESCRIPTION

[0063] The specific implementation mode of the present invention is described below in conjunction with embodiments:

[0064] It should be noted that the structures, proportions, sizes, etc. shown in this specification are only used to match the contents disclosed in the specification so that people familiar with this technology can understand and read them, and are not used to limit the conditions under which the present invention can be implemented. Any structural modification, change in proportional relationship or adjustment of size should still fall within the scope of the technical content disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0065] At the same time, the terms such as "upper", "lower", "left", "right", "middle" and "one" cited in this specification are only for the convenience of description and are not used to limit the scope of implementation of the present invention. Changes or adjustments to their relative relationships should be regarded as the scope of implementation of the present invention without substantially changing the technical content.

[0066] Definitions of Abbreviations and Key Terms

[0067] Electric vehicle (EV)

[0068] Electric vehicle aggregator (EVA)

[0069] Demand response (DR)

[0070] Price-based demand response (PBDR)

[0071] Incentive-based demand response (IBDR)

[0072] Demand response inference network (DRIN)

[0073] Temporal convolutional network (TCN)

[0074] Embodiment 1:

[0075] Technical solution of prior art 1:

[0076] Given the capacity of various units and the maximum load, the existing technology usually constructs a two-layer optimization model, fixing the adjustable load capacity participating in PBDR and IBDR. The upper layer optimizes the overall operating cost from the perspective of the grid operator, and the optimization variables are the unit output and the market clearing price. Even the market clearing price is not optimized, and only the time-of-use electricity price is used. From the perspective of EVA, the lower layer optimizes the demand response strategy of EVA based on the market clearing price or current time-of-use electricity price obtained by the upper layer, and feeds back to the upper layer. The upper and lower layers iterate and interact multiple times until the model converges.

[0077] Disadvantages of prior art 1:

[0078] The above technology analyzes the impact of demand response on the power system under the premise of a fixed load scale participating in demand response, and does not analyze the impact on the cost of grid operators from the perspective of the type and scale of adjustable loads participating in demand response. In addition, it only considers the uncertainty of new energy output and load forecasting, and lacks modeling and consideration of the response uncertainty of users participating in demand response.

[0079] Technical solution of the second prior art:

[0080] Traditional demand response modeling methods mainly include price elasticity coefficient, consumer psychological model, etc. The price elasticity coefficient reflects the impact of price changes at different times on the adjustable load at different times by constructing an elasticity coefficient matrix; the consumer psychological model describes the responsiveness of the adjustable load at different prices based on the principles of consumer psychology through nonlinear models such as logistic functions.

[0081] Optimization methods that consider response uncertainty include: random optimization optimizes user demand response behavior by introducing several typical scenarios and the corresponding probability of occurrence of the scenarios through objectives such as expected minimum; robust optimization gives the form of uncertainty sets through uncertainty intervals to optimize user demand response behavior under the worst conditions.

[0082] Disadvantages of the second prior art:

[0083] Traditional demand response modeling methods ignore the response uncertainty of users participating in demand response.

[0084] The setting of uncertainty parameters in uncertainty optimization methods such as stochastic optimization and robust optimization is relatively subjective and difficult to reflect the real user intentions.

[0085] In view of the shortcomings of technology 1, a two-layer matching optimization model is constructed that comprehensively considers source-load uncertainty and response uncertainty. The upper-layer grid operator optimizes the IBDR matching and electricity prices under different scenarios, proposes a dynamic pricing mechanism for electricity prices, and introduces the imbalance of demand response into the energy price and incentive price pricing process to reflect the impact of demand response conditions at different times on grid operator pricing and EVA response strategies. The lower-layer EVA optimizes the response strategy and provides a reference for grid operation and planning.

[0086] Aiming at the shortcomings of the second technique, a data-driven demand response modeling method is proposed. The transferable load characteristics are modeled as physical information loss to guide model training, and the response strategy of EVA under PBDR and IBDR is inferred using probabilistic modeling, and the mean and variance of the EVA response strategy are output.

[0087] Embodiment 2:

[0088] The present invention proposes an electric vehicle price-incentive demand response ratio optimization method that comprehensively considers source-load uncertainty and response uncertainty, and constructs a two-layer ratio optimization model that comprehensively considers source-load uncertainty and response uncertainty. The upper layer is based on the power grid operator. In order to consider the source-load uncertainty, a random optimization method is used to construct a demand response ratio optimization model, optimize the IBDR ratio and the energy price and incentive price and pass it to the lower layer EVA, where the IBDR ratio is defined as the proportion of the EVA load participating in the IBDR to the total EVA load. The lower layer is based on EVA, and a data-driven demand response inference network (DRIN) is proposed. For PBDR and IBDR, the EVA response strategy is inferred, the mean and variance of the EVA response strategy are output, and the mean is used to represent the EVA demand response strategy considering the response uncertainty, which is returned to the upper layer power grid operator. The upper and lower layers are iterated repeatedly until the maximum number of iterations is reached or the total cost of the power grid operator changes less than the threshold, and the IBDR ratio and the total cost of the power grid operator are finally output.

[0089] Upper-level demand response ratio optimization model:

[0090] The upper-level demand response ratio optimization model aims to minimize the total cost of the grid operator, including power generation cost, incentive cost, penalty for renewable energy curtailment, and penalty for demand response deviation. The decision variables are energy price, incentive price, and IBDR ratio.

[0091]

[0092] Where: F DR represents the upper optimization objective function; r represents the proportion of EVA load participating in IBDR in the total EVA load, that is, the IBDR ratio; S represents the number of typical scenarios; p s represents the probability of a typical scenario; T represents the total optimization period in a day; C DR represents the total EVA load scale; a, b, c represent the power generation cost coefficients of conventional units respectively; Indicates the maximum output of the new energy unit at time t under scenario s (scenario s and time t will not be repeated in the following text); Respectively represent the output of thermal power units and new energy units; represents the energy price; represents the incentive electricity price; pen It indicates the penalty price for curtailment of renewable energy and insufficient regulation; Indicates DR deviation.

[0093] The constraints are as follows:

[0094]

[0095] Where: Indicates the maximum output of conventional units; It represents the per unit value of EVA load participating in PBDR; It represents the per unit value of EVA load participating in IBDR; It represents the target control curve of the total EVA load (including both EVA participating in PBDR and EVA participating in IBDR); Indicates non-adjustable load.

[0096] In order to consider the impact of demand response deviation at different times on electricity prices, the present invention proposes a dynamic pricing mechanism for electricity prices, which takes the marginal energy price of the power grid when no demand response is performed as the basic energy price, and adjusts the price according to the deviation between the actual load curve and the target control curve at different times. Corrected energy price When there is a positive load deviation at time t, that is, the actual load is greater than the target regulated load, the energy price is increased to guide the EVA participating in PBDR and the EVA participating in IBDR to reduce the load at time t. When there is a negative load deviation at time t, the opposite is true.

[0097] Incentive electricity price The pricing mechanism adopts the average clearing electricity price in the spot market.

[0098]

[0099] Where: E represents the energy price sensitivity coefficient; represents the basic energy electricity price; Indicates the maximum / minimum limit of energy price.

[0100] The lower layer is the demand response strategy reasoning model based on the demand response reasoning network (DRIN):

[0101] In order to characterize the uncertainty of EVA response, this paper proposes DRIN to infer the response strategy of EVA participating in PBDR and IBDR. In order to accurately capture the complex nonlinear mapping relationship between input data and output data, this paper introduces the attention mechanism and temporal convolutional network (TCN) to improve the network's ability to analyze the characteristics of input data. The attention mechanism is used to analyze the importance and correlation between multiple input data, and TCN is used to analyze the correlation between temporal features at different times. This paper further uses TCNs of different sizes to design a multi-temporal feature extraction network to gradually extract the fluctuation characteristics of short, medium and long time series, as shown in Figure 1(a). The temporal deconvolution network with a symmetrical structure is designed to restore the different temporal features obtained to the same dimension, and output the probability distribution parameters of the demand response strategy (mean and standard deviation of the response strategy). The DRIN structure is shown in Figure 1(b).

[0102] The specific parameters of the multi-time series feature extraction network and the time series deconvolution network are shown in Table 1:

[0103] Table 1 - Multi-time series feature extraction network parameters

[0104]

[0105] Table 2 - Temporal Deconvolution Network Parameters

[0106]

[0107]

[0108] The input data set X and the real output data set Y are shown below. The data dimensions of the samples on day d are 24*12 and 24*1 respectively. The input data set includes DR type, DR instruction, energy price, incentive price, power and energy upper and lower bounds, baseline power, baseline energy, etc. The real output data set includes deterministic demand response strategy.

[0109]

[0110] Y={P d} (59)

[0111] Where: Indicates the type of EVA participating in demand response on day d, 0 represents PBDR and 1 represents IBDR; It represents the IBDR regulation signal on the dth day, and its value ranges from -1 to 1 (the dth day will not be mentioned in the following text); represents the energy price; represents the incentive electricity price; P d,max / min Indicates the EVA power upper / lower limit per unit value; Indicates the per unit value of the reference load; Indicates the average load per unit value; E d,max / min Indicates the EVA energy upper / lower limit per unit value; Indicates the EVA reference energy per unit value; ΔE d P represents the per-unit value of PBDR energy transition caused by EV entering and leaving the grid; d It represents the per unit value of EVA load after demand response.

[0112] DRIN transforms the transferability constraint into physical information loss. The smaller the physical information loss, the smaller the transferability error. This makes the PBDR strategy output by the proposed DRIN conform to the actual transferability load characteristics. The total network loss function L DRIN As shown in formula (12).

[0113]

[0114] Where: L MLE represents the maximum likelihood function loss; L PI Indicates physical information loss; μ d Represents the mean of the normal distribution that the response strategy follows; σ d represents the standard deviation of the corresponding normal distribution.

[0115] After N trainings, the last network weight is saved for inferring the EVA demand response strategy. , judge whether EVA participates in PBDR or IBDR, and output the corresponding response strategy. Further, based on the EVA energy boundary, the EVA response strategy output by DRIN is corrected to ensure that the response strategy will not cause EVA to cross the energy boundary. The final response strategy mean μ d Return to the upper grid operator. In summary, the optimization process of the double-layer demand response ratio optimization model considering double uncertainty proposed in the present invention is as follows: Figure 2It can be understood that the present invention constructs an electric vehicle price-incentive demand response ratio optimization model considering dual uncertainty, uses stochastic optimization to consider source-load uncertainty, uses DRIN to consider response uncertainty, optimizes the ratio of EVA participating in PBDR and IBDR, and reduces the total cost of power grid operators.

[0116] Embodiment 3:

[0117] The present invention provides an electric vehicle demand response ratio optimization system considering dual uncertainty, and the system can implement an electric vehicle demand response ratio optimization method considering dual uncertainty, specifically including:

[0118] Acquisition unit: used to obtain the target control curve and basic energy price;

[0119] Model building unit: used to build a two-layer ratio optimization model that comprehensively considers source load uncertainty and response uncertainty;

[0120] The model building unit comprises a first building subunit and a second building subunit;

[0121] The first construction subunit is used to construct an upper-layer demand response ratio optimization model, and the second construction subunit is used to construct a lower-layer demand response strategy reasoning model based on a demand response reasoning network DRIN;

[0122] Judgment unit: When the maximum number of iterations is reached or the total cost change of the power grid operator is less than the threshold, the incentive-based demand response IBDR ratio and the corresponding electricity price operating cost are obtained.

[0123] Embodiment 4:

[0124] This embodiment provides a terminal device, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for an operation of a method for optimizing the ratio of electric vehicle demand response considering dual uncertainty, including the following steps:

[0125] A two-layer matching optimization model is constructed that comprehensively considers source-load uncertainty and response uncertainty. The upper layer is based on the grid operator. In order to consider source-load uncertainty, a stochastic optimization method is used to construct a demand response matching optimization model, optimize the IBDR matching, energy price, and incentive price, and pass them to the lower layer EVA, where the IBDR matching is defined as the proportion of EVA load participating in IBDR to the total EVA load. The lower layer is based on EVA, and a data-driven demand response inference network (DRIN) is proposed. For PBDR and IBDR, the EVA response strategy is inferred, the mean and variance of the EVA response strategy are output, and the mean is used to represent the EVA demand response strategy considering response uncertainty, which is returned to the upper layer grid operator. The upper and lower layers are iterated repeatedly until the maximum number of iterations is reached or the total cost of the grid operator changes less than the threshold, and finally the IBDR matching and the total cost of the grid operator are output.

[0126] Embodiment 5:

[0127] This embodiment provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It is understandable that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory.

[0128] The processor may load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the above-mentioned embodiment regarding a method for optimizing the ratio of electric vehicle demand response considering dual uncertainty; the processor may load and execute the following steps of one or more instructions in the computer-readable storage medium:

[0129] A two-layer matching optimization model is constructed that comprehensively considers source-load uncertainty and response uncertainty. The upper layer is based on the grid operator. In order to consider source-load uncertainty, a stochastic optimization method is used to construct a demand response matching optimization model, optimize the IBDR matching, energy price, and incentive price, and pass them to the lower layer EVA, where the IBDR matching is defined as the proportion of EVA load participating in IBDR to the total EVA load. The lower layer is based on EVA, and a data-driven demand response inference network (DRIN) is proposed. For PBDR and IBDR, the EVA response strategy is inferred, the mean and variance of the EVA response strategy are output, and the mean is used to represent the EVA demand response strategy considering response uncertainty, which is returned to the upper layer grid operator. The upper and lower layers are iterated repeatedly until the maximum number of iterations is reached or the total cost of the grid operator changes less than the threshold, and finally the IBDR matching and the total cost of the grid operator are output.

[0130] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0131] The present invention is described with reference to the flowchart and / or block diagram of the method, device (system), and computer program product according to the embodiment of the present invention. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the process and / or box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0132] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0133] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0134] The preferred embodiments of the present invention are described in detail above, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

[0135] Many other changes and modifications may be made without departing from the concept and scope of the present invention.It should be understood that the present invention is not limited to the specific embodiments, and the scope of the present invention is defined by the appended claims.

Claims

1. A method for optimizing the ratio of electric vehicle demand response considering dual uncertainty, characterized in that: The method comprises: The target control curve and basic energy electricity price are obtained, and a two-layer ratio optimization model that comprehensively considers source-load uncertainty and response uncertainty is constructed, including an upper-layer demand response ratio optimization model and a lower-layer demand response strategy reasoning model based on the demand response inference network DRIN. After the maximum number of iterations is reached or the total cost change of the power grid operator is less than the threshold, the incentive-based demand response IBDR ratio and the corresponding electricity price operating cost are obtained.

2. The method for optimizing the ratio of electric vehicle demand response considering dual uncertainty according to claim 1 is characterized in that: The constructed two-layer matching optimization model comprehensively considers source-load uncertainty and response uncertainty. The upper layer is based on the power grid operator. In order to consider the source-load uncertainty, a random optimization method is used to construct a demand response matching optimization model, optimize the IBDR matching, energy electricity price, and incentive electricity price and pass them to the lower layer EVA, where the IBDR matching is defined as the proportion of EVA load participating in IBDR to the total EVA load; the lower layer is based on EVA, and uses the data-driven demand response inference network DRIN to construct an EVA response strategy for PBDR and IBDR, output the mean and variance of the EVA response strategy, and use the mean to represent the EVA demand response strategy considering the response uncertainty, which is returned to the upper layer power grid operator. The upper and lower layers are iterated repeatedly until the maximum number of iterations is reached or the total cost of the power grid operator changes less than a threshold, and finally the IBDR matching and the total cost of the power grid operator are output.

3. The method for optimizing the ratio of electric vehicle demand response considering dual uncertainty according to claim 1 is characterized in that: Constructing the upper-layer demand response ratio optimization model specifically includes: The upper-level demand response ratio optimization model aims to minimize the total cost of the grid operator, including power generation cost, incentive cost, penalty for renewable energy abandonment, and penalty for demand response deviation; the decision variables are energy price, incentive price, and IBDR ratio; Where: F DR represents the upper optimization objective function; r represents the proportion of EVA load participating in IBDR in the total EVA load, that is, the IBDR ratio; S represents the number of typical scenarios; p s represents the probability of a typical scenario; T represents the total optimization period in a day; C DR represents the total EVA load scale; a, b, c represent the power generation cost coefficients of conventional units respectively; Indicates the maximum output of the new energy unit at time t under scenario s; Respectively represent the output of thermal power units and new energy units; represents the energy price; represents the incentive electricity price; pen It indicates the penalty price for curtailment of renewable energy and insufficient regulation; Indicates DR deviation; The constraints of the upper-level demand response ratio optimization model are as follows: Where: Indicates the maximum output of conventional units; It represents the per unit value of EVA load participating in PBDR; It represents the per unit value of EVA load participating in IBDR; It represents the target control curve of the total EVA load, including both EVA participating in PBDR and EVA participating in IBDR; Indicates non-adjustable load.

4. The method for optimizing the ratio of electric vehicle demand response considering dual uncertainty according to claim 3 is characterized in that: Considering the impact of demand response deviation at different times on electricity prices, a dynamic pricing mechanism for electricity prices is constructed. The marginal energy price of the power grid when there is no demand response is used as the basic energy price. The deviation between the actual load curve and the target control curve at different times is used as the basic energy price. Corrected energy price When there is a positive load deviation at time t, that is, the actual load is greater than the target load, the energy price is increased to guide the EVA participating in PBDR and the EVA participating in IBDR to reduce the load at time t; when there is a negative load deviation at time t, the opposite is true, and the incentive price is increased. The pricing mechanism takes the average clearing price of the spot market as: Where: E represents the energy price sensitivity coefficient; represents the basic energy electricity price; Indicates the maximum / minimum limit of energy price.

5. The method for optimizing the ratio of electric vehicle demand response considering dual uncertainty according to claim 1 is characterized in that: Construct the demand response strategy reasoning model based on the demand response reasoning network DRIN, which includes: Using the data-driven demand response inference network DRIN, the response strategy of EVA participating in PBDR and IBDR is constructed; the attention mechanism and the time series convolution network TCN are introduced. The attention mechanism is used to analyze the importance and correlation between multiple input data, and TCN is used to analyze the correlation between time series features at different times; TCNs of different sizes are used to design a multi-time series feature extraction network to gradually extract the fluctuation characteristics of short, medium and long time series; The input data set X and the real output data set Y are shown in the following formula. The data dimensions of the samples on the dth day are 24*12 and 24*1 respectively. The input data set includes DR type, DR instruction, energy price, incentive price, power energy upper and lower bounds, baseline power, and baseline energy. The real output data set includes deterministic demand response strategy. Where: Indicates the type of demand response that EVA participates in on day d, 0 represents PBDR and 1 represents IBDR; It represents the IBDR regulation signal on day d, with a value between -1 and 1; represents the energy price; represents the incentive electricity price; P d,max / min Indicates the EVA power upper / lower limit per unit value; Indicates the per unit value of the reference load; Indicates the average load per unit value; E d,max / min Indicates the EVA energy upper / lower limit per unit value; Indicates the EVA reference energy per unit value; ΔE d P represents the per-unit value of PBDR energy transition caused by EV entering and leaving the grid; d It represents the per unit value of EVA load after demand response; DRIN converts the transferability constraint into physical information loss. The smaller the physical information loss, the smaller the transferability error. This makes the PBDR strategy output by DRIN conform to the actual transferability load characteristics. The total network loss function L DRIN As shown in formula (12): Where: L MLE represents the maximum likelihood function loss; L PI Indicates physical information loss; μ d Represents the mean of the normal distribution that the response strategy follows; σ d represents the standard deviation of the corresponding normal distribution; After N trainings, the last network weight is saved for inferring the EVA demand response strategy. DRIN , judge whether EVA participates in PBDR or IBDR, and output the corresponding response strategy. Further, based on the EVA energy boundary, the EVA response strategy output by DRIN is corrected, and the final response strategy mean μ d Return to upper grid operator.

6. An electric vehicle demand response ratio optimization system considering dual uncertainty, characterized in that: The system comprises: Acquisition unit: used to obtain the target control curve and basic energy price; Model building unit: used to build a two-layer ratio optimization model that comprehensively considers source load uncertainty and response uncertainty; The model building unit comprises a first building subunit and a second building subunit; The first construction subunit is used to construct an upper-layer demand response ratio optimization model, and the second construction subunit is used to construct a lower-layer demand response strategy reasoning model based on a demand response reasoning network DRIN; Judgment unit: When the maximum number of iterations is reached or the total cost change of the power grid operator is less than the threshold, the incentive-based demand response IBDR ratio and the corresponding electricity price operating cost are obtained.

7. The electric vehicle demand response ratio optimization system considering dual uncertainty according to claim 6 is characterized in that: The first construction subunit specifically includes: constructing the upper layer demand response ratio optimization model: The upper-level demand response ratio optimization model aims to minimize the total cost of the grid operator, including power generation cost, incentive cost, penalty for renewable energy abandonment, and penalty for demand response deviation; the decision variables are energy price, incentive price, and IBDR ratio; Where: F DR represents the upper optimization objective function; r represents the proportion of EVA load participating in IBDR in the total EVA load, that is, the IBDR ratio; S represents the number of typical scenarios; p s represents the probability of a typical scenario; T represents the total optimization period in a day; C DR represents the total EVA load scale; a, b, c represent the power generation cost coefficients of conventional units respectively; Indicates the maximum output of the new energy unit at time t under scenario s; Respectively represent the output of thermal power units and new energy units; represents the energy price; represents the incentive electricity price; pen It indicates the penalty price for curtailment of renewable energy and insufficient regulation; Indicates DR deviation; The constraints of the upper-level demand response ratio optimization model are as follows: Where: Indicates the maximum output of conventional units; It represents the per unit value of EVA load participating in PBDR; It represents the per unit value of EVA load participating in IBDR; It represents the target control curve of the total EVA load, including both EVA participating in PBDR and EVA participating in IBDR; Indicates non-adjustable load; Considering the impact of demand response deviation at different times on electricity prices, a dynamic pricing mechanism for electricity prices is constructed. The marginal energy price of the power grid when there is no demand response is used as the basic energy price. The deviation between the actual load curve and the target control curve at different times is used as the basic energy price. Corrected energy price When there is a positive load deviation at time t, that is, the actual load is greater than the target load, the energy price is increased to guide the EVA participating in PBDR and the EVA participating in IBDR to reduce the load at time t; when there is a negative load deviation at time t, the opposite is true, and the incentive price is increased. The pricing mechanism takes the average clearing price of the spot market as: Where: E represents the energy price sensitivity coefficient; represents the basic energy electricity price; Indicates the maximum / minimum limit of energy price.

8. The electric vehicle demand response ratio optimization system considering dual uncertainty according to claim 6 is characterized in that: The first construction subunit specifically includes: constructing a lower-layer demand response strategy reasoning model based on a demand response reasoning network DRIN: Using the data-driven demand response inference network DRIN, the response strategy of EVA participating in PBDR and IBDR is constructed; the attention mechanism and the time series convolution network TCN are introduced. The attention mechanism is used to analyze the importance and correlation between multiple input data, and TCN is used to analyze the correlation between time series features at different times; TCNs of different sizes are used to design a multi-time series feature extraction network to gradually extract the fluctuation characteristics of short, medium and long time series; The input data set X and the real output data set Y are shown in the following formula. The data dimensions of the samples on the dth day are 24*12 and 24*1 respectively. The input data set includes DR type, DR instruction, energy price, incentive price, power energy upper and lower bounds, baseline power, and baseline energy. The real output data set includes deterministic demand response strategy. And={P d } (23) Where: Indicates the type of demand response that EVA participates in on day d, 0 represents PBDR and 1 represents IBDR; It represents the IBDR regulation signal on day d, with a value between -1 and 1; represents the energy price; represents the incentive electricity price; P d,max / min Indicates the EVA power upper / lower limit per unit value; Indicates the per unit value of the reference load; Indicates the average load per unit value; E d,max / min Indicates the EVA energy upper / lower limit per unit value; Indicates the EVA reference energy per unit value; ΔE d P represents the per-unit value of PBDR energy transition caused by EV entering and leaving the grid; d It represents the per unit value of EVA load after demand response; DRIN converts the transferability constraint into physical information loss. The smaller the physical information loss, the smaller the transferability error. This makes the PBDR strategy output by DRIN conform to the actual transferability load characteristics. The total network loss function L DRIN As shown in formula (12): Where: L MLE represents the maximum likelihood function loss; L PI Indicates physical information loss; μ d Represents the mean of the normal distribution that the response strategy follows; σ d represents the standard deviation of the corresponding normal distribution; After N trainings, the last network weight is saved for inferring the EVA demand response strategy. DRIN , judge whether EVA participates in PBDR or IBDR, and output the corresponding response strategy. Further, based on the EVA energy boundary, the EVA response strategy output by DRIN is corrected, and the final response strategy mean μ d Return to upper grid operator.

9. A computer device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, an electric vehicle demand response ratio optimization method considering dual uncertainty as described in any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements a method for optimizing electric vehicle demand response ratio considering dual uncertainty as described in any one of claims 1 to 5.

Images