Retail package electricity price decision-making method, model training method and related device

Abstract

The invention provides a training method and device and an application method and device of a retail package electricity price decision model, and relates to the technical field of retail package electricity price decision in an electricity market, and the method comprises the steps: obtaining data for retail package electricity price decision in a preset historical time period; the retail package electricity price decision data comprises time period type division data, electricity selling subject wholesale side data, retail user characteristic data and signed retail package agreed electric quantity data between an electricity selling subject and each retail user; based on an electricity selling subject income model and a retail user electricity utilization utility model, taking an electricity selling subject as an intelligent agent, and constructing a Markov decision process; and training the Markov decision process by using a near-end strategy optimization algorithm based on the data for the retail package electricity price decision to obtain a retail package electricity price decision model. According to the method, the benefits of the electricity selling main body and the electricity utilization utility of the retail users can be taken into consideration, and flexible decision support is provided for formulating the electricity price of the combined retail package in an uncertain environment.

Description

Technical Field

[0001] The present invention relates to the technical field of retail package electricity price decision-making in the power market, and in particular to a training method, an application method and a device for a retail package electricity price decision-making model. Background Art

[0002] In the electricity retail market, building a reasonable retail package pricing mechanism can ensure the fair, just and orderly operation of the electricity retail market. The existing retail package pricing mechanism is a fixed retail package pricing mechanism, which often only considers the interests of the electricity seller, formulates the retail package with the goal of maximizing the profit of the electricity seller, and the deployed pricing strategy is predetermined by the abstract model.

[0003] However, only considering the interests of the electricity sellers while ignoring the electricity utility and satisfaction of retail users makes it difficult for retail package price signals to motivate users to adjust their electricity consumption behavior, resulting in problems such as waste of adjustable regulation resources and excessive electricity costs for users; and, since wholesale electricity prices show variable and complex trends at different stages of the market, and the dynamic correlation between wholesale electricity prices and retail electricity prices shows highly nonlinear and uncertain characteristics, traditional linear or deterministic modeling methods are difficult to fully capture the complex interactions between the two and cannot fully reflect the actual dynamics of the market, thus adopting a fixed retail package pricing mechanism, which has the problem of low flexibility in retail electricity price decision-making; and, the existing retail package pricing mechanism mostly customizes different retail package prices according to the electricity consumption and needs of different types of retail users. When multiple types of retail users coexist, the existing scheme is difficult to maximize the interests of multiple parties.

[0004] Based on this, how to balance the interests of electricity sales entities and the electricity efficiency of retail users, and at the same time provide flexible decision-making support for the formulation of joint retail package electricity prices (i.e., retail package electricity prices for multiple types of retail users) under an uncertain environment, has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] In view of this, in order to solve the above technical problems, the present invention provides a retail package electricity price decision method, a model training method and related devices.

[0006] The present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a training method for a retail package electricity price decision model, comprising:

[0008] Obtaining data for retail package electricity price decision-making within a preset historical time period; the data for retail package electricity price decision-making includes time period type division data, wholesale side data of the electricity sales entity, retail user's own characteristic data and retail package agreed electricity volume data signed between the electricity sales entity and each retail user;

[0009] Based on the revenue model of the electricity sales entity and the electricity utility model of the retail users, a Markov decision process is constructed with the electricity sales entity as an intelligent agent;

[0010] Based on the retail package electricity price decision-making data, the Markov decision process is trained using a proximal strategy optimization algorithm to obtain a retail package electricity price decision-making model.

[0011] Optionally, the power selling entity's profit model takes into account the power selling entity's medium- and long-term market power purchasing costs, spot market power purchasing costs, and retail market power selling revenues.

[0012] Optionally, the process of constructing the revenue model of the electricity sales entity includes:

[0013] A medium- and long-term market electricity purchase cost model for the electricity sales entity is established; the formula of the medium- and long-term market electricity purchase cost model is as follows:

[0014]

[0015] Among them, C z T is the medium- and long-term market electricity purchase cost of the electricity sales entity; f , T p and T g They are peak time, normal time and valley time respectively; P f , P p and P g are the monthly average transaction prices of the peak period, normal period and valley period of the medium- and long-term wholesale market of the electricity sales entity; I is the set of contracted retail users of the electricity sales entity; q i,t The retail package agreed electricity consumption of the i-th retail user in the t-th preset unit time period;

[0016] A spot market electricity purchase cost model for the electricity sales entity is established; the formula of the spot market electricity purchase cost model is as follows:

[0017]

[0018] Among them, C x is the spot market electricity purchase cost of the electricity seller; T is the number of preset unit time periods included in a day; is the day-ahead market-cleared electricity quantity of the electricity sales entity in the tth preset unit time period; r,tThe day-ahead settlement price of the electricity sales entity in the t-th preset unit time period; is the optimal power consumption of the ith retail user in the tth preset unit time period; P s,t is the real-time settlement price of the electricity seller in the tth preset unit time period;

[0019] A retail market electricity sales revenue model for the electricity sales entity is established; the formula of the retail market electricity sales revenue model is as follows:

[0020]

[0021] Among them, R sale The retail market electricity sales revenue of the electricity sales entity; i,t The electricity retail package price for the i-th retail user in the t-th preset unit time period is determined;

[0022] The following formula is used to construct the revenue model C of the electricity sales entity: total :

[0023] C total =R sale -C z -C x .

[0024] Optionally, the retail user electricity utility model takes into account the retail package electricity price and the retail user's own characteristics.

[0025] Optionally, the process of constructing the retail user electricity utility model includes:

[0026] Establish an optimal electricity consumption behavior model for the retail user; the formula of the optimal electricity consumption behavior model is as follows:

[0027]

[0028] in, The optimal electricity consumption of the retail user under the current retail package electricity price; is the rigid load of the ith retail user in the tth preset unit time period; is the adjustable load of the ith retail user in the tth preset unit time period; α i is the price elasticity parameter of the ith retail user; β i is the preference level parameter of the ith retail user for electricity consumption;

[0029] Based on the optimal electricity consumption behavior model, the retail user electricity consumption utility model is established. The formula of the retail user electricity consumption utility model is as follows:

[0030]

[0031] Among them, U i,t is the electricity consumption utility of the ith retail user in the tth preset unit time period; ω i is the diminishing marginal utility parameter brought by constant electricity to the i-th retail user.

[0032] Optionally, based on the revenue model of the electricity sales entity and the electricity utility model of the retail users, the electricity sales entity is used as an intelligent agent to construct a Markov decision process, which specifically includes:

[0033] The state space S of the agent is designed as:

[0034] S={s t |t=1,2,...,T}

[0035]

[0036]

[0037] Among them, T * The time period type used to identify the current time period, wherein the time period types include peak time period, normal time period and valley time period;

[0038] Design the action space A of the agent as:

[0039] A={a t |t=1,2,...,T}

[0040] a t ={p i∈I,t}

[0041] The reward function r of the agent is designed as:

[0042]

[0043] Among them, η is the proportional magnification factor of the electricity utility value.

[0044] In a second aspect, the present invention provides a method for deciding a retail package electricity price, comprising:

[0045] Obtain data for retail package electricity price decision-making within the decision-making day;

[0046] Based on the retail package electricity price decision-making data, a preset retail package electricity price decision model is used to generate retail package electricity price data for retail users within each preset unit time period of the decision-making day; the retail package electricity price data for retail users include the electricity retail package price corresponding to each retail user of the electricity sales entity within the current preset unit time period; the retail package electricity price decision model is trained using the training method of the retail package electricity price decision model as described above.

[0047] In a third aspect, the present invention provides a training device for a retail package electricity price decision model, comprising:

[0048] A training data acquisition module is used to acquire data for retail package electricity price decision-making within a preset historical time period; the data for retail package electricity price decision-making includes time period type division data, wholesale side data of the electricity sales entity, retail user's own characteristic data and retail package agreed electricity volume data signed between the electricity sales entity and each retail user;

[0049] A construction module is used to construct a Markov decision process based on the revenue model of the electricity sales entity and the electricity utility model of the retail user, taking the electricity sales entity as an intelligent agent;

[0050] The training module is used to train the Markov decision process based on the retail package electricity price decision-making data using a proximal strategy optimization algorithm to obtain a retail package electricity price decision-making model.

[0051] In a fourth aspect, the present invention provides a retail package electricity price decision device, comprising:

[0052] A decision data acquisition module, used to acquire data for retail package electricity price decision-making within the decision-making day;

[0053] A decision module is used to generate retail package electricity price data for retail users within each preset unit time period of the day to be decided based on the retail package electricity price decision-making data and using a preset retail package electricity price decision model; the retail package electricity price data for retail users includes the electricity retail package price corresponding to each retail user of the electricity selling entity within the current preset unit time period; the retail package electricity price decision model is trained using the training method of the retail package electricity price decision model as described above.

[0054] The present invention adopts the above technical scheme, a training method for a retail package electricity price decision model, including: obtaining data for retail package electricity price decision within a preset historical time period; the data for retail package electricity price decision includes time period type division data, wholesale side data of the electricity sales entity, retail user's own characteristic data and contracted retail package agreed electricity volume data between the electricity sales entity and each retail user; based on the electricity sales entity revenue model and the retail user electricity utility model, a Markov decision process is constructed with the electricity sales entity as an intelligent body; based on the retail package electricity price decision data, a proximal strategy optimization algorithm is used to train the Markov decision process to obtain a retail package electricity price decision model. Based on this, since a Markov decision process is constructed based on the revenue model of the electricity sales entity and the electricity utility model of the retail users, the present invention can take into account the interests of the electricity sales entity and the electricity utility of the retail users. In addition, since a Markov decision process is constructed with the electricity sales entity as an intelligent body, a proximal strategy optimization algorithm is used to train the Markov decision process based on the data for retail package electricity price decision-making, and a retail package electricity price decision-making model is obtained. This enables the retail package electricity price decision-making model to provide flexible decision support for the formulation of joint retail package electricity prices under an uncertain environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 It is a flowchart of a training method for a retail package electricity price decision model provided by an embodiment of the present invention;

[0057] Figure 2 It is a flowchart of a method for deciding a retail package electricity price provided by an embodiment of the present invention;

[0058] Figure 3 It is a structural schematic diagram of a training device for a retail package electricity price decision model provided by an embodiment of the present invention;

[0059] Figure 4 It is a structural schematic diagram of a retail package electricity price decision-making device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0061] Example

[0062] Figure 1 FIG. 1 is a flow chart of a method for training a retail package electricity price decision model provided by an embodiment of the present invention. Figure 1 As shown, the training method of the retail package electricity price decision model includes:

[0063] Step 101: Obtain data for retail package electricity price decision-making within a preset historical time period; the data for retail package electricity price decision-making includes time period type classification data, wholesale side data of power sales entities, retail users' own characteristic data, and retail package agreed electricity volume data signed between the power sales entity and each retail user.

[0064] Specifically, the time period type division data belongs to the market operation parameters, which specifically include the peak period T of the power market in the preset historical time period. f , valley period T g PeacetimeT p The preset historical time period can be several consecutive months.

[0065] The wholesale side data of the electricity sales entity may specifically include: the monthly average transaction price P during the medium and long-term wholesale market peak period of the electricity sales entity; f , the average monthly transaction price of the electricity sales entity in the medium and long-term wholesale market during normal periods P p The average monthly transaction price of the medium- and long-term wholesale market valley period of the electricity sales entity P g , the day-ahead market-cleared electricity quantity dataset of the power sales entity within the preset historical time period, the day-ahead settlement price dataset of the power sales entity within the preset historical time period, and the real-time settlement price dataset of the power sales entity within the preset historical time period.

[0066] The retail user's own characteristic data may specifically include: a rigid load data set and an adjustable load data set of each retail user within a preset historical time period, a price elasticity parameter, a preference level parameter for electricity consumption, and a marginal diminishing utility parameter.

[0067] Step 102: Based on the revenue model of the electricity sales entity and the electricity utility model of the retail users, a Markov decision process is constructed with the electricity sales entity as the intelligent agent.

[0068] In the embodiments of this specification, the revenue model of the electricity seller takes into account the medium- and long-term market electricity purchase costs, spot market electricity purchase costs and retail market electricity sales revenue of the electricity seller.

[0069] In the embodiments of this specification, the process of constructing the revenue model of the electricity sales entity includes:

[0070] (1) Establish a medium- and long-term market electricity purchase cost model for electricity sellers; the formula for the medium- and long-term market electricity purchase cost model is as follows:

[0071]

[0072] Among them, C z The medium- and long-term market electricity purchase cost of the electricity sales entity; T f , T p and T g They are peak time, normal time and valley time respectively; P f , P p and P g are the monthly average transaction prices of the peak, normal and valley periods of the medium- and long-term wholesale market of the electricity sales entity; I is the set of retail users contracted by the electricity sales entity; q i,t The agreed electricity consumption of the retail package of the ith retail user in the tth preset unit time period.

[0073] (2) Establish a spot market electricity purchase cost model for electricity sellers; the formula of the spot market electricity purchase cost model is as follows:

[0074]

[0075] Among them, C x is the electricity purchase cost of the electricity seller in the spot market; T is the number of preset unit time periods included in a day. When the length of the preset unit time period is 1 hour, the value of T is 24; P is the day-ahead market clearing electricity quantity of the electricity sales entity in the tth preset unit time period; r,t The day-ahead settlement price of the electricity sales entity in the t-th preset unit time period; is the optimal power consumption of the ith retail user in the tth preset unit time period; P s,t It is the real-time settlement price of the electricity seller in the tth preset unit time period.

[0076] In the embodiment of this specification, the process of establishing the spot market electricity purchase cost model includes:

[0077] (2.1) Based on the deviation between the day-ahead market clearing electricity volume and the retail contract electricity volume of the electricity sales entity and the day-ahead market settlement electricity price of the electricity sales entity, a day-ahead market purchase cost model of the electricity sales entity is established. The day-ahead market purchase cost model of the electricity sales entity is C rThe formula is as follows:

[0078]

[0079] (2.2) Based on the deviation between the actual electricity consumption of retail users and the market clearing electricity of the electricity sellers on the previous day, and the real-time market settlement electricity price of the electricity sellers, a real-time market electricity purchase cost model for the electricity sellers is established. The real-time market electricity purchase cost C s The model formula is:

[0080]

[0081] (2.2) Adding formula (3) and (4) together, we get the spot market electricity purchase cost model:

[0082]

[0083] (3) Establish a retail market electricity sales revenue model for electricity sales entities; the formula for the retail market electricity sales revenue model is as follows:

[0084]

[0085] Among them, R sale The retail market electricity sales revenue of the electricity sales entity; i,t The electricity retail package price for the i-th retail user in the t-th preset unit time period is determined.

[0086] (4) Use the following formula to construct the revenue model C of the electricity sales entity. total :

[0087]

[0088] In the embodiment of the present invention, the retail user electricity utility model takes into account the retail package electricity price and the retail user's own characteristics.

[0089] In the embodiment of the present invention, the process of constructing the retail user electricity utility model includes:

[0090] (1) Establish an optimal electricity consumption behavior model for retail users; the formula of the optimal electricity consumption behavior model is as follows:

[0091]

[0092] in, The optimal electricity consumption for retail users under the current retail package electricity price; is the rigid load of the ith retail user in the tth preset unit time period; is the adjustable load of the ith retail user in the tth preset unit time period; α iis the price elasticity parameter of the ith retail user; β i is the preference level parameter of the ith retail user for electricity consumption.

[0093] (2) Based on the optimal electricity consumption behavior model, a retail user electricity consumption utility model is established. The formula of the retail user electricity consumption utility model is as follows:

[0094]

[0095] Among them, U i,t is the electricity consumption utility of the ith retail user in the tth preset unit time period; ω i is the diminishing marginal utility parameter brought by constant electricity to the i-th retail user.

[0096] Based on this, since the present invention takes into account the economic law of diminishing marginal utility, the established retail user electricity utility model can be used to evaluate the retail package electricity price generated by the decision-making method, the retail user's electricity satisfaction and sensitivity to price changes after adjusting the user's electricity behavior.

[0097] In the embodiment of the present invention, based on the revenue model of the electricity sales entity and the electricity utility model of the retail user, the electricity sales entity is taken as an intelligent agent to construct a Markov decision process, which may specifically include:

[0098] (1) Considering that the settlement price and cleared electricity quantity of the electricity seller on the wholesale side, the characteristics of the retail users themselves, and the agreed electricity quantity of the retail package have an observable influence on the formulation of the retail electricity price, the state space S of the intelligent agent is designed as:

[0099] S={s t |t=1,2,...,T}......(9)

[0100]

[0101] Among them, T * Used to identify the time period type of the current time period. Time period types include peak time period, normal time period, and valley time period.

[0102] (2) The retail package electricity price formulation between the electricity sales entity and each retail user at different times is designed as the action of the intelligent agent, and the action space A of the intelligent agent is designed as:

[0103] A={a t |t=1,2,...,T}......(10)

[0104] a t ={p i∈I,t}

[0105] (3) Taking into account the interests of the electricity sales entity and the utility of users, based on the aforementioned electricity sales entity revenue model and retail user electricity utility model, the reward function r of the intelligent agent is designed as:

[0106]

[0107]

[0108] Among them, η is the proportional magnification factor of the electricity utility value.

[0109] Step 103: Based on the retail package electricity price decision-making data, a proximal policy optimization (PPO) algorithm is used to train the Markov decision process to obtain a retail package electricity price decision-making model.

[0110] Specifically, step 103: based on the retail package electricity price decision-making data, using the proximal strategy optimization algorithm, training the Markov decision process to obtain the retail package electricity price decision-making model may specifically include:

[0111] (3.1) The retail package electricity price decision-making data within the preset historical time period obtained in step 101 is input into the PPO algorithm, and the data required for iterative update of the PPO algorithm is generated, which specifically includes the following steps:

[0112] (3.1.1) Randomly select one day of data from the retail package electricity price decision-making data within the preset historical time period {s t |t=1,2,...,T}, and replace s at t=1 t as the initial state input.

[0113] (3.1.2) t As the input of the PPO algorithm actor network, the actor network perceives the state input data and outputs action a t .

[0114] (3.1.3) Based on the generated action value a t The optimal electricity consumption behavior model of retail users established above is used to calculate the optimal electricity consumption of each retail user.

[0115] (3.1.4) Use the above formula (11) to calculate the reward value r of this step t .

[0116] (3.1.5) The currently obtained {s t ,a t ,r t ,s t+1} is stored in the memory pool D as the experience required for iterative training of the algorithm.

[0117] (3.1.6) Execute t=t+1, and repeat steps (3.1.2) to (3.1.5) until t=25.

[0118] (3.2) According to the data generated in step (3.1), the PPO algorithm is iteratively trained, including the following steps:

[0119] (3.2.1) Sample a batch of experience B from the experience pool D and calculate the loss function L(φ) of the critic network. The calculation formula is:

[0120]

[0121] Where γ is the reward discount factor; V θ (s t ) is state s t The value estimate under ; N is the number of trajectories taken from the experience.

[0122] (3.2.2) Based on the loss function value calculated in the previous step (3.2.1), back propagation is performed to update the critic network parameter φ by gradient descent to achieve the update iteration of the critic network. The update formula is:

[0123]

[0124] Among them, α is the learning rate.

[0125] (3.2.3) Calculate the advantage function A of the PPO algorithm π (s t ,a t ), the calculation formula is:

[0126]

[0127] (3.2.4) Calculate the loss function of the actor network The calculation formula is:

[0128]

[0129] Among them, π θ (a t |s t ) is based on the state s t Generate action a t The probability of ; ε is a hyperparameter.

[0130] (3.2.5) Based on the loss function value calculated in step (3.2.4), the actor network parameter θ is updated by gradient ascent to realize the update iteration of the actor network. The update formula is:

[0131]

[0132] (3.3) Repeat steps (3.1) and (3.2) to iteratively train the PPO algorithm until the reward value converges, thereby obtaining a retail package electricity price decision model.

[0133] The present invention adopts the above technical scheme. Since a Markov decision process is constructed based on the revenue model of the electricity sales entity and the electricity utility model of the retail users, the present invention can take into account the interests of the electricity sales entity and the electricity utility of the retail users. In addition, since the Markov decision process is constructed with the electricity sales entity as the intelligent body, the Markov decision process is trained based on the data for retail package electricity price decision-making and the proximal strategy optimization algorithm is used to obtain the retail package electricity price decision model, so that the retail package electricity price decision model can provide flexible decision support for the formulation of joint retail package electricity prices under uncertain environments.

[0134] Based on a general inventive concept, the present invention also provides a method for deciding a retail package electricity price. Figure 2 FIG. 1 is a flow chart of a method for deciding a retail package electricity price provided by an embodiment of the present invention. Figure 2 As shown, the retail package electricity price decision-making method includes:

[0135] Step 201: Obtain data for retail package electricity price decision-making within the decision-making day.

[0136] Step 202: Based on the retail package electricity price decision-making data, use the preset retail package electricity price decision model to generate retail package electricity price data for retail users in each preset unit time period on the decision-making day; the retail package electricity price data for retail users includes the electricity retail package price corresponding to each retail user of the electricity sales entity in the current preset unit time period; the retail package electricity price decision model is trained using the training method of the retail package electricity price decision model described in the aforementioned embodiment.

[0137] Based on a general inventive concept, the present invention also provides a training device for a retail package electricity price decision model. Figure 3 Schematic diagram of the structure of a training device for a retail package electricity price decision model provided by an embodiment of the present invention. Figure 3 As shown, the training device of the retail package electricity price decision model includes:

[0138] The training data acquisition module 31 is used to obtain data for retail package electricity price decision-making within a preset historical time period; the data for retail package electricity price decision-making includes time period type classification data, wholesale side data of the power selling entity, retail user's own characteristic data and retail package agreed electricity volume data signed between the power selling entity and each retail user.

[0139] The construction module 32 is used to construct a Markov decision process based on the revenue model of the electricity sales entity and the electricity utility model of the retail users, with the electricity sales entity as the intelligent agent.

[0140] The training module 33 is used to train the Markov decision process based on the retail package electricity price decision-making data using the proximal strategy optimization algorithm to obtain the retail package electricity price decision-making model.

[0141] Optionally, the power selling entity's profit model takes into account the power selling entity's medium- and long-term market power purchasing costs, spot market power purchasing costs, and retail market power selling revenues.

[0142] Optionally, the training device for the retail package electricity price decision model of the present invention may further include a power sales entity revenue model construction module, which is used to:

[0143] (1) Establishing a medium- and long-term market electricity purchase cost model for the electricity sales entity; the formula of the medium- and long-term market electricity purchase cost model is as follows:

[0144]

[0145] Among them, C z T is the medium- and long-term market electricity purchase cost of the electricity sales entity; f , T p and T g They are peak time, normal time and valley time respectively; P f , P p and P g are the monthly average transaction prices of the peak period, normal period and valley period of the medium- and long-term wholesale market of the electricity sales entity; I is the set of contracted retail users of the electricity sales entity; q i,t The agreed electricity consumption of the retail package of the ith retail user in the tth preset unit time period.

[0146] (2) Establishing a spot market electricity purchase cost model for the electricity sales entity; the formula of the spot market electricity purchase cost model is as follows:

[0147]

[0148] Among them, C x is the spot market electricity purchase cost of the electricity seller; T is the number of preset unit time periods included in a day; is the day-ahead market-cleared electricity quantity of the electricity sales entity in the tth preset unit time period; r,t The day-ahead settlement price of the electricity sales entity in the t-th preset unit time period; is the optimal power consumption of the ith retail user in the tth preset unit time period; P s,t It is the real-time settlement price of the electricity sales entity in the tth preset unit time period.

[0149] (3) Establishing a retail market electricity sales revenue model for the electricity sales entity; the formula of the retail market electricity sales revenue model is as follows:

[0150]

[0151] Among them, R sale The retail market electricity sales revenue of the electricity sales entity; i,t The electricity retail package price for the i-th retail user in the t-th preset unit time period is determined.

[0152] (4) Use the following formula to construct the revenue model C of the electricity sales entity. total :

[0153] C total =R sale -C z -C x .

[0154] Optionally, the retail user electricity utility model takes into account the retail package electricity price and the retail user's own characteristics.

[0155] Optionally, the training device for the retail package electricity price decision model of the present invention may further include a retail user electricity utility model building module, which is used to:

[0156] (1) Establishing the optimal electricity consumption behavior model of the retail user; the formula of the optimal electricity consumption behavior model is as follows:

[0157]

[0158] in, The optimal electricity consumption of the retail user under the current retail package electricity price; is the rigid load of the ith retail user in the tth preset unit time period; is the adjustable load of the ith retail user in the tth preset unit time period; α i is the price elasticity parameter of the ith retail user; β i is the preference level parameter of the ith retail user for electricity consumption.

[0159] (2) Based on the optimal electricity consumption behavior model, the retail user electricity consumption utility model is established. The formula of the retail user electricity consumption utility model is as follows:

[0160]

[0161] Among them, U i,t is the electricity consumption utility of the ith retail user in the tth preset unit time period; ω i is the diminishing marginal utility parameter brought by constant electricity to the i-th retail user.

[0162] Optionally, the construction module 32 may be specifically used for:

[0163] (1) Design the state space S of the agent as:

[0164] S={s t |t=1,2,...,T}

[0165]

[0166] Among them, T * The time period type used to identify the current time period, where the time period types include peak time period, normal time period, and valley time period.

[0167] (2) Design the action space A of the agent as:

[0168] A={a t |t=1,2,...,T}

[0169] a t ={p i∈I,t}

[0170] (3) Design the reward function r of the agent as:

[0171]

[0172] Among them, η is the proportional magnification factor of the electricity utility value.

[0173] Based on a general inventive concept, the present invention also provides a retail package electricity price decision-making device. Figure 4 Schematic diagram of the structure of a retail package electricity price decision device provided by an embodiment of the present invention. Figure 4 As shown, the retail package electricity price decision-making device includes:

[0174] A decision data acquisition module 41 is used to acquire retail package electricity price decision data within the decision-making day;

[0175] The decision module 42 is used to generate retail package electricity price data for retail users in each preset unit time period of the decision-making day based on the retail package electricity price decision data and using a preset retail package electricity price decision model; the retail package electricity price data for retail users includes the electricity retail package price corresponding to each retail user of the electricity sales entity in the current preset unit time period; the retail package electricity price decision model is trained using the training method of the retail package electricity price decision model described in the aforementioned embodiment of the present invention.

[0176] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0177] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0178] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0179] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0180] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0181] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0182] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0183] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0184] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A training method for a retail package electricity price decision model, characterized in that: include: Obtain data for retail package electricity price decision-making within a preset historical time period; The retail package electricity price decision-making data includes time period type division data, wholesale side data of the electricity sales entity, retail user's own characteristic data and retail package agreed electricity data between the electricity sales entity and each retail user; Based on the revenue model of the electricity sales entity and the electricity utility model of the retail users, a Markov decision process is constructed with the electricity sales entity as an intelligent agent; Based on the retail package electricity price decision-making data, the Markov decision process is trained using a proximal strategy optimization algorithm to obtain a retail package electricity price decision-making model.

2. The training method of the retail package electricity price decision model according to claim 1, characterized in that: The electricity sales entity's profit model takes into account the electricity sales entity's medium- and long-term market electricity purchase costs, spot market electricity purchase costs and retail market electricity sales revenue.

3. The training method of the retail package electricity price decision model according to claim 2 is characterized in that: The construction process of the electricity sales entity revenue model includes: A medium- and long-term market electricity purchase cost model for the electricity sales entity is established; the formula of the medium- and long-term market electricity purchase cost model is as follows: Among them, C z T is the medium- and long-term market electricity purchase cost of the electricity sales entity; f 、T p and T g They are peak time, normal time and valley time respectively; P f , P p and P g are the monthly average transaction prices of the peak period, normal period and valley period of the medium- and long-term wholesale market of the electricity sales entity; I is the set of contracted retail users of the electricity sales entity; q i,t The retail package agreed electricity consumption of the i-th retail user in the t-th preset unit time period; A spot market electricity purchase cost model for the electricity sales entity is established; the formula of the spot market electricity purchase cost model is as follows: Among them, C x is the spot market electricity purchase cost of the electricity seller; T is the number of preset unit time periods included in a day; is the day-ahead market-cleared electricity quantity of the electricity sales entity in the tth preset unit time period; r,t The day-ahead settlement price of the electricity sales entity in the t-th preset unit time period; is the optimal power consumption of the ith retail user in the tth preset unit time period; P s,t is the real-time settlement price of the electricity seller in the tth preset unit time period; A retail market electricity sales revenue model for the electricity sales entity is established; the formula of the retail market electricity sales revenue model is as follows: Among them, R sale The retail market electricity sales revenue of the electricity sales entity; i,t The electricity retail package price for the i-th retail user in the t-th preset unit time period is determined; The following formula is used to construct the revenue model C of the electricity sales entity: total : C total =R sale -C z -C x 。 4. The training method of the retail package electricity price decision model according to claim 3 is characterized in that: The retail user electricity utility model takes into account the retail package electricity price and the retail user's own characteristics.

5. The training method of the retail package electricity price decision model according to claim 4 is characterized in that: The process of constructing the retail user electricity utility model includes: Establish an optimal electricity consumption behavior model for the retail user; the formula of the optimal electricity consumption behavior model is as follows: in, The optimal electricity consumption of the retail user under the current retail package electricity price; is the rigid load of the ith retail user in the tth preset unit time period; is the adjustable load of the ith retail user in the tth preset unit time period; α i is the price elasticity parameter of the ith retail user; β i is the preference level parameter of the ith retail user for electricity consumption; Based on the optimal electricity consumption behavior model, the retail user electricity consumption utility model is established. The formula of the retail user electricity consumption utility model is as follows: Among them, U i,t is the electricity consumption utility of the ith retail user in the tth preset unit time period; ω i is the diminishing marginal utility parameter brought by constant electricity to the i-th retail user.

6. The training method of the retail package electricity price decision model according to claim 5, characterized in that: Based on the revenue model of the electricity sales entity and the electricity utility model of the retail users, the electricity sales entity is taken as an intelligent agent to construct a Markov decision process, which specifically includes: The state space S of the agent is designed as: S={s t |t=1,2,...,T} Among them, T * The time period type used to identify the current time period, wherein the time period types include peak time period, normal time period and valley time period; Design the action space A of the agent as: A={a t |t=1,2,...,T} a t ={p i∈I,t } The reward function r of the agent is designed as: Among them, η is the proportional magnification factor of the electricity utility value.

7. A method for deciding a retail package electricity price, characterized in that: include: Obtain data for retail package electricity price decision-making within the decision-making day; Based on the retail package electricity price decision-making data, a preset retail package electricity price decision model is used to generate retail package electricity price data for retail users within each preset unit time period on the day to be decided; the retail package electricity price data for retail users includes the electricity retail package price corresponding to each retail user of the electricity sales entity within the current preset unit time period; the retail package electricity price decision model is trained using the training method for the retail package electricity price decision model described in any one of claims 1 to 6.

8. A training device for a retail package electricity price decision model, characterized in that: include: A training data acquisition module, used to acquire data for retail package electricity price decision-making within a preset historical time period; The retail package electricity price decision-making data includes time period type division data, wholesale side data of the electricity sales entity, retail user's own characteristic data and retail package agreed electricity data between the electricity sales entity and each retail user; A construction module is used to construct a Markov decision process based on the revenue model of the electricity sales entity and the electricity utility model of the retail user, taking the electricity sales entity as an intelligent agent; The training module is used to train the Markov decision process based on the retail package electricity price decision-making data using a proximal strategy optimization algorithm to obtain a retail package electricity price decision-making model.

9. A retail package electricity price decision-making device, characterized in that: include: A decision data acquisition module, used to acquire data for retail package electricity price decision-making within the decision-making day; A decision module is used to generate retail package electricity price data for retail users within each preset unit time period of the decision-making day based on the retail package electricity price decision-making data and using a preset retail package electricity price decision model; the retail package electricity price data for retail users includes the electricity retail package price corresponding to each retail user of the electricity selling entity within the current preset unit time period; the retail package electricity price decision model is trained using the training method of the retail package electricity price decision model described in any one of claims 1 to 6.

Images